SATELLITE BASED HYDROLOGICAL

APPLICATIONS AND MODELING

Space Applications Centre, ISROAhmedabad-380015

India as seen from RISAT-1 (blue colour shows

mapped water bodies)

Lecture Notes on training

SATELLITE BASED HYDROLOGICAL

APPLICATIONS AND MODELING

Organized By

Earth-ecosystems Research and Training Division

VEDAS Research Group (EPSA)

Space Applications Centre (ISRO)

Ahmedabad-380058

(08-11 August, 2017)

CONTENTS

1. BASICS OF REMOTE SENSING AND APPLICATIONS

R P Singh

2. REMOTE SENSING FOR HYDROLOGICAL APPLICATIONS

R P Singh and P K Gupta

3. SATELLITE ALTIMETRY OVER LAND

S Chander

4. HYDROLOGICAL MODELLING AND REMOTE SENSING

P K Gupta

5. EVAPOTRANSPIRATION: TOOLS AND TECHNIQUES

R Pradhan

6. WATER QUALITY MONITORING FROM SPACE

A Gujrati

7. DIGITAL IMAGE PROCESSING

V B Jha

BASICS OF REMOTE SENSING AND APPLICATIONS

R. P. SINGH

Land Hydrology Division

Geosciences, Hydrology, Cryosphere Sciences and Applications Group (EPSA)

Space Applications Centre, ISRO

Ahmedabad-380015

Remote sensing technique is being currently used for providing solution to many resource

management issues as well as carry out scientific investigations to explore newer dimensions in

field of global carbon cycle, water cycle, improved weather prediction and planetary studies.

Satellite observations in different electromagnetic regions allow detection of various geophysical

parameters of the earth and planetary environment. This article reviews basic concepts involved

in remote sensing as well discusses various earth observation applications in India

Key Words: Remote Sensing, Spectral Signatures, Synthetic Aperture Radar, Passive

Microwave Radiometer, Scatterometry.

1.0 Introduction

Remote Sensing usually refers to the technology

of acquiring information about the earth's surface

(land and ocean) and atmosphere using sensors

onboard airborne (aircraft, balloons) or space

borne (satellites, space shuttles) platforms. In

remote sensing, the sensors are not in direct

contact with the objects or events being observed.

The electromagnetic radiation is normally used as

an information carrier in remote sensing.

Electromagnetic radiation is a self-propagating

wave in space with electric and magnetic

components. These components oscillate at right

angles to each other and to the direction of

propagation. Every object reflects/scatters a

portion of the electromagnetic energy incident on

it depending upon its physical properties. In

addition, objects emit radiation depending on

their temperature and emissivity. If we study the

reflectance/emittance of any object at different

wavelengths, we get a reflectance/emittance

pattern which is characteristic of that object.

Visual perception of objects is the best example

of remote sensing. We see an object by the light

reflected from the object falling on the human

eye. Modern remote sensing is an extension of

this natural phenomenon. However, apart from

visible light, the electromagnetic radiation

extending from the ultraviolet to the far infrared

(IR) and the microwave regions are also used for

remote sensing of the earth resources.

Remote sensing employs passive and/or active

sensors. Passive sensors are those, which sense

natural radiations, either emitted or reflected

from the earth. On the other hand, sensors, which

produce their own electromagnetic radiation, are

called active sensors (LIDAR, SAR). Remote

sensing can also be broadly classified as optical,

thermal and microwave. In optical remote

sensing, sensors detect solar radiation in the

visible and near infrared wavelength regions,

reflected or scattered from the earth, forming

images resembling photographs taken by a

camera located up high in the space. Thermal

remote sensing deals with detection of emitted

thermal radiation (generally in 3- 15 um spectral

range). This information is used to estimate,

temperature, humidity, cloud properties, thermal

inertia, surface mineral composition etc.

Microwave remote sensing is done by observing

passive emission or backscattered signal in 1-200

GHz spectral range.

Figure 1. Remote sensing in which solar radiation reflected from different surface features are observed by

satellite sensor.

2.0 Spectral Signature

Different land cover features, such as water, soil,

vegetation, cloud and snow reflect visible and

infrared light in different ways. The interpretation

of optical images requires the knowledge of the

spectral reflectance signatures of the various

materials (natural or man-made) covering the

surface of the earth. Any set of observable

characteristics (such as wavelength wise

variation of target reflectance) which directly or

indirectly leads to the identification of an object

and/or its condition is termed as signature.

Spatial, spectral and temporal variations are

important characteristics of target, which is used

for discrimination. Spectral signature of

vegetation is uniquely characterized by

absorption in the blue and red bands due to

chlorophyll. Generally leaf pigments in visible

region, cell structure in near infrared region

(NIR) and water content in short wave infra red

(SWIR) region are the dominant controlling

factor in leaf/canopy spectra. Infrared sensors

which measure the thermal infrared radiation

emitted from the earth help in estimation of land

or sea surface temperature. Figure 2. Wavelength

wise distribution of reflectance (spectral

signature) of different earth features

superimposed by vertical lines showing different

bands of MOS-B sensor.

Satellite sensor measures radiance which is the

radiant flux per unit solid angle leaving an

extended source in a giving direction per unit

projected source area in that direction. Unit of the

radiance is watts per meter square per micron, per

steradian (Wm-²um-1sr-¹). Observed radiances are

converted into reflectance which is the fractional

part of the incident radiation that is reflected by

the surface. Spectral reflectance is the reflectance

measured within a specific wavelength interval.

The reflection from a surface, which follows

Snell’s Law of reflection (angle of incidence =

angle of reflection, both measured from the

surface normal) is called specular reflection.

Here, the direction of the outgoing or reflected

ray is completely determined by the incoming

direction. If the angular distribution of the

reflected ray varies with the surface property and

does not follow Snell’s law then such reflection

is said to be diffuse. The reflection from a

Lambertian surface (whose intensity varies as

cosine of the angle measured from the normal to

the surface) is diffuse in nature. Bidirectional

Reflectance Distribution Function (BRDF)

describes the directional dependence of reflected

optical radiation. It characterizes the radiance

reflected into a specific view direction as a result

of the radiant flux incident upon a surface.

In optical remote sensing of the earth, the optical

sensors are looking through a layer of atmosphere

lying in between the sensors and the Earth's

surface being observed. Hence, it is essential to

understand the effects of atmosphere on the

electromagnetic radiation traveling from the

Earth to the sensor through the atmosphere. The

atmospheric constituents cause wavelength

dependent absorption and scattering of radiation.

These effects degrade the quality of images.

Some of the atmospheric effects can be corrected

before the images are subjected to further

analysis and interpretation. Absorption in the

atmosphere mostly occur when the EM radiation

interact with the atmospheric atoms or molecules

so as to excite the molecule to a higher energy

level. In this process, the incident radiation

transfers all or part of its energy to molecule.

A consequence of atmospheric absorption is that

certain wavelength bands in the electromagnetic

spectrum are strongly absorbed and effectively

blocked by the atmosphere. The wavelength

regions in the electromagnetic spectrum usable

for remote sensing are determined by their ability

to penetrate the atmosphere. These regions are

known as atmospheric transmission windows.

Remote sensing systems are often designed to

operate within one or more of the atmospheric

windows. Atmospheric molecules are

responsible for selective absorption in different

wavelength.

Even in the regions of atmospheric windows, the

scattering by the atmospheric molecules and

aerosols produces spatial redistribution of energy.

The scattered / diffused radiance entering the

field of view of a remote sensor, other than that

from the target of interest, is called path radiance.

Scattering is a multiple reflection of

electromagnetic waves by particles or surfaces.

Energy is not lost to the medium but the radiation

is scattered out to other directions, thereby

reducing the amount of radiation in the original

direction. The sum total of absorption and

scattering is known as attenuation. Broadly there

are three type of scattering process in atmosphere.

1) Molecular or Rayleigh Scattering - This occurs

when the particles causing the scattering are

smaller in size than the wavelengths of radiation

in contact with them. This type of scattering is

therefore wavelength dependent. As the

wavelength decreases, the amount of scattering

increases. It is the Rayleigh scattering that is

responsible for the sky appearing blue.

2) Particle or Mie Scattering - Mie scattering is

caused by pollen, dust, smoke, water droplets,

and other particles in the lower portion of the

atmosphere. It occurs when the size of particles

causing the scattering are similar or slightly

larger than the wavelengths of radiation in

contact with them. Turbid appearance of sky is

due to Mie scattering caused by suspended

aerosols.

3) Non-selective Scattering - It occurs in the

lower portion of the atmosphere when the

particles are much larger than the incident

radiation. This type of scattering is not

wavelength dependent. Scattering of optical light

in cloud is associated with non selective

scattering.

Figure 3 Atmospheric transmittance due to absorption of different atmospheric molecules.

3.0 Thermal Remote Sensing

Any object above absolute zero temperature

emits electromagnetic radiation. Thus the objects

we see around, including ourselves are thermal

radiators. An ideal substance is called blackbody

which absorbs the entire radiant energy incident

on it and emits radiant energy at the maximum

possible rate per unit area at each wavelength for

any given temperature. No actual substance is a

true blackbody, although some substances

approach its properties. The radiance being

emitted by a blackbody at given wavelength ( )

and Temperature T is given by Planck’s

Radiation Law.

Mλ = 2лhc2

λ5 [exp(hc/λkT)-1]

Where k Boltzmann’s constant, h is Planck’s

constant, c is velocity of light, and T is the

absolute temperature in Kelvin.

A black body is an ideal surface such that

1. It absorbs all incident radiation

regardless of the wavelength or direction of the

incident radiation;

2. For a given temperature and wavelength,

nobody can emit more energy than a black body;

3. Emission from a black body is

independent of direction, that is, the black body

is a diffuse emitter.

The total emission within all the wavelengths

Mtotal can be found out by integrating the Planck’s

equation from λ = 0 to λ = ∞ and works out to be

Mtotal = σT4 Wm-2

Where σ is Stefan –Boltzman constant.

Another useful expression in thermal remote

sensing is Wien’s Displacement Law, which

gives the wavelength λmax at which the exitance is

maximum and is related to the temperature as

λmax T = constant

if λmax is expressed in micrometer and T in 0K,

then the constant is 2897.

The heat energy is converted to radiation at a

maximum rate as per Planck’s law. However, a

real surface does not emit at this maximum rate.

The emission from a real surface is characterized

with respect to a black body. In order to do so, a

term called emissivity is used which compares,

the ‘radiating capability’ of a surface to that of a

black body (an ideal radiator).

Figure 4. The Planck’s radiation distribution of blackbody at different temperatures

Emissivity () defined as the ratio of radiant

exitance of the material of interest (Mm) to the

radiant exitance of a black body (Mb) at the same

temperature.

= Mm / Mb

For a black body, = 1, for all wavelengths. For

a gray body < 1, and can vary with wavelength

and direction.

Brightness temperature (TB) of the surface is the

temperature of a blackbody surface which, when

placed in front of the receiver aperture, would

produce the same received flux within the

spectral band of receiver. Brightness temperature

(TB) is defined a

TB = T

where is emissivity of the target and the T is

absolute physical temperature.

Figure 5. Observations of thermal emission on Martian surface showing (a) radiance in

W/cm2/um/sr, (b) Surface temperature in K and (c) surface Emissivity estimated from THEMIS

observations from band 3 (7.93 um).

4.0 Microwave Remote Sensing

Microwave remote sensing is highly useful as it

provides the observations of earth‘s surface

regardless of day/night and atmospheric

conditions. Microwave remote sensing makes use

frequency range from 1 to 300 GHz of the

spectrum. The electromagnetic waves in this

range are relatively less affected by the

atmosphere and hence provide useful data in

overcast or turbid environment. The active

sensors in microwave consist of transmitter and

receiver. Scatterometers, Synthetic Aperture

Radars (SAR) and altimeters are some of the

examples of active microwave sensors. The

transmitted energy is reflected and /or scattered

from the target. The signal with a propagation

delay is received and processed to deduce and

understand the target properties. Radar equation

expresses the fundamental relationship between

radar parameters, target characteristics and the

received signal. For monostatic radars, it is given

by

𝑃𝑟 =𝜆2

(4𝜋)3∫𝑃𝑡𝐺

2𝜎0

𝑅4𝑑𝐴

Where Pr is the average power returned to the

radar antenna from the extended target, Pt is the

power transmitted by radar, G is the gain of the

antenna, R is the distance of the antenna from the

target, λ is the wavelength of the radar, o is the

radar scattering coefficient of the target. The

integration is over the illuminated area A. The

backscattering coefficient is defined as the ratio

of the energy received by the sensor, over the

energy that the sensor would have received if the

surface scattered the energy incident upon it in

isotropic fashion. It is represented as Decibels

(dB).

In the active microwave remote sensing,

information about the object’s physical structure

and electrical property is retrieved by analyzing

the backscattering The microwave signature of

the object are governed by sensor parameters

(frequency, polarization, incidence angle) and

physical (surface roughness, feature orientation)

and electrical (dielectric constant) property of the

target. A given surface may appear very rough at

higher frequency compared to lower frequency.

Generally backscattering coefficient increases

with increasing frequency. In addition, the signal

penetration depth increases with wavelength in

microwave region. Use of multi- frequency

allows distinction between roughness types. The

backscattering also depends on the polarization of

(a) (b) (c)

the incident wave. A vegetation canopy

consisting of short vertical scatter over a rough

surface can be considered as short vertical

dipoles. In such case, vertically polarized incident

wave interact strongly with canopy. The multiple

scattering and volume scattering from a complex

surface, such as forest cause depolarization. The

radar backscattering coefficient from a terrain is

strongly dependent on angle of incidence. The

angular dependency of backscattering coefficient

is primarily due to surface roughness. The surface

water extent during flood is detectable on radar

backscatter image due to high contrast between

smooth water and rough land surface.

SAR interferometry is an extremely powerful tool

for mapping the Earth’s land, ice and even the sea

surface topography. The basic idea is that the

position of a point on the Earth’s surface can be

reconstructed from the phase difference

(interferogram) between two complex-valued

SAR images achieved by coherently processing

the backscattered signals (phase) recorded by the

two antennas.

A Passive sensor consists of only a receiver.

Emitted radiation from manmade or natural

targets is received and processed by the

radiometer to infer the target properties. Raleigh-

Jeans law describe the spectral radiance Mλ(T)

from a black body in microwave at a given

temperature through classical arguments. At

microwave frequencies, Planck’s equation gets

approximated to Raleigh-Jeans law as

Mλ = 2лckT

λ4

where K is Boltzmann constant, is emissivity of

the body at absolute temperature T and

wavelength λ. Emission from a body in

microwave region at particular wavelength is

proportional to the brightness temperature.

Brightness temperature observed over a region

which is product of physical temperature and

surface emissivity is used to infer many

geophysical properties.

Figure 6. Spatial variability of different ecosystem as observed by IRS-LISS-III data

Figure 7. Supervised classification of different land cover classes in parts of Madhya Pradesh, India.

5.0 Remote Sensing Applications

The output of a remote sensing system is

usually an image representing the scene being

observed. Image analysis and modeling is used

in order to extract useful information from the

image. Remote sensing images are normally in

the form of digital images. There are many

image analysis techniques (image

transformation, enhancements, pattern

recognition, fusion, merging etc) available for

analysis the data. Suitable techniques are

adopted for a given area and land cover

characteristics, depending on the requirements

of the specific problem.

Identification of terrain categories is done by

digital processing of data acquired by

multispectral scanners. Classification is a process

of assigning individual pixels of an image to

categories, generally on the basis of spectral

seperability analysis. Classification is generally

carried out by supervised or unsupervised

technique. Supervised Classification is digital-

information extraction technique in which the

operator provides training-site information that

the computer uses to assign pixel to categories. It

generates the decision rule and assigns the classes

accordingly, Unsupervised classification is

digital information extraction technique in which

the computer assigns pixels to categories through

clustering techniques without a priori field

information of classes.

Remote sensing help mapping, monitoring and

management of various resources like

agriculture, forestry, geology, wetlands, ocean

etc. It further enables monitoring of environment

and thereby helping in conservation. In the last

four decades it has grown as a major tool for

collecting information on almost every aspect on

the earth. Some of the important projects carried

out in the country include Groundwater Prospects

Mapping under Drinking Water Mission,

Forecasting Agricultural output using Space,

Agrometeorology and Land based observations

(FASAL), Forest Cover/Type Mapping, Wetland

Mapping, Biodiversity Characterization, Snow &

Glacier Studies, Land Use/Cover mapping,

Coastal Studies, Coral and Mangroves Studies,

Wasteland Mapping etc. The information

generated by large number of projects is used by

various departments, industries and others for

different purposes like development planning,

monitoring, conservation etc.

Forest Crop WL/Fallow Water SandForest Crop WL/Fallow Water Sand

REMOTE SENSING FOR HYDROLOGICAL APPLICATIONS

R P SINGH and P K GUPTA

Land Hydrology Division

Geosciences, Hydrology, Cryosphere Sciences and Applications Group (EPSA)

Space Applications Centre, ISRO

Ahmedabad-380015

Although 70 percent of the Earth’s surface is covered with water, the amount of fresh water

available on land surfaces is a tiny fraction of the total. Less than 1 percent is fresh water, present

in the form of groundwater, soil moisture and, river/lakes on the land surface for domestic,

agriculture, aquatic and other purposes. Water demand is growing by twice of the population

growth. One third of the population would be under water stress by 2025 and 2/3 of population

would be under the water scarcity by 2050. Therefore, there is a need for global, satellite-based

observations of terrestrial surface waters to diagnose where is water stored in the Earth’s land

surfaces, and how does this storage vary in space and time? Satellite remote sensing provides a

means of observing repetitive and continuous coverage of the Earth surface and atmosphere with

high spatial resolutions. Many earth observation satellites are in use; polar satellites which turn

around the poles, and geostationary satellites which have a fixed position with respect to the earth

surface. A series of satellites are being used for observing hydrological state variables for the

assessment and management of water resources. Such variables are rainfall from

microwave/thermal data, river/reservoir/lakes water levels from altimeter and scatterometer,

surface soil moisture from active/passive microwave data, flood inundation areas from

microwave, wetland areas using optical data, irrigated areas from optical data, groundwater

prospects using optical/microwave data etc. Different hydrological variables namely rainfall,

runoff, wetlands, groundwater and soil moisture has been discussed.

Key Words: Satellite Remote Sensing, Hydrology, Rainfall, Runoff, Soil Moisture, Irrigation

class, Water level, Groundwater, Microwave, Altimeter, Scatterometry.

1.0 Introduction

The average annual rainfall including snowfall in

India is 4000 Billion Cubic Meters (Rakesh

Kumar 2005) but availability of per capita fresh

water is major concern as the population continue

to increase in future (R K Mall 2006). Water

security and management requires information

ranging from regular inventory of surface water

bodies to assessment of rainfall, soil moisture,

evapotranspiration, ground water and snow melt

runoff (V.V. Rao 2013). Hydrological drought

and flood forecast during extreme weather events

become important for planning and water

resource management.

Modeling different components of water cycle

requires measurement of water which can exist in

all the three phases of matter i.e. solid, liquid and

gas (R.O. Green 2006). Satellite provides an

important platform from where measurements

can be done in any part of electromagnetic

spectrum suitable to detect different phases of

water over a large region not feasible through

sparse network of surface-based instruments.

Instruments on Earth Observation systems have

been extensively used to measure hydrologic and

hydraulic variables such as water spread area,

elevation of water surface (h), its slope (∂h/∂x)

and temporal changes (∂h/∂t) (E A Douglas,

2007). Seasonal monitoring of reservoir spread,

conjunctive water uses and water management

practices associated to transplanting operations

helps in irrigation scheduling (P K Gupta 2009

and 2010). Satellite observations are regularly

used to generate snow cover map which help in

snow melt runoff estimation (S.K. Jain 2010).

Satellite data has been used to study the changes

in the extent of Himalayan glaciers inventory and

their monitoring in terms of whether they are

retreating or being stable over the time is another

important contribution of space technology

towards understanding the climate change signals

(I.M. Bahuguna 2014). To provide safe drinking

water to hundred thousands of villages, ground

water prospect maps were generated using

satellite data conjunctively with ground

information showing probable regions where

wells can be drilled (R.K. Jaiswal 2003).

Groundwater prospects maps at 1:50,000 scale

using IRS-1C/ 1D LISS-III data for entire country

was generated under the project titled "Rajiv

Gandhi National Drinking Water Mission".

Remote sensing data along with Geospatial

technology have helped planning of water

resources by the respective water management

boards.

Hydrological remote sensing is carried out using

measurements from various Indian satellite

platforms such as SARAL-Altika Mission

(R.M.Gairola 2015) (Inland Water level),

RISAT-1 SAR Mission (T. Misra 2013) (Surface

water spread, Soil Moisture), Resourcesat-1/2

Missions (M.R. Pandya 2007) (Snow cover,

Wetlands, Land use Land cover, Water quality),

Cartosat Missions (DEM), Kalpana, Megha

Tropiques and INSAT-3D Missions (Rainfall,

Solar Radiation etc.) (R. R. Navalgund 2010).

Global Missions such as Landsat Program,

Sentinel Program, Jason program,

SRTM/ASTER topography missions, MODIS

instruments on Earth Observation Terra and Aqua

Missions, GRACE Mission (S. Swenson 2008),

Soil Moisture and Ocean Salinity (SMOS) (Y.H.

Kerr 2001) Mission, Soil Moisture Active

Passive (SMAP) (D. Entekhabi 2010)Mission

and Tropical Rainfall Measuring Mission

(TRMM) etc. also provide valuable datasets to

model the water fluxes over India. Scientific

rationale of section of various hydrological

parameters using satellite data is discussed in

subsequent sections.

2.0 Physical Basis for Remote Sensing in

Hydrology

Remote sensing generally refers to the

technology of observations of earth using sensors

placed on aircraft or satellites platforms. These

sensors employ active as well passive sensing

system. Active systems have their own source of

illumination (Radar, Scatterometer, Altimeter)

whereas passive systems sense natural radiations,

either reflected or emitted from the earth. In

active remote sensing, generally instruments

(Radar, Scatterometer) measure back scattered

signals from target however Altimeters use the

time delay in propagation of the incident signal to

infer the topography (elevation, water level) of

the surface. Short pulse of electromagnetic wave

is transmitted and received by Radar from space

platform and range between the satellite and earth

surface is measured by precise orbit computation

and correcting atmospheric and geophysical

signals.

In the active microwave remote sensing,

information about the object’s physical structure

and electrical property is retrieved by analyzing

the backscattering signal. Microwave remote

sensing provides the observations of earth‘s

hydrological variables regardless of day/night

and atmospheric conditions. Water being a polar

molecule has very high sensitivity in microwave

wavelengths due to orientation polarization

property. The electromagnetic radiation is used as

an information carrier in remote sensing.

Instruments operate in optical (Cameras, Spectro-

radiometer), infrared (Thermal radiometers) and

microwave (Radar, Altimeter, Scatterometer etc.)

wavelength ranges. Microwave remote sensing

provides the observations of earth‘s hydrological

cycle regardless of day/night and atmospheric

conditions. Water being a polar molecule has

very high sensitivity in microwave wavelengths

due to orientation polarization property. The

microwave signature of the object is governed by

sensor parameters (frequency, polarization,

incidence angle) and physical (surface roughness,

feature orientation) and electrical (dielectric

constant) property of the target (H.S. Srivastava

2009). The surface water extent is detectable on

radar backscatter image due to high contrast

between smooth water and rough land surface.

Radar operated in interferometry mode helps in

mapping the surface topography. Radar

interferometry works on the principle that the

position of a point on the Earth’s surface can be

reconstructed from the phase difference

(interferogram) between two complex-valued

SAR images achieved by coherently processing

the backscattered signals (phase) recorded by the

two antennas. Shuttle Radar Terrain Mapper

(SRTM) used interferometric technique to derive

global digital elevation model (DEM). DEM

information when coupled with other remote

sensing data and terrain models provide valuable

inputs in hydrology.

A Passive sensor consists of only a receiver.

Emitted radiation from manmade or natural

targets is received and processed by the

radiometer to infer the target properties. Any

object above absolute zero temperature emits

electromagnetic radiation. Thus the objects we

see around, including ourselves are thermal

radiators. An ideal substance is called blackbody

which absorbs the entire radiant energy incident

on it and emits radiant energy at the maximum

possible rate per unit area at each wavelength for

any given temperature. Brightness temperature

observed over a region which is product of

physical temperature and surface emissivity is

used to infer many geophysical properties.

Brightness temperature are lower when measured

over moist surface as compared to brightness

temperature observed over dry surface.

Modeling of multi frequency vertical as well as

horizontal polarization brightness temperature

(from SSMI, AMSR-E type of radiometers) with

respect to varying soil moisture content and other

surface characteristics is carried out to retrieve

the surface soil moisture (R.P.Singh 2005).

3. Techniques for Estimation of Hydrological

variables

3.1 Rainfall

Rainfall can be estimated using combination of

optical and thermal infrared data and also using

the microwave data.

3.1.1 Optical-infrared approach

Existing techniques have concentrated on using

visible and thermal infrared channels for cloud-

top reflectance and temperature. In the visible

band, it is possible to determine the type of cloud

cover based on the textural characteristics of the

cloud within the satellite image. The physical

basis for this method is that high cloud top

brightness in the visible bands means that there is

a greater probability of rain due to large cloud

thickness. Similarly, low cloud top temperature in

the thermal band implies high cloud tops and also

implies a correspondingly large cloud thickness.

Therefore, precipitating clouds can be

distinguished from others on the basis of their

brightness or infrared temperature characteristics.

A strong quantitative relationship exists between

rainfall amount and calibrated cold cloud

duration derived from thermal infrared data.

Precipitation intensity is directly proportional to

the area of the upper surface of the cloud at

temperature of less than –15 0C. Decreases in

brightness temperature is directly correlated to

cumulative rainfall amounts.

Satellite data: Kalpana-1, INSAT-3A, MODIS,

ASTER, NOAA etc.

3.1.2 Microwave approach

Passive microwave remote sensing and active

microwave remote sensing are being used to

estimate rainfall. This is largely because clouds

are relatively transparent at microwave

wavelengths but hydrometeors are not, resulting

in a relatively direct link between hydrometeors

and attenuation of upwelling microwave

radiation. Radar measurements of power of

electromagnetic waves backscattered by

raindrops are directly related to a physical

quantity called reflectivity. Estimation of rainfall

amounts involves using reflectivity via a

reflectivity-rainfall relationship. How is the

relationship selected, there are two approaches. In

the first approach, which we will term the drop

size distribution approach, relations are derived

from raindrop size distribution observations

made at the surface. The second approach is

similar in relying on statistical estimation

procedures to relate measured values of radar

reflectivity to rainfall rate.

Satellite data: SSM/I, TMI, Tropical Rainfall

Measurement Mission Precipitation Radar

Satellite Products: NOAA CPC, JAXA, TRMM,

METEOSAT, INSAT etc.

3.2 Surface Runoff

What is the global distribution of runoff water

delivered to the oceans and what is its inter-

seasonal and inter-annual variability? Runoff is

an integrated effect of rainfall, topography, soil

and land cover conditions. Efforts are being made

to derive the runoff directly using remote sensing

technique such as combined use of altimeters and

microwave data. Presently remote sensing

technology has been used to derive very crucial

inputs to runoff modeling. Remote sensing

derived precipitation, topography, LULC and

land surface characteristics are used for the

computation of runoff using simple rainfall-

runoff simulation approach;

3.2.1 Altimetry

River water levels estimated using altimeter data

(SARAL-ALTIKA, Jason-2) can be used for the

validation of the model results. Water level is

important hydrological quantity required to

budget the fresh water availability. Satellite

altimetry is important active remote sensing

technique for systematic monitoring of water

levels of reservoirs, lakes and rivers. Satellite

altimetry technique was originally started for

assessment of ocean topography but recent

instruments on JASON, SARAL Altika, Jason-3,

Sentinel-3 and future missions such as Surface

Water and Ocean Topography (SWOT) mission

are designed to study the inland water bodies also.

Radars onboard satellite emit pulses towards

nadir and receive the echo by water surface. The

half time span for pulse reflected back to mission

corresponds to distance ( 𝜌 ) between satellite and

earth surface. The height H of the reflecting water

body with reference to geodetic reference is given

as

𝐻 = 𝑎s − 𝜌 + 𝐶iono + 𝐶dry + 𝐶wst + 𝐶st + 𝐶pt (4)

Where as is the satellite altitude with reference to

reference ellipsoid. Other terms reference to

corrections related with delayed propagation

through the atmosphere (Cdry and Cwet), the

interaction with ionosphere (Ciono) and solid earth

tides (Cst and Cpt ).

Radar altimeter onboard ISRO/CNES SARAL –

AltiKa mission provides important information

of water level for rivers and large reservoirs at 35-

day repeat interval (P.K.Gupta 2015) (A.K.

Dubey 2015) (F. Papa 2012). Observations of

water level in Ukai Reservoir from 2013 to 2016

shows lowest water level condition in 2016

indicating hydrological drought situation.

Analysis showed that amount of water volume

availability was found to be less by19.8% in the

current year (April 2016) as compared to last year

(April2015) whereas it is 88.5% less in

comparison to 2014 for the same time frame over

the Ukai Reservoir. A flood wave of 5.93 m,

which is highest since inception of SARAL-

Altika (Feb 2013) between two passes (4th May to

9th June 2016) over Brahmaputra was estimated.

Fig. Water Level Retrieval using Satellite

Altimetry over Ukai reservoir, India

Satellite altimetry (A. K. Dubey 2014) has been

used to study the river stage and its discharge

using rating curve relationship. Papa et al. (2010)

(F.Papa 2010) estimated monthly discharges

from Ganga and Brahmaputra rivers using

TOPEX-Poseidon (T/P), ERS-2 and ENVISAT

satellite altimetry data. Biancamaria et al. (2011)

(S. Biancamaria 2011) also studied water levels

at upstream locations in India using T/P altimetry

data to forecast the water levels of Ganges and

Brahmaputra rivers. Frappart et al. (2005) (F.

Frappart 2005) have determined spatio temporal

variations of water volume in Negro River basin

using area variations from SAR data and changes

in water level from T/P altimetry data.

3.2.2 Scatterometery

A Scatterometer is an active microwave

instrument (radar) actively transmit

electromagnetic pulses to the Earth's surface and

measure the backscatter response/power of the

return pulse scattered back to the antenna.

Scatterometers average the detected returns from

a sequence of pulses, a process known as

postdetection integration often achieving ± 0.10

to 0.15 dB accuracy. calibrated to less than a few

tenths of a decibel; ample to capture inter-

seasonal diff. with 1 to 2 dB changes.

Satellite based remote sensing of river

hydrodynamics is important application in

hydrology. Basin level rainfall and associated soil

wetness influences the fluctuations in river water

levels through process of surface runoff in the

downstream (Wagner 1999). Fig. present water

level at gauging site Dhubri (downstream of

river) using OSCAT/SCATSAT-1 (high

resolution SIR datasets) data for the year 2013.

Fig. Scatterometer (blue line, every 15 days) and

altimeter (red line, every 35 days) estimated

water levels for Dhubri gauging site for the year

2013.

88

90

92

94

96

98

100

102

104

0 30 60 90 120 150 180

Wa

ter

leve

l fr

om

msl

, m

Julian Days

Multi-year water level fluctuation in Ukai reservoir using

SARAL-Altika

2013

2014

2015

2016

25

26

27

28

29

30

31

32

33

22-M

ay

26

-Ju

n

31

-Ju

l

04

-Se

p

09

-Oct

13-N

ov

18

-De

c

22

-Ja

n

Wat

er

leve

ls, m

Altimeter pass dates

Altimeter_scatterometer Altimeter_linear fit

3.3 Wetlands

Wetland identification, mapping and analyses are

often done by means of remote sensing. Remote

sensing of wetlands may be studied using a range

of sensors such as panchromatic or colour-

infrared aerial sensors, broadband multi-spectral

to narrowband hyperspectral besides microwave

sensors. The mapping and inventory is the first

step in the conservation and management of

wetlands. Several studies appeared from time-to-

time using fine resolution (5 to 100 m) multi-

spectral satellite data like Landsat-MSS (Multi

Spectral Satellites), Thematic Mapper, Enhanced

TM; SPOT; Linear Imaging Self Scanning (LISS-

I, LISS-II, LISS-III) on IRS series of satellites;

AVNIR on ADEOS satellite for mapping and

classification of wetlands. Data from Advanced

Very High Resolution Radiometer (AVHRR) on-

board NOAA is currently available operational

representative coarse resolution system besides

Moderate Resolution Imaging Spectro-

radiometer (MODIS) on-board Terra and Aqua

satellites was used for global wetland mapping

and estimation of aerial extent. The importance

of this has been recognized over the last forty

years and various governmental agencies have

laid emphasis on mapping the spatial

extents/distribution of wetlands to feed towards

the functional characterization like hydrologic,

biogeochemical and maintenance of habitat and

food webs. Space applications Centre has done

National level wetland mapping comprising of

natural and manmade water bodies covering

inland and coastal areas using RESOURESAT-1

LISS-III data at 1:50000 scales. A total of 201503

large wetlands (> 2.25 ha) and 555557 small

wetlands (< 2.25 ha) covering an area of 15.26

Mha have been mapped and classified.

Synthetic Aperture Radar (SAR) which is an

active sensor onboard Microwave Satellites

(RISAT-1; Mishra et al., 2013), ERS, Radarsat

and Envisat) has many characteristics that make

it useful in water spread (land-water demarcation)

mapping and monitoring activities over time

because of its ability to image even during severe

weather conditions and day/night acquisition

capability (Lee K S, 2003; Brisco et al., 2008).

But, estimation of flood extent varies with the

type of polarization since the intensity of the

radar-backscattered signal depends upon the

wavelength, incident angle and polarization of

the signal and the geometric

(structure/composition) and electrical (dielectric

constant) properties of the ground features it

impends upon. Land-water demarcation using

HH polarization data has better capability as

compared to HV polarization. Fig. shows water

spread area of Kamleshwar Dam in Gir forest

region of Gujarat India using multi-date RISAT-

1 MRS datasets for 2015.

Fig. Water spread using multi-date RISAT-1

dataset over Kamleshwar Dam in Gir forest

regions of Gujarat. Different colors show water

spread change from April to October 2015.

3.4 Groundwater

Groundwater potential zones and its short term

depletion/surplus areas can be identified using the

remote sensing technology.

3.4.1 Groundwater potential zone

April

July

September

October

The occurrence and movement of groundwater in

an area is mainly controlled by secondary

porosity caused by fracturing of the underlying

rocks. Lithology, geomorphology, fractures,

lineament, soil, landuse, drainage and slope all

play an important role in groundwater

replenishment. High relief, steep slopes and high

drainage density impart higher runoff causing

less infiltration, while low relief, gentle slope and

low drainage density result in low runoff and

comparatively high infiltration. Surface water

bodies, such as rivers, ponds and canals, can act

as recharge zones enhancing the groundwater

potential in the vicinity. Therefore, identification

delineation and quantization of surface

hydrological features are vital for groundwater

potential zones investigation.

Remote sensing technology has been widely used

for groundwater resource management. Satellite

data (LISS-III, AWiFS, RISAT, CARTO_DEM)

are useful for extracting various hydro-geological

thematic maps required for groundwater

assessment. These thematic maps are assigned

suitable weights and different rankings to the

individual classes within each thematic map. The

integrated thematic maps can be used to compute

the Groundwater Potential Index (GWPI). Very

poor GWPI zones show low water yields (0.5- 1.5

lps) while excellent GWPI zones show high water

yields (5–7 lps). The results can be subsequently

cross-checked with resistivity survey. Space

applications centre has carried out hydro-

geological studies in varied geological setups

under the Rajiv Gandhi drinking water mission.

3.4.2 Groundwater from gravity anomaly

Satellite-based observational technique allows us

to directly monitor regional changes in stored

water. It allows us to produce an up-to-date

quantitative estimate of the temporal and spatial

variability of groundwater in the region, which is

a first step towards management of sustainable

water resources for the populated places on the

globe. The Gravity Recovery and Climate

Experiment (GRACE) satellite mission,

measures temporal variations in the gravity field

due to mass fluctuations which GRACE record

remarkably from the space. Withdrawals for

irrigation and other uses are depleting the

groundwater reserves of Rajasthan, Punjab and

Haryana at a rate of 4.0 ± 1.0 cm yr-1 equivalent

height of water, or 17.7 ± 4.5 km3 yr-1, recorded

by GRACE (Tiwari 2009).

3.5 Soil moisture

Hydrological variable soil moisture at the

regional scales is very dynamic and changes

conditions rapidly. Satellite based sensors offer

the advantages of large area mapping and long-

term repetitive coverage. Active and passive

remote sensing techniques are used for the

estimation of soil moisture.

3.5.1 Active remote sensing

Near surface (0-5 cm) soil moisture can be

estimated at microwave frequencies. The most

important feature is the large contrast in

emissivity between water and land. This is due to

the large dielectric constant of water (80) while

that of most dry minerals or soils is <5 and for

mixture of soil water is 30-40 which constitute a

base for observing emissivity and estimating soil

moisture through remote sensing at microwave

frequencies. It has been found that the longer

wavelengths are better for increased sampling

depth and reduced effects of noise factors such as

vegetation and surface roughness. The range of

emissivity variation to be expected is from 0.95

for dry soils to less than 0.6 for smooth wet soils.

The main factors which affect the accuracy are

vegetation cover (25%) and roughness (15%).

Example RADARSAT (Radio Detection and

Ranging Satellite), Radar Imaging satellite

(RISAT) etc.

3.5.2 Passive remote sensing

Passive microwave methods measure the natural

thermal emission of the land surface using very

sensitive detectors, the intensity of this emission

is generally expressed as a brightness

temperature. The brightness temperature of the

surface is related to its emissivity, physical

temperature and contributions from the

intervening atmosphere. Microwave emission is

mainly influenced by the dielectric and roughness

properties of the surface. As the fractional

amount of water increases, the emissivity

decreases and the slope of the emissivity between

two or more frequencies can be used to determine

the surface wetness condition. Examples Special

Sensor Microwave/Imager (SSM/I), Advanced

Microwave Scanning Radiometer (AMSR), Soil

Moisture and Ocean Salinity (SMOS) etc.

A statistical approach as well as a forward model

based inversion technique is mainly used in the

estimation of soil moisture. Statistical techniques

rely on regression analysis between measured

backscattering coefficient/brightness temperature

and surface soil moisture. The slope and intercept

of the regression line are dependent on land cover

variables, which can be estimated from ancillary

data. In the forward model based inversion

technique, a model is used to simulate remotely

sensed output signal (backscattering coeff.,

brightness temperature) on the basis of input land

surface characteristics (e.g. soil moisture, soil

texture, roughness, vegetation water content).

Inversion methods based on iterative

minimization between forward model simulation

and observation are used to estimate the soil

moisture. The statistical approaches are simpler

to use in comparison to the physical approaches

but require a region specific coefficient as

microwave scattering/emissions are determined

by the soil (texture, roughness) and vegetation

characteristics. Based on the sensor involved in

the soil moisture estimation, remote sensing

techniques can be categorized in broad three

categories (1) Radar Based Technique, (2)

Scatterometer based Techniques and (3) Passive

Microwave Radiometer based Techniques.

3.5.2.1 Shifting Irrigation Practices

Over exploitation of ground water in the recent

past is well known fact in the Punjab and Haryana

region and has been reported by several studies

using the satellite based gravity anomaly from

GRACE mission and also using observed data.

This decline in groundwater has enforced

“Punajb Sub-soil water act in 2009” by the

Punjab Govt. This water act has changed

irrigation practices in the region. Passive

microwave radiometer (AMSR-E) soil moisture

data from 15 March to 30 September was

analyzed during 2002 to 2011 period.

It has been found that there is gradual shift in the

early soil wetness pattern and associated change

in the irrigation practices. Shifting is delayed by

14 ± 1 days in central Punjab from 167 Julian day

to 181 Julian day from pre to post “water act”

period, respectively. Fig. presents the shifting

irrigation practices over Punjab and Haryana

regions. in Julian days for pre and post water act

time periods.

Fig. presents the shifting irrigation practices over

Punjab and Haryana regions. in Julian days for

pre and post water act time periods.

Legend

state_bnd_geo

ndvi_max2008_subset.tif

Value

High : 0.9

Low : 0.5

sm_maxdiff10_subset.tif

Value

High : 300

Low : 50

day_maxdiff10subset.tif

Value

High : 30

Low : 0

195

175

155

135

Julian Days

Pre “water Act”

Post “water Act”

3.6 Conjunctive water uses

Improving the water use efficiency in irrigated

systems integrating satellite inputs is the thrust

area of research. Hence, conjunctive use of water

from different sources i.e., precipitation, canal

water, groundwater, surface water etc., is gaining

attention. Thenkabail et al. (2005) used near

continuous time series remote sensing moderate

resolution imaging spectrometer (MODIS) data

to map the irrigated areas of the Indo-Gangetic

region. Global irrigation mapping effort was

undertaken by Siebert et al. 2007. Multi-date high

temporal IRS-1D data have been used to identify

the various sources of irrigation such as surface

water, groundwater, wetland etc. by analyzing the

space time spectral curves over the Damodar

command area in West Bengal (P K Gupta 2009).

Fig. Conjunctive water use (irrigation

class/source) pattern (GW: groundwater, SW:

surface water) in Damodar command area using

multi-date IRS-1D data.

3.7 River morphologic features

Understanding of river and floodplain interaction

is very limited. Generally, physical scaled models

are used to investigate the interaction. These

physical models are unable to present the

morphological change and unsteady effect

accurately. This river dynamics (change in the

river morphological features) can be identified

using the microwave remote sensing techniques.

River models integrating the remote sensing

inputs will provide us depth of water and

duration. A decision rule based classifier was

developed to delineate the river morphological

features using the multi-date RISAT-1 MRS HH

polarized (gives better discrimination as

compared to VV polarization) data during July to

September 2012.

Fig. River morphological features delineated

using multi-date RISAT-1 MRS HH data over

Brahmaputra river reach nearby Guwahati.

4.0 Conclusions and Future Directions

Present trend in remote sensing of hydrology is to

develop methodologies for retrieval of various

hydro-meteorological parameters from satellite

data and assimilate the information in physically

based distributed hydrological models.

Integrating remote sensing derived inputs such as

DEM, rainfall, crop growth parameters (Leaf area

index), land use land cover, wetlands, soil

moisture, snow cover area, vegetation indices,

river network, cross sections, river water level

fluctuations, insolation, Albedo etc. into semi-

distributed and distributed hydrological models is

a challenge. As volume of input data and

computational requirements are enormous, use of

these models has not yet reached the operational

status in many developing countries. Future need

in water management is to develop and assess the

use of remote sensing methodologies, in

combination with in situ data and hydrological

Command boundary

Groundwater

Canal water

Wetland

Non crop

Water

Canal network

Turbulent Water

Calm WaterIsland Water

FloodplainInundation

Island Vegetation

Sand Bar

Others

models. In the present scenario hydrological

research is oriented towards addressing the

particular process instead there is need to use

hydrological system approach to study the

integrated effect of hydrology and eco-system

processes for updating predicted hydrologic state

variables. Study scale needs to be up scaled from

basin to global nature to account for the

atmospheric phenomenon and hydro-

teleconnections. Development and testing of

remote sensing algorithms and generation of

environmental forcing fields needed to drive the

hydrologic model for various emerging

applications such as eco-hydrology, monitoring

of reservoirs/lakes storage from satellite altimetry

etc. which can lead to generate operational

hydrological products.

In future, there is requirements to improve the

assessment of water level, soil moisture,

bathymetry, Rainfall and evapotranspiration.

Present capability of repetivity (10-35 days) for

water level retrieval needs to be improved to

daily. Availability of soil moisture at 10 km need

to be improved to 1 km with capability of soil

moisture profiling. It is required to have LIDAR

measurements for river bathymetry and submeter

level DEM with vertical accuracy of 0.5m for

flood inundation studies. Present capability to

measure rainfall need to be improved to 1km

spatial resolution. Hydrological modeling need

simultaneous measurement of LST, NDVI,

Albedo and soil moisture to know ET at high

spatial resolution (100m). An integrated satellite

system dedicated for Hydrological application on

Indian Mission is needed with simultaneous

measurements from wide SAR, Nadir Altimeter,

Passive Radiometer with active rain radar.

5.0 References

1. A. K. Dubey, P. K. Gupta, S. Dutta, and B.

Kumar. 2014. "Evaluation of Satellite

altimetry derived river stage variation for the

brided Brahmaputa river." International

Journal of Remote Sensing 23: 7815-7827.

2. A.K. Dubey, P.K. Gupta, S. Dutta, and R.P.

Singh. 2015. "Water level retrieval using

SARAL/Altika observations in braided

Brahmaputra river." Marine Geodesy 38 (1):

549-567.

doi:10.1080/01490419.2015.1008156.

3. D. Entekhabi, E.G. Njoku, P.E. ONeill, K. H.

Kellogg, W.T. Crow, W.N. Edelstein, J.K.

Entin, S.D. Goodman, T.J. Jakson, J. Johson,

J. Kimball, J. R. Piepmeier R.D. Koster, N.

Martin, K.C. McDonald, M. Moghaddam, S.

Moran, R. Reichle, J.C. Shi and Team. 2010.

"The Soil Moisture Active Passive (SMAP)

Mission." Proc. of IEEE 98 (5): 704-716.

doi:10.1109/JPROC.2010.2043918.

4. E. A. Douglas, E. Rodriguez and D. P.

Lettenmaier. 2007. "Measuring surface water

from Space". Reviews of Geophysics, 45: 1-

24.

5. F. Frappart, F. SeyLER, J.M. Martinez, J.G.

Leon, and A. Cazenave. 2005. "Flood Plain

water storage in the Negro river basin

estimated from Microwave Remote Sensing

of inundation area and water levels." Remote

Sensing of Environment 99: 387-399.

6. F. Papa, S.K. Bala, R.K. Pandey, F. Durand,

V.V. Gopalkrishna, A. Rahman, and W.B.

Rossow. 2012. "Ganga- Brahmaputra river

discharge from Jason-2 radar altimetry: An

update to the long term satellite derived

estimation of continental fresh watr forcing

flux into Bay of Bengal." Journal of Geophys.

Res. C11021. doi:10.1029/2012JC008158.

7. F.Papa, F. Durand, W. B. Rossow, A.

Rahman, and S.K. Bala. 2010. "Satellite

altimeterderived monthly discharge of

Ganga-Brahmaputra river and its seasonal to

interannual variations from 1993 to 2008."

Journal of Geophysical Research 115

(C12013). doi:10.1029/2009JC006075.

8. H.S. Srivastava, P. Patel, Y. Sharma and R.

R. Navalgund. 2009. "Large area soil

moisture estimation using multi incidence

angle RADARSAT-1 SAR data." IEEE

transaction on Geosciences and Remote

Sensing 47: 2528-2535.

9. I.M. Bahuguna, B.P. Rathore, R. Brahmbhatt,

M. Sharma, S. Dhar, S.S. Randhawa, K.

Kumar, S. Romshoo, R.D. Shah, R.K.

Ganjoo and Ajai. 2014. "Are the Himalayan

glaciers retreating ?" Current Science 106

(7): 1008-1013.

10. M. R. Pandya, D. B. Shah, H.J. Trivedi, N.P.

Darji, R. Ramakrishnan, S. Panigrahy, J. S.

Parihar, A.S. Kirankumar. 2014. "Retrieval

of land surface temperature from the

Kalpana-1 VHRR data using a single-channel

algorithm and its validation over western

India." ISPRS Journal of Photogrammetry

and Remote Sensing 94: 160–168.

11. P.K. Gupta, S.Chauhan, and M.P. Oza. 2016.

"Modelling Surface Runoff and Trends

Analysis over India." Journal of Earth

System Science (under review).

12. P.K.Gupta, A.K. Dubey, N. Goswami, R.P.

Singh, P. Chauhan. 2015. "Use of

Saral/Altika observations for modeling river

flow." Marine Geodesy 38 (1): 614-625.

doi:10.1080/01490419.2015.1008157.

13. P. K. Gupta, S. R. Oza and S. Panigrahy

2010. Monitoring Transplanting Operation of

Rice crop using Passive Microwave

Radiometer Data. International Journal of

Bio-Systems Engineering 108 (2011): 28-35.

14. P. K. Gupta, S. Dutta and S. Panigrahy

(2009). Mapping of Conjunctive Water use

Productivity Pattern in an Irrigation

Command using Temporal IRS WiFS Data.

Journal of Water Resources Management 24

(1): 157-171 .

15. P. S. Thenkabail, S. Mitchell, H. Turral.

2005. "Ganges and Indus river basin

landuse/land cover (LULC) and irrigated

area mapping using continuous streams of

MODIS data". Remote Sens Environ 95:317–

341. doi:10.1016/j.rse.2004.12.018.

16. R. P. Singh, D.R. Mishra, A.K. Sahoo, and S.

Dey. 2005. "Spatial and Temporal variability

of soil moisture over India using IRS P4

MSMR data." International Journal of

Remote Sensing 26: 2241-2247.

17. R. R. Navalgund, R.P. Singh. 2010. "The

evolution of the earth observation system in

India." 90 (4).

18. R.K. Jaiswal, S.Mukherjee, J. Krishnamurthy

and R. Saxena. 2003. "Role of Remote

sensing and GIS techniques for generation of

ground water prospect zones towards rural

development- An Approach." International

Jouranl of Remote Sensing 24 (5): 993-1008.

19. R.K. Mall, A Gupta, R. Singh, R. S. Singh

and L. S. Rathore. 2006. "Water resources

and climate change: An Indian perspective."

Current Science 90 (12): 1610-1626.

20. R.M.Gairola. 2015. "SARTAL/ALTIKA

Science and Applications: Summary of

Achievements." NNRMS (B) Bulletin 39: 1-

11.

21. R.O. Green, T.H. Painter, D.A. Roberts and

J. Dozier. 2006. "Measuring the expressed

abundance of the three phases of water with

an imaging spectrometer over melting snow."

water Resources 42 (W10402).

doi:10.1029/2005WR004509.

22. R.P.Singh, S.R. Oza, K.N. Chaudhari, and

V.K. Dadhwal. 2005. "Spoatial and

Temporal Patterns of surface soil Moisture

over India estimated using surface wetnes

index from SSM/I microwave radiometer."

International Journal of Remote Sensing 26:

1269-1276.

23. Rakesh Kumar, R.D. Singh and K.D. Singh.

2005. "Water Resources of India." Current

Science 89 (5): 794-811.

24. S. Swenson, J. Famiglietti, M. Rodell, J.

Chen, and C.R. Wilson. 2008. "Estimating

profile soil moisture and ground water

variations using GRACE and Oklahoma

Mesonet soil moisture data." Water

Resources Research 44 (W01413).

25. S.K. Jain, A. Goswami and A.K. Saraf. 2010.

"Snow melt runoff modelling in Himalyan

basin with aid of satellite data." Internation

Journal of Remote Sensing 31 (24): 6603-

6618.

26. S.R. Oza, R.P. Singh, V.K. Dadhwal and P.S.

Desai. 2006. "Large area soil moisture

estimation using spaceborne multi frequency

passive microwave data." Journal od Indian

Society of Remote Sensing 34: 343-350.

27. S. Siebert, P. Döll, J. Hoogeveen, J.M.

Faures, K. Frenken, S. Feick. 2005.

"Development and validation of the global

map of irrigation areas". Hydrol Earth Syst

Sci 9:535–547

28. T. Misra, S.S. Rana, N.M. Desai, D.B. Dave,

Rajeevjyoti, R.K. Arora, C.V.N. Rao, B.V.

Bakori, R. Neelakantan, and J.G. Vachchani.

2013. "Synthetic Aperure Radar payload on

boardRISAT-1: Configuration, Technology

and Performance." Current Science 104 (4):

446-461.

29. V.M. Tiwari, J. Wahr and S. Swensen. 2009.

"Dwindling ground water resources in

Northern India from satellite gravity

observations." Geophys. Res. Lett. 36

(L18401). doi:10.1029/2009GL039401.

30. V.V. Rao, J.R. Sharma and V.K. Dadhwal.

2013. "Water Resources of India- Critical

Issues and Satellite technology Options."

NNRMS Bulletin 38: 1-9.

31. Y.H. Kerr, P. Waldteufel, J.P.Wigneron, J.M.

Martinuzzi, J. Font, and M. Berger. 2001.

"Soil Moisture Retrieval from Space: The

Soil Moisture and Ocean Salinity (SMOS)

Mission." IEEE Transactions on Gepsciences

and Remote Sensing 39 (8): 1729-1735.

32. Wagner, W, Lemoine, G, & Rott, H (1999) A

method for estimating soil moisture from

ERS scatterometer and soil data. Remote

Sensing of Environment, 70, 191–207.

doi:10.1016/S0034-4257 (99) 00036-X.

SATELLITE ALTIMETRY OVER LAND

S. CHANDER

Land Hydrology Division

Geosciences, Hydrology, Cryosphere Sciences and Applications Group (EPSA)

Space Applications Centre, ISRO

Ahmedabad-380015

Radar altimetry over Ocean is a mature science with centimeter level accuracy. But over inland

water bodies there is still scope of improvement by analyzing the raw altimeter measurement

“waveform”. Complete retrieval algorithm is needs to be modified taking into consideration the

land contamination within the altimeter foot-print. The methodology can be broadly divided into

three major categories on waveform classification, waveform retracking and dedicated inland

range corrections algorithms. The 40 Hz waveforms can be classified based on the pattern

matching, maximum likelihood estimation, linear discriminant analysis (LDA) and Bayesian

classifier. Waveforms can be retracked using Brown, Ice-2, Threshold, and Offset Centre of

Gravity methods. Range corrections can be modified by ECMWF operational, ERA reanalysis

pressure fields and global ionosphere maps. With modified methodology it is possible to estimate

inland water level measurement within 10 cm level accuracy. We have tested these results with

the gauge measurements and GPS measured water levels over the validation sites. This water

level product is now being disseminated over 50 major Indian inland water bodies through

VEDAS site.

Key Words: Altimetry, Waveform Retracking, Hydrology, Geophysical range corrections,

ionospheric total electron content, SARAL/AltiKa radar altimeter, SWOT.

1.0 Introduction

Satellite altimetry was begun with the

ocean/gravitational science community needs in

the 1960's. The principle of radar altimeters is

deceptively straightforward. The altimeter

transmits a short pulse of electromagnetic

radiation with known power towards the earth's

surface. The pulse interacts with the surface and

part of the incident radiation is reflected back to

the altimeter. This return radar power received by

the altimeter is recorded through time, producing

an altimetric "waveform". By analyzing

amplitude and shape of return waveform, various

characteristics of the surface such as backscatter

coefficient, surface roughness, wave height and

wind speed over oceans, etc. can be retrieved. In

Figure 1 explains the basic principle of the

altimeter. As shown in the figure the radar pulse

need to be corrected due to atmospheric and

ionospheric effects. Onboard radiometer is used

to correct the signal due to highly spatial and

temporal variable water vapor. But the foot-print

of the radiometer is an order of larger than the

altimeter foot-print, so is more prone to land

contamination in the coastal and inland water

bodies. Meteorological model derived water

vapor information can be used than to estimate

the wet tropospheric correction. All the altimeter

measurements are provided with respect to single

reference ellipsoid, i.e. WGS84. This reference

can be converted to mean sea level and geoid

once the high resolution information is available

over the reference site.

Fig.1: Altimeter Principle

Altimetric measurements have a wide range of

applications in Oceanography, Geodesy,

Geophysics, Glaciology, and Continental

Hydrology. Satellite altimetry for inland water

applications has evolved from investigation of

water height retrieval to its monitoring since last

two decades. Monitoring water level of inland

water bodies has become essential for studying

the effects of global climate change and

increasing population pressure on the fresh water

resources. The traditional monitoring of river

gauges is limited to few locations. But, non-

availability of in-situ measurements at desired

locations imposes serious constraints on basin

scale hydrology study. Besides this, in-situ

discharge measurement is expensive and time

consuming. Satellite Altimetry has the potential

to monitor inland water bodies continuously and

consistently over a long period of time. Thus

altimetry can support subsequent hydrologic-

hydraulic modeling for planning and

management of water resources at regional

scales. Altimetry derived reservoir/ river levels

can subsequently be used to deal with key inland

water resources problems such as flood, rating

curve generation for remote locations, reservoir

operations, and calibration of river/lake models.

2.0 Limitations over Land

Altimeter waveform can have different

characteristic shapes according to reflecting

surfaces like Open Ocean, ice caps, inland water,

and continental surfaces. Waveforms over open

water show a characteristic shape, often referred

to as "ocean-like" or "Brown-like", featuring a

sharp rise up to a maximum value, followed by a

gentle sloppy plateau. But over inland water

bodies and land waveform may have different

shapes with more than one peak due to more than

one reflecting surface within the altimeter foot-

print. More details about the waveform shapes

and classification can be found elsewhere. If

these waveform are not classified can lead to

height inaccuracy of the order of few meters. One

such example of SARAL altimeter multi-peak

waveform is shown in figure 2 over one of the

Inland water body.

Fig.2: Multi-peak waveforms

Altimeter can only measure within a narrow

range window vertically, called "analysis

window". As the satellite-to-Earth distance

changes with the satellite motion along the orbit

(due to Earth surface topography), the position of

the analysis window must be adjusted to ensure

that the altimeter samples at the time when the

pulse hits the surface. This is done, by "on-board

tracker", a predictive device that minimizes the

risk of the altimeter losing track of the surface.

The tracker on-board the altimeter tries to align

the midpoint of the leading edge of the waveform

at the nominal tracking gate that is fixed for a

given altimeter. But generally it is unable to

accurately predict the range and thus, the

midpoint of the leading edge is not always

aligned at the nominal tracking gate. Over ocean

this correction rarely rises more than few

centimeter. In order to achieve very high

accuracy in range the waveforms acquired are

down linked to earth where they are retracked

manually to improve the range estimates.

Over land the situation is more complicated, and

if along track slope is higher than few meter the

altimeter can lost its track. Thus for land

altimetry, the waveform must be post processed

to exactly estimate the range. This is called

"waveform retracking". Various model (based on

physics) and empirical retrackers have been

developed for characterizing the waveforms from

open-ocean, coastal and continental ice. The

formation of the theoretical shape of an echo over

the surface was given by Brown & Hayne [1].

Beta-5 retracker [2] is a five parameter functional

model designed to derive geophysical parameters

for Brown like waveforms obtained over the

ocean and large water bodies by trying to fit the

acquired waveform to the model curve. Offset

Centre of Gravity Retracker (OCOG) is normally

used for rectangular type waveforms where no

model can be fitted [3]. It attempts to find the

Centre of Gravity, Amplitude and Width of the

waveform. The threshold retracker was first

developed by Davis [4]. The amplitude of the

waveform is determined as the OCOG retracker.

Then the leading edge is determined by

computing the gate corresponding to threshold of

the amplitude after taking care of the thermal

noise component. Different threshold levels were

suggested based on the scattering mechanism, i.e.

50% threshold level for waveforms dominated by

surface scattering and 25% threshold for volume

scattering.

3.0 Sub-waveform based waveform retracker

A modified sub-waveform based retracker is

developed and implemented to see the

improvement in retrieved range. Since most

categories of waveforms namely brown, specular

and rectangular can be accurately retracked by

existing empirical and physically based

retrackers, it remains to be seen whether

multipeak waveforms can be retracked correctly

using the subwaveform based retracker.

Subwaveform based retracking technique like the

improved threshold exist but it requires some

reference data of water level height to identify the

appropriate subwaveform and determine the

correct range. This new retracker attempts to find

the first leading edge in the waveform and retrack

the subwaveform using the 50% threshold

method without trying to explain the full shape of

the waveform. The basic assumption is that the

first leading edge is from the surface closest to the

satellite and for flat surfaces it is from the nadir.

Figure 3 shows the improvement in range

estimation using this modified retracker with

comparison to previous retrackers. The range can

be corrected by order of 5 m, as in the case of

multipeak waveform [8].

Fig.3: Comparison between retracking algorithms

4.0 Geophysical Range Corrections

The range estimated after waveform retracking

needs to be corrected for atmospheric effects and

sea-atmospheric interaction. The presence of dry

gasses, water vapor in the troposphere and free

electron in the ionosphere reduces the speed of

the radar pulse causing the observed range to

become longer and the sea surface height to be

too low. These atmospheric range correction

algorithms were modified for inland water

bodies.

The Dry Tropospheric correction (DTC) effect

can be derived from the sea level pressure data set

mainly from model re-analysis or forecasts. The

DTC has strong height dependence and since

satellite altimetry is primarily designed for ocean

applications, for which no such dependence

exists, altimeter products fail to provide the DTC

appropriate for inland water studies. The

correction was computed from Sea Level

Pressure (SLP) grids, using equation [5]:

𝐷𝑇𝐶 =0.0022768 𝑃0

1 − 0.0026 𝐶𝑜𝑠 2∅ − 0.28 × 10−6 ℎ𝑠

Here 𝑃0 is the sea level pressure, ∅ is the latitude

and ℎ𝑠 is the surface height.

The path delay due to the presence of water vapor

in the atmosphere, the wet tropospheric

correction (WTC), is one of the major error

sources in satellite altimetry. On-board

microwave radiometers corrections are hampered

by the contamination of the surrounding lands.

Alternatively, the vertically integrated water

vapor required for the wet tropospheric range

correction could be obtained from meteorological

model analyses. The WTC was calculated from

global grids of two single-level parameters

provided by global atmospheric models from the

following expression [6]:

𝑊𝑇𝐶 = − (0.101995 +1725.55

50.44+0.789 𝑇0)

𝑇𝐶𝑊𝑉

100

Here 𝑇0 is the near-surface air temperature (two-

meter temperature) and 𝑇𝐶𝑊𝑉 is the total column

water vapor.

Atmospheric refraction from free electrons and

ions in the upper atmosphere is related to the

dielectric properties of the ionosphere. This

columnar electron density can be approximated

by the Total Electron Content (TEC), i.e. the

integrated electron density. Therefore, the

ionospheric delay can be calculated using

following equation [7]:

𝐼𝐶 = −40.3 𝑇𝐸𝐶

𝑓𝐾𝑎2

For smaller inland water bodies earth tide is

applied, but elastic-ocean and ocean-loading tides

are only applicable for larger water bodies (~

1000 km2). The inverse barometric correction is

not applied because the lakes/reservoirs are

closed systems. The sea state bias (SSB)

correction is also not applied because wind

effects tend to be averaged out along-track.

Model predicted load tide and solid earth tide was

directly taken from the SARAL SGDR products.

Fig. 4: Flow chart to estimate dynamic height

Figure 4 shows the complete flow chart to

estimate the dynamic water level using altimeter

dataset. With the modified retrieval algorithms

the water level was retrieved over the Ukai

reservoir and found to be matching within RMSE

15 cm, and validated with the gauge dataset and

GPS measured water levels. The results are

shown in figure 5 [9].

Fig.5: Retrieved water level over Ukai reservoir

5.0 Conclusions and Future Directions

Over inland water bodies due to land

contamination within the altimeter foot-print,

generally altimeter gets very complex multi-

peaked waveforms. Due to complex waveforms,

the accuracy of the retrieved water level is limited

to few tens of centimeter, but with the proposed

methodology we have achieved retrieval

accuracy better than 15 cm. Knowledge about the

waveform shape prior to retracking is useful for

optimizing the choice of the retracking method.

Waveform retracker should be robust so that it

can take care for a number of waveform shapes

that generally found in nature. The range

corrections dedicated to inland water bodies are

important to remove the seasonal trend from the

water level retrieval. Overall, the altimeter

retrieved water level was found to be matched

well both with the gauge measurement. Altimeter

has a limitation that it only gives information

along the nadir track, but generally over inland

water bodies we require across track information

as well. Surface water and Ocean topography

(SWOT) is a proposed altimeter that will utilize

Ka band radar interferometry technique to

estimate the slope and extent information of the

water surface in across track direction.

References

1 G. S. Hayne, “Radar altimeter mean return

waveforms from near-normal-incidence ocean

surface scattering,” IEEE Transactions on

Antennas Propagations 28(5), 687–692 (1980).

2 T. V. Martin, H. J. Zwally, A. C. Brenner, and

R. A. Bindschadler, “Analysis and retracking of

continental ice sheet radar altimeter waveforms,”

Journal of Geophysical Research: Oceans

88(C3), 1608–1616 (1983).

3 D. J.Wingham, C. G. Rapley, and H. Griffiths,

“New techniques in satellite tracking system,”

Proceedings of IGARSS’86 symposium, 1339–

1344 (1986).

4 C. H. Davis, “A robust threshold retracking

algorithm for measuring ice-sheet surface

elevation change from satellite radar altimeters,”

IEEE Transactions on Geoscience and Remote

Sensing 35(4), 974–979 (1997).

5 E. K. Smith and S. Weintraub, “The Constants

in the Equation for Atmospheric Refractive Index

at Radio Frequencies,” Proceedings of the IRE

41(8), 1035–1037 (1953).

6 M. Bevis, S. Businger, S. Chiswell, T. A.

Herring, R. A. Anthes, C. Rocken, and R. H.

Ware, “GPS meteorology - Mapping zenith wet

Delays onto Precipitable Water,” J. Appl. Meteor.

33, 379–386 (1994).

7 N. Picot, K. Case, S. Desai, and P. Vincent,

“AVISO and PODAAC User Handbook, IGDR

and GDR Jason Products,” (2003).

8 D. Ganguly, S. Chander, S. Desai, and P.

Chauhan, “A Subwaveform-Based Retracker for

Multipeak Waveforms: A Case Study over Ukai

Dam/Reservoir,” Marine Geodesy 38(1), 581–

596 (2015).

9 S. Chander and D. Ganguly, “Development of

water level estimation algorithms using

SARAL/Altika dataset & Validation over the

Ukai Reservoir, India”, accepted in Journal of

Applied Remote Sensing (2016).

HYDROLOGICAL MODELLING AND REMOTE SENSING

P. K. GUPTA

Land Hydrology Division

Geosciences, Hydrology, Cryosphere Sciences and Applications Group (EPSA)

Space Applications Centre, ISRO

Ahmedabad-380015

Water in our planet is available in the atmosphere, the oceans, on land and within the soil and

fractured rock of the earth’s crust Water molecules from one location to another are driven by the

solar energy. Moisture circulates from the earth into the atmosphere through evaporation and then

back into the earth as precipitation. In going through this process, called the Hydrologic Cycle,

water is conserved – that is, it is neither created nor destroyed. All hydrological models are

simplified representations of the real world. Models can be either physical (e.g. laboratory scale

models), electrical analogue or mathematical. The physical and analogue models have been very

important in the past. However, the mathematical group of models is by far the most easily and

universally applicable, the most widespread and the one with the most rapid development with

regard to scientific basis and application. A hydrological model is composed of two main parts, a

hydrological core and a technological shell. The hydrological core is based on a certain

hydrological scientific basis providing the definitions of variables, the process descriptions and

other aspects. The technological shell is the programming, user interface, pre- and post-processing

facilities etc. Recent advances in the remote sensing technologies such as altimetery,

scatterometery, LANDASAT8, passive radiometers, multiple sensors in a single platform, SMAP,

SMOS etc. have made it possible to estimate various hydrological parameters such as rainfall,

river water levels, ET, soil moisture, groundwater etc. Current research is focused on

parameterization of various hydrologic-hydraulic models using remote sensing derived

hydrological variables.

Key Words: Satellite Remote Sensing, Hydrological models, flood modeling, water balance,

river flow models, curve number.

1.0 Key issues in water Resources

Effects of exploitation of water resources

Periodic or permanent lowering of

groundwater table

Increased concentration of pollutants in

the aquifer

Increased risk of salt water intrusion and

land subsidence

Irrigation

Continuing low efficiency of irrigation

projects

Environmental concerns due to

excessive irrigation

Land Degradation and Soil Erosion

Desertification due to increased human

and livestock population

Negligence of upland catchment

management

Surface and Groundwater Pollution

Contamination of water due to waste

disposal

Nitrogen and Pesticides pollutions due

to agricultural activities

Floods and Droughts

Changes in the land use

Changes in the hydrological regime

2.0 Hydrological processes

Following hydrological processes need to be

considered for the development of the

hydrological models;

• Canopy rain Interception

• Infiltration

• Depression storage

• Soil moisture storage

• Shallow sub-surface runoff / baseflow

• Preferential /macropore flow

• Runoff generation processes

• Infiltration excess (Horton flow)

• Saturation excess (Dunne flow)

• Flow routing

• Soil evaporation

• Vegetation transpiration

• Water surface evaporation

• Canopy water evaporation

• Groundwater flow

• Lakes

• Wetlands

• Structures that control waters

• Snowmelt

Precipitation

Precipitation occurs when atmospheric moisture

becomes too great to remain suspended in clouds.

It denotes all forms of water that reach the earth

from the atmosphere, the usual forms being

rainfall, snowfall, hail, frost and dew. Once it

reaches the earth’s surface, precipitation can

become surface water runoff, surface water

storage, glacial ice, water for plants,

groundwater, or may evaporate and return

immediately to the atmosphere. Rainfall

measurements can be done using rain gauge and

Fig. processes involve for the development of the

hydrological system model.

satellite remote sensing. Sources from which

remote sensing derived rainfall is available is

Climate Prediction Centre (NOAA), tropical

rainfall measurement mission (TRMM),

meteorological satellite (METEOSAT) etc.

Interception:

Interception is defined as the process whereby

precipitation is retained on the leaves, branches,

and stems of vegetation. This intercepted water

evaporates directly without adding to the

moisture storage in the soil. The interception

process is modelled as an interception storage,

which must be filled before stem flow to the

ground surface takes place. The size of the

interception storage capacity, depends on the

vegetation type and its stage of development,

which is characterised by the leaf area index.

Runoff:

Runoff is the water that flows across the land

surface after a storm event. As rain falls over

land, part of that gets infiltrated the surface and

RAINFALL

Through fall

Canopy storage

INFILTRATION

Soil water store

Groundwater discharge

Direct

rain onto stream

From Open waterFrom

Canopy storage

From soil

EVAPORATION

Geological lenses

From crops

TRANSPIRATION

STREAM FLOW

Interception

Figure: Processes for modelling hydrological cycle in the forest system

remaining water flows as overland flow. As the

flow bears down, it notches out rills and gullies

which combine to form channels. The

geographical area which contributes to the flow

of a river/channel is called a catchment of that

river/channel.

Storage:

Portion of the precipitation falling on land surface

which does not flow out as runoff gets stored as

either surface water bodies like Lakes, Reservoirs

and Wetlands or as sub-surface water body like

soil moisture and Ground water.

The following definitions may be useful:

Lakes: Large, naturally occurring inland body of

water

Reservoirs: Artificial or natural inland body of

water used to store water to meet various

demands.

Wet Lands: Natural or artificial areas of shallow

water or saturated soils that contain or could

support water–loving plants.

Unsaturated Zone (Vadose Zone):

Zone between ground surface to groundwater

table is known as unsaturated zone. The

unsaturated zone is usually heterogeneous and

characterized by cyclic fluctuations in the soil

moisture as water is replenished by rainfall and

removed by evapotranspiration and recharge to

the groundwater table. Unsaturated flow is

primarily vertical since gravity plays the major

role during infiltration.

Soil water constants

For a particular soil, certain soil water

proportions are defined which dictate whether the

water is available or not for plant growth. These

are called the soil water constants, which are

described below.

• Saturation capacity: this is the total water

content of the soil when all the pores of the soil

are filled with water. It is also termed as the

maximum water holding capacity of the soil. At

saturation capacity, the soil moisture tension is

almost equal to zero.

• Field capacity: this is the water retained by an

initially saturated soil against the force of gravity.

Hence, as the gravitational water gets drained off

from the soil, it is said to reach the field capacity.

At field capacity, the macro-pores of the soil

are drained off, but water is retained in the

micropores. Though the soil moisture tension at

field capacity varies from soil to soil, it is

normally between 1/10 (for clayey soils) to 1/3

(for sandy soils) atmospheres.

• Permanent wilting point: plant roots are able

to extract water from a soil matrix, which is

saturated up to field capacity. However, as the

water extraction proceeds, the moisture content

diminishes and the negative (gauge) pressure

increases. At one point, the plant cannot extract

any further water and thus wilts.

The Saturated Zone (SZ):

Saturated subsurface flow or ground water table

is known as saturated zone. Ground water storage

is the water infiltrating through the soil cover of

a land surface and traveling further to reach the

huge body of water underground. As mentioned

earlier, the amount of ground water storage is

much greater than that of lakes and rivers.

However, it is not possible to extract the entire

groundwater by practicable means. It is

interesting to note that the groundwater also is in

a state of continuous movement – flowing from

regions of higher potential to lower. The rate of

movement, however, is exceptionally small

compared to the surface water movement.

Evapotranspiration

Evapotranspiration is actually the combination of

two terms – evaporation and transpiration. The

first of these, that is, evaporation is the process of

liquid converting into vapour, through wind

action and solar radiation and returning to the

atmosphere. Evaporation is the cause of loss of

water from open bodies of water, such as lakes,

rivers, the oceans and the land surface. It is

interesting to note that ocean evaporation

provides approximately 90 percent of the earth’s

precipitation. However, living near an ocean does

not necessarily imply more rainfall as can be

noted from the great difference in the amount of

rain received between the east and west coasts of

India.

Transpiration is the process by which water

molecules leaves the body of a living plant and

escapes to the atmosphere. The water is drawn up

by the plant root system and part of that is lost

through the tissues of plant leaf (through the

stomata). In areas of abundant rainfall,

transpiration is fairly constant with variations

occurring primarily in the length of each plants

growing season. However, transpiration in dry

areas varies greatly with the root depth.

Evapotranspiration, therefore, includes all

evaporation from water and land surfaces, as well

as transpiration from plants.

Potential evapotranspiration (PET)

Pan evaporation The evaporation rate from pans

filled with water is easily obtained. In the absence

of rain, the amount of water evaporated during a

period (mm/day) corresponds with the decrease

in water depth in that period. Pans provide a

measurement of the integrated effect of radiation,

wind, temperature and humidity on the

evaporation from an open water surface.

Although the pan responds in a similar fashion to

the same climatic factors affecting crop

transpiration, several factors produce significant

differences in loss of water from a water surface

and from a cropped surface. Reflection of solar

radiation from water in the shallow pan might be

different from the assumed 23% for the grass

reference surface. Storage of heat within the pan

can be appreciable and may cause significant

evaporation during the night while most crops

transpire only during the daytime.

Overland flow:

The amount of rainfall in excess of the infiltrated

quantity flows over the ground surface following

the land slope. This is the overland flow. The

portion that infiltrates moves through an

unsaturated portion of the soil in a vertical

direction for some depth till it meets the water

table, which is the free surface of a fully saturated

region with water (the ground water reserve).

3.0 Remote Sensing and GIS

For water resources engineer, locating aerial

extent of water bodies like lakes, rivers, ponds,

etc. from remotely sensed data is an important

task. The spectral response from a water body is

complex, as water in any quantity is a medium

that is semi-transparent to electromagnetic

radiation. Electromagnetic radiation incident on

water may be absorbed, scattered and transmitted.

The spectral response also varies according to the

wavelength, the nature of the water surface (calm

or wavy), the angle of illumination and

observation of reflected radiation from the

surface and bottom of shallow water bodies. Pure

clear water has a relatively high reflectance in the

visible wavelength bands between 0.4 and 0.6μm

with virtually no reflectance in the near-infrared

(0.7μm) and higher wavelengths. Thus clear

water appears dark on an infrared image.

Therefore, location and delineation of water

bodies from remotely sensed data in the higher

wave bands can be done very accurately.

The satellite Remote Sensing provides

information in spatial and temporal domains,

with high resolution about the processes of the

land phase of the hydrological cycle, which is

very crucial for successful model analysis,

prediction and validation (Jagadeesha, 1999). RS

techniques can be extremely useful in estimating

a number of key variables of DHMs, particularly

for large basins with sparse data. RS technologies

are often considered as innovative ways of

obtaining data at a reduced cost (Koblinsky et al.,

1992) and replace the conventional techniques. In

addition to that RS can provide time series of data

relatively easily and enabling periodically

updating of variables. Benefit/cost ratio ranging

from 75:1 to 100:1 can be realised in using

remotely sensed data in hydrology and water

resources management (Kite and Pietroniro,

1996).

Table: RS based hydrological variables

The use of RS technology involves large amount

of spatial data management. The GIS technology

provides suitable alternatives for efficient

management of large and complex databases. The

possibility of rapidly combining data of different

types in a GIS has led to significant increase in its

use in hydrological applications. The use of RS

data, in combination with DHM, provides new

possibilities for deriving spatially distributed

time series of input variables, as well as new

means for calibration and validation of the

hydrological model (Bastiaanssen et al., 2000;

Fortin et al., 2001).

Table: Hydrological processes and address

through remote sensing derived variables.

3.0 Hydrological models classifications

Models are classified based on the process

description.

Fig. classification of hydrological models based

on the process description.

Hydrologic parameters Sensor Technology Resolution Repeat ivity

Rainfall TRMM, INSAT, NOAA CPC,

JAXA

Precip. Radar (JAXA)TMI,

VIRS VHRR

0.01 to 0.25 Deg Daily

3 hourly

Soil moisture SSMI, AMSR Radiometers 12-56 km 5-day

Groundwater GRACE gravity 100,000 km2 30 days

Lake/reservoir levels Jason-2, ALTIKA, Sentinel-3 Altimetric radar 350 m 10 day

Evapotranspiration MODIS, INSAT Visible/NIR 1 km to 8 km 1-2 days

Stream discharge Jason-2, ALTIKA Altimetric radar 350m, 175m 10 -35 day

Leaf area index INSAT, MODIS Visible/NIR 1 km 8 day comp.

Topography CARTOSAT-1, SRTM,

GTOPO, ASTER

Optical, microwave 10 m to 1 km -

Insolation INSAT, MODIS VHRR 1 km to 8 km daily

Land Surface Temp. INSAT, MODIS Thermal Infrared 8 km hourly

Land use/cover RESOURCESAT-2, MERIS,

MODIS, SPOT

Optical 56 m to 1 km yearly

Lakes/Wetland extents RESOURCESAT-2, MODIS Optical 23 m to 250 m yearly

Snow covered area RESOURCESAT-2, MODIS Optical 56 m 8 day

Albedo INSAT, MODIS Optical 1 km 16 day

NDVI INSAT, MODIS, SPOT Optical 1 km 16 day

Wind speed , humidity INSAT-

3D/MEGHSTROPICS

Optical/sounder Daily

GPP INSAT, MODIS Optical 1 km 8 day

NPP INSAT, MODIS Optical 1 km yearly

S.N. Processes RS parameters Other Data

1 Canopy rain Interception LAI

2 Infiltration LULC, Rainfall Soil

3 Depression storage DEM

4 Soil moisture LULC, Soil Moisture, DEM Soil

5 Base flow DEM Geological formation

6 Preferential flow LULC, DEM Macroporosity

7 Runoff (Saturation excess-Infiltration excess)

DEM, LULC, Rainfall Soil

8 Soil Evaporation LST, Humidity Soil

9 Water Evaporation LST, Water Bodies, Humidity

10 Canopy water Evaporation LAI

11 Transpiration LAI, Humidity, Water level (Altika) Root depth and density

12 Flow routing DEM, LULC

13 Ground Water Flow Ground Water Anomaly (GRACE) Lethology

14 Lake/Wetlands Land Water Mask

15 Structural control water Reservoir Extent Dam Locations

16 Snow Melt Albedo, Insulation, Snow cover, Short and long wave radiation, LST

Weather Parameters

Hydrological Simulation Model

Deterministic

Empirical

Stochastic

Lumped

Conceptual

Distributed

Physically based Joint Stochastic-

Deterministic

Some important definitions in hydrological

modeling:

Model: a conceptual or physically based

procedure for numerically solving the

hydrological processes.

A mathematical model: is a set of mathematical

expressions and logical statements combined in

order to simulate the natural system.

Static model: empirical and regression model in

which time is not independent variable

Dynamic model: require different equation with

time as independent variable and this shows the

time variability of output.

Why simulation model: are used to understand

how system work and interact one another.

Deterministic model: is a model where two

equal sets of input always yields the same output

if run through the model under identical

conditions. A deterministic model has no inner

operations with a stochastic behaviour.

Empirical model: is a model developed without

any consideration of the physical processes that

we otherwise associate with the catchment. The

model is merely based on the analysis of the

concurrent input and output time series. Also,

known as black box model.

Lumped model: is a model where the catchment

is regarded as one unit. The variables and

parameters are thus representing average values

for the entire catchment. Thus the virtual world is

just reduced to just one object.

Conceptual model: physically sound knowledge

and empirically derived equations. Physical

significance is not clear that is why not possible

to assess the parameters from direct

measurement.

Stochastic model: has at least one component of

random character which is not explicit in the

model input, but only implicit or hidden.

Therefore, identical inputs will generally results

in different outputs if run through the model

under, externally seen, identical conditions.

Physical based: description of natural system

using the basic mathematical representation of

the flows of mass, momentum and various forms

of energy. Known as white box model. These

model consists of linked Partial Differential

Equations (PDE’s) with parameters, which in

principle have direct physical significance and

can be evaluated by independent measurements.

Physical processes like conservation of mass and

momentum acting upon input variables are taken

into account.

Distributed: able to take spatial variation of

variables and parameters into account which are

spatially interactive on cell by cell basis.

Simulations: is time varying description of the

natural system computed by the hydrological

model. A simulation may be seen as the models

imitation of the behaviour of the natural system

Parameter: a parameter is a constant in the

mathematical expressions or logical statements of

the mathematical model. It remains constant in

the virtual time.

Variable: is a quantity which varies in space and

time. It can be a series of inputs to and outputs

from the model, but also a description of

conditions in some component of the model.

Modelling system: is defined as a generalized

software package, which, without program

changes, can be used to establish a model with the

same basic types of equation (but allowing

different parameter values) for different

catchments.

Model qualification: an estimation of the

adequacy of the conceptual model to provide an

acceptable level of agreement for the domain of

intended application.

Model Verification: substantiation that a

computerized model is in some sense a true

representation of a conceptual model within

certain specified limits or ranges of application

and corresponding accuracy.

Model calibration: involves manipulation of a

specific model parameters to reproduce the

response of the catchment under study within the

range of accuracy specified in the performance

criteria. It is important to assess the uncertainty in

the estimation of model parameters, for example

from sensitivity analysis.

Model validation: is the processes of

demonstrating that a given site specific model is

capable of making sufficiently accurately

predictions. This implies the application of the

calibrated model without changing the

parameters values that were set during the

calibration when simulating the response for

another period than the calibration period. The

model is said to be validated if its accuracy and

predictive capability in the validation period have

been proven to lie within acceptable limits

Distributed Hydrological Models (DHMs) can

serve as a tool to simulate the hydrological water

balance in the command/watershed, which is

essential to reassess the crop water demand both

in space and time. But these DHMs face the

problem of inadequate field data to describe the

processes of hydrological cycle accurately. The

amount of information available using the

conventional method is often very less than the

ideal to run a spatially distributed model

(Vachaud and Chen, 2002). Secondly,

development of more complex, physically

realistic, distributed hydrological models has

dramatically increased the demand for spatial

data (Pietroniro and Leconte, 2000).

Contrary to the lumped conceptual models, a

distributed physically based model does not

consider the water flows in an area to take place

between a few storage units. Instead, the flows of

water and energy are directly calculated from the

governing continuum (partial differential)

equations, such as for instance the Saint Venant

equations for overland and channel flow,

Richards’ equation for unsaturated zone flow and

Boussinesq’s equation for groundwater flow.

Distributed physically-based models have been

used for a couple of decades on a routine basis for

the simulation of hydrological processes.

Today, several general-purpose catchment model

codes of this type exist such as SHE (Abbott et

al., 1986), MIKE SHE (Refsgaard and Storm,

1995), IHDM (Beven et al. 1987). Distributed

physically based models give a detailed and

potentially more correct description of the

hydrological processes in the catchment than do

the other model types. Moreover, they are able to

exploit the quasi-totality of all information and all

knowledge that is available concerning the

catchment that is being modelled. The distributed

physically based models can in principle be

applied to almost any kind of hydrological

problem. However, in practice, they will be used

complementary to the other model types for cases

where the other models are not suitable. Some

examples of typical applications are:

Prediction of the effects of catchment changes

due to human interference in the hydrological

cycle, such as changes in land use (including

urbanization), groundwater development and

irrigation.

Prediction of runoff from ungauged catchments

and from catchments with relatively short

records.

As opposed to the lumped conceptual models,

which require long historical time series of

rainfall, runoff and evaporation data for the

assessment of parameters, the parameters of the

distributed physically-based models may be

assessed from intensive, short-term field

investigations.

Water quality and soil erosion modelling for

which a more detailed and physically correct

simulation of water flows is important.

4.0 Calibration and validation of the models

In the present study, resistance number and

seepage loss are considered as the model

calibration parameters. Resistance number used

in this model is defined as the reciprocal of the

Manning's roughness coefficient, which is

otherwise known as Strickler's coefficient. For

calibrating the model, an initial run is made with

default value of global resistance number and

seepage loss. Selection of locations for

calibration is done based on the availability of

observed flow data. Based on the comparison

between the observed and simulated flows, global

resistance number and seepage loss are adjusted.

This process is continued until the observed and

simulated values are in close agreement. For

further refinement of results, local resistance

numbers are used in the system and simulations

are done till the better match between observed

and simulated flows.

Calibrated model is validated for the period other

than considered for the calibration of the model.

Two goodness-of-fit criteria recommended by the

ASCE Task Committee (ASCE, 1993a,b), i.e.,

percent deviation of flow volume DV and Nash-

Sutcliffe coefficient R2, are considered to draw a

better conclusion from the comparison of

observed and simulated flow values. Percent

deviation of flow volume is calculated by using

the following formula.

100

V

VVD

O

SOV

where, Vs = simulated flow volume (m3); and Vo

= observed flow volume (m3). The value of Dv

should be zero for a perfect model.

The Nash-Sutcliffe Coefficient is calculated as

follows:

2

avO

2

SO2

QQ

QQ1R

where, Qo = observed discharge (m3/s); Qs =

simulated discharge (m3/s); and Qav = mean of the

observed discharge (m3/s). The value of R2 is 1

for the perfect model.

In addition, Student's t-value is also computed to

test the significance of the difference of means of

observed and simulated flows. The Student's t-

value is estimated by using the following

formula.

1

0

s

n

S

ddt

where, ts = computed t-value, d = mean of the

residuals; d0 = hypothesized mean which is

considered as zero; S = standard deviation of

residuals and n1 = number of data.

Status of Application of hydrological

modelling systems to various problems

5.0 Conclusions and Future Directions

Remote sensing has a strong technological basis

for the development of advanced sensors and

processing systems from various satellite

platforms whereas hydrology is more of science

oriented describing the various processes

involved in the water cycle. Therefore, current

research focus on integrating remote sensing

derived hydrological variables into various

hydrologic and hydraulic models to bridge the

gap between the point measurements and

mathematical model simulations.

Parameterization of a hydrological system model

demands for various datasets. Some of the crucial

hydrological parameters which are estimated

from remote sensing may be utilize for the

initialization, boundary and for the calibration

and validation of models. In near future, with the

evolving satellite technology (Surafce Water

Ocean Topography (SWOT, NASA-ISRO multi

frequency SAR (NISAR), some of the

hydrological processes such as surface runoff,

river flow etc. shall be addressed from satellite

platforms.

6.0 References

1. ASCE Task Committee on Definition of

Criteria for Evaluation of Watershed Models

(1993a). Criteria for evaluation of watershed

models. J. Irrig. and Drain. Engrg., ASCE,

119 ( 3): 429-442.

2. ASCE Task Committee on Irrigation Canal

System Hydraulic Modeling. (1993b).

Unsteady flow modeling of irrigation canals.

J. Irrig. and Drain. Engrg., ASCE, 119 (4):

615-630.

3. Bastiaanssen, W. G. M., Molden, D. J. and

Makin I. W. (2000). Remote sensing for

irrigated agriculture: examples from research

and possible applications. Agric. Water

Mgmt., 46: 130-155.

4. Bos, M. G. (1997). Performance indicators

for irrigation and drainage. Irrig. and Drain.

Sys., 11: 119-137.

5. DHI. (1988). MIKE 11 Scientific

documentation and user guide, DHI,

Copenhagen, Denmark.

6. Fortin, P. J., Turcotte, R., Massicotte, S.,

Moussa, R., Fitzback, J. and Villeneuve, P. J.

(2001). Distributed watershed model

compatible with remote sensing and GIS data

II: Application to Chaudie’re watershed. J.

Hydrol. Engrg., ASCE, 6(2): 100-108.

7. Jagadeesha, C. J. (1999). Water resources

development and management.

GISdevelopment, 3(6): 20-22.

8. Kite, G. W and Piteroniro, A. (1996). Remote

sensing applications in hydrological

modeling. Hydrol. Sci. J., 41(4): 561-591.

9. Koblinsky, C.J., Gaspar, P., Lagerloef, G.

(eds). 1992. The Future of Spaceborne

Altimetry: Oceans and Climate Change. Joint

Field Status of Application

Adequacy1

of

Scientific

Basis

Scientifically1

well Tested

?

Validation2

on Pilot

Schemes

?

Practical3

Applications

Major4

Constraints

for Practical

Applications

Water Resources Assessment

*Groundwater Good Good Adequate Standard/Part Administrative

*Surface Water Very Good Very Good Adequate Standard/Part Administrative

Irrigation Good Good Partially Very Limited Techno/Admin

Soil Erosion Fair Fair Very

Limited

Nil Science

Surface Water

Pollution

Good Good Adequate Some Cases Administrative

Groundwater Pollution

*Point Source

(Landfills)

Good Good Partially Standard/Part Techno/Admin

*Non-point

(Agriculture)

Fair Fair Very

Limited

Very Limited Techno/Admin

Effect of Land Use Changes

* Flows Good Fair Fair Very Limited Science

*Water Quality Fair Fair Fair Nil Science

Oceanographic Institutions Incorporated:

Washington, DC; 75 pp.

10. Ministry of Finance, (2002). Union budget

and economic survey, 2002-2003.

Government of India, New Delhi, India.

11. Swaminathan, M.S. (2000) Natural resources

management- for an evergreen revolution. In

The Hindu - Survey of Indian Agriculture

2000. pp. 9-16.

12. Ongley, E. D. (1996). Control of water

pollution from agriculture. Irrig. and Drain.

Paper No. 55, Food and Agric. Organization,

Rome.

13. Palanisami, K., 1984. Irrigation water

management : The determinants of canal

water distribution in India - A micro

analysis. Agricole Publishing Company,

New Delhi, 120 p.

14. Pietroniro, A. and Leconte, R. (2000). A

review of Canadian remote sensing

applications in hydrology, 1995-1999.

Hydrol. Process., 14: 1641-1666.

15. Refsgaard, J. C. and Storm, B. (1995). MIKE

SHE. Computers Models in Watershed

Hydrology, V. P. Singh (ed.), Water

Resources Publications, Colorado, USA,

806-846.

16. Sanmugnathan, K. and Bolton, P. (1988).

Water management in third world irrigation

schemes. ODA Bulletin, No. 11, Hydraulic

Research, London, UK.

17. Vachaud, G. and Chen, T. (2002). Sensitivity

of a large-scale hydrologic model to quality

of input data obtained at different scales;

Distributed versus stochastic non-distributed

modeling. J. Hydrol., 264: 101–112.

EVAPOTRANSPIRATION: TOOLS AND TECHNIQUES

ROHIT PRADHAN

Land Hydrology Division

Geosciences, Hydrology, Cryosphere Sciences and Applications Group (EPSA)

Space Applications Centre, ISRO

Ahmedabad-380015

Evapotranspiration(ET) is the combination of two processes where water is lost on one hand from

soil surface by evaporation and on the other hand from vegetation by transpiration.

Evapotranspiration forms a key component of the hydrological cycle which directly influences

the climate system. Estimation of ET is also important for management of irrigation in agricultural

systems, for addressing water balance in catchments of reservoirs and rivers etc. ET and soil

moisture greatly influence the climate and are now being used extensively for improving weather

forecasts. This note provides a brief introduction to the processes of evaporation and transpiration

and its influencing parameters followed by the various models used in estimation of potential and

actual ET over land surface. This is followed by a section on estimating ET using remotely sensed

data.

Key Words: Evapotranspiration, Potential ET, evaporation models, remote sensing.

1. Introduction

As the name suggests, ET involves two

components: (a) Evaporation is the process

whereby liquid water is converted to water

vapour (vaporization) and removed from the

evaporating surface (vapour removal). Water

evaporates from a variety of surfaces, such as

lakes, rivers, soil and wet vegetation. (b)

Transpiration consists of the vaporization of

liquid water contained in plant tissues and the

vapour removal to the atmosphere. Plants

predominately lose their water through stomata

(stomata are small openings on the plant leaf

through which gases and water vapour pass).

Nearly all water taken up by plants is lost by

transpiration and only a tiny fraction is used

within the plant.

Evaporation and transpiration occur

simultaneously and there is no easy way of

distinguishing between the two processes. The

driving force to remove water vapour from the

evaporating surface is the difference between the

water vapour pressure at the evaporating surface

and that of the surrounding atmosphere.

1.1 Factors affecting evapotranspiration:

(i) Weather parameters: Net radiation, air

temperature, humidity and wind speed.

(ii) Crop factors: Differences in resistance

to transpiration, crop height, crop

roughness, albedo and crop rooting

characteristics.

(iii) Environmental conditions: Ground

cover, plant density and the soil water

content.

The evaporation process over any vegetated

landscape is linked by two fundamental

equations:

(a) Water balance: Evapotranspiration can

be determined by measuring the various

components of the soil water balance.

This approach consists of assessing the

incoming and outgoing water flux into

the crop root zone over some time period.

P = Eact + Q + ∆S

Where, P is rainfall, Eact is actual

evaporation, Q is runoff and ∆S is change

in moisture storage of soil.

(b) Energy balance: Evaporation of water

requires relatively large amounts of

energy. The process of evaporation is

governed by energy exchange at the

vegetation/soil surface and is limited by

the amount of energy available. The

amount of energy arriving at a surface

must equal the energy leaving the surface

for the same time period.

R = H + λEact + G

Where, R is net radiation received at

surface, H is sensible heat flux, λEact is

outgoing energy as actual evaporation

and G is heat conduction into soil. λ is the

latent heat of vaporization.

ET rate is expressed as millimeters (mm) per unit

time. This rate expresses the amount of water lost

from a surface in units of water depth. For e.g. if

we say that ET is 1 mm/day it implies that 10

cubic meter of water is lost per hectare per day

from that place. In terms of energy, 2.45 MJ of

energy is required to vaporize 1 kg water at 20oC.

This is the value of latent heat of vaporization i.e.

the amount of energy required to vaporize 1 kg of

water at given temperature.

2. Measurement and Estimation of ET

ET can be measured at field level using

instruments like lysimeters, evaporimeters, eddy

flux towers etc. However, these are expensive

and cumbersome to maintain and provide point

measurements of ET which is not valid over

larger areas. To overcome this issue, various

models have been developed over the years for

estimation of ET. These models range from

simple empirical equations to the much advanced

radiation based models used today. Most of these

models are region-specific and the empirical

coefficients apply to those locations only limiting

their use in other regions. However, these models

have been widely used by hydrologists all around

the world for estimation of ET in their study area

and later calibrated for those regions.

Before proceeding to the various models

employed for estimating ET, it is necessary to

define some of the basic terms used frequently in

ET modelling. The terms and their short forms

given below will be used throughout this section.

Potential Evapotranspiration (PET): Dingman

(1992) defines PET as the rate at which

evapotranspiration would occur from a large area

completely and uniformly covered with growing

vegetation which has access to an unlimited

supply of soil water, and without advection or

heating effects.

Reference crop Evapotranspiration: It is the

evapotranspiration from a crop with specific

characteristics and which is not short of water.

FAO-56 adopts the specific characteristics of a

reference crop with certain height (0.12 m),

surface resistance (70 s m-1) and albedo (0.23)

and then determines the reference ET using the

Penman-Monteith Equation.

Actual Evapotranspiration (AET): AET is

defined as the quantity of water transferred as

water vapour to the atmosphere from an

evaporating surface. This surface can refer to

anything from the real world eg open lake, bare

soil, vegetated surface etc.

2.1 Models for estimating PET

A brief description of some of the most widely

used models for estimating potential and crop

reference ET is given below:

(a)Thornthwaite (1948): In the Thornthwaite

evaporation method, the only meteorological data

required to compute mean monthly potential

evapotranspiration is mean monthly air

temperature. It is expressed by:

𝐸𝑡ℎ,𝑗 = 16 (ℎ𝑟

12) (

ⅆ𝑎𝑦

30) (

10𝑇𝑗

𝐼)

𝛼𝑡ℎ

Where, Eth,j is the estimate of PET for month j, hr

is the mean daylight hours in month j, day is

number of days in month j, Tj is the mean

monthly air temperature (in oC) and I is annual

heat index. Once the mean monthly temperature

is known, the mean monthly PET can directly be

derived for each of the twelve months of a year.

(b)Penman (1948): Penman used an energy

equation based on net incoming radiation. This

approach does not require the surface temperature

variable.

Where, EPen is daily PET from a saturated surface,

Rn is net daily radiation to the evaporating

surface, Ea is a function of daily average wind

speed and vapour pressure, ∆ is the slope of

vapour pressure curve at air temperature, γ is the

psychrometric constant and λ is the latent heat of

vaporization.

(c)Penman-Monteith (1981): This is the most

widely used model for estimating PET from a

vegetated surface. It is expressed as:

Where, ETPM is the Penman-Monteith PET, Rn is

net daily radiation at the vegetated surface, G is

soil heat flux, ρa is mean air density at constant

pressure, ca is specific heat of the air, ra is

aerodynamic resistance, rs is surface resistance.

FAO has provided an excellent guide for

estimating crop reference ET in agricultural

regions based on Penman-Monteith equations.

This was formulated to make a universal method

of estimating PET. It is called the FAO-56 model

and is represented by the equation:

Where ETRC is the reference crop ET, and Ta is

mean daily air temperature. All other variables

have the same meaning as mentioned in previous

models.

(d)Priestley-Taylor (1972): This model computes

PET in terms of energy fluxes without an

aerodynamic component using the following

equation:

Where, EPT is the Priestley-Taylor PET, αPT is the

Priestley-Taylor constant which was taken as

1.26 in the original.

(e)FAO-24 Blaney-Criddle: The FAO-24

Reference Crop version of Blaney - Criddle is

defined as:

Where, ETBC is the Blaney-Criddle reference crop

ET, RHmin is minimum relative daily humidity,

n/N is measured sunshine hours to possible

sunshine hours, py is percentage of actual daytime

hours for the day compared to the daylight hours

for the entire year. ei (i=0 to 5) are the empirical

coefficients.

(f)Turc (1961): The Turc method is one of the

simplest empirical equations used to estimate

reference crop ET.

Where, ETTurc is Turc reference ET, Rs is the

incoming solar radiation and Ta is the average air

temperature.

(g)Hargreaves-Samani (1985): The HS equation

which estimates reference crop ET is as follows:

Where, ETHS is the reference crop ET, CHS is

empirical coefficient, Ra is extra-terrestrial

radiation, Tmax, Tmin and Ta are the maximum,

minimum and average air temperatures

respectively. This equation is used for weekly or

monthly time scales for better results.

(h)Morton (1985): Morton’s approach uses

iterative solution of the following two energy-

balance and vapour transfer equations for PET at

the equilibrium temperature.

Where ETMo is the Morton’s estimate of PET, fv is

vapour transfer coefficient and is a function of

atmospheric stability, ɛs is the surface emissivity,

σ is Stefan-Boltzmann constant, Te and Ta are

equilibrium and air temperatures. This is a part of

the CRAE model which computes both PET and

AET.

The above mentioned models showcase only a

few of the ones being used today for estimation

of PET. Hydrologists continuously make

improvements in these models to better represent

their study area by conducting vigorous

calibration or introducing subsequent correction

terms.

2.2 Models for estimating AET

Once the potential of evaporation is determined

at a place using weather data, AET or actual

evapotranspiration is computed by incorporating

the limiting factors like actual soil moisture, state

of vegetation on ground etc. The following

models are used regularly for AET estimation:

(a)Morton models: Bouchet (1963) stated that

PET and AET depend on one another via

feedback from land and atmosphere

simultaneously, given that the area is sufficiently

large and homogeneous with no or little advective

heat and moisture. This complementary

relationship (CR) is given by:

ETAct = 2 ETWet - ETPot

ETwet is potential or wet environment

evapotranspiration and ETpot is the point

potential ET at a place whose area is so small that

its heat and water vapour fluxes have no effect on

the overpassing air.

Morton’s CRAE (Complementary Relationship

Areal Evapotranspiration) model computes AET

for land environments. It calculates point

potential ET (ETPot) using the equation in

previous section and wet-environment areal

evapotranspiration or ETwet by modifying the

Priestley-Taylor approach as:

Where, ETwet

Mo is wet-environment areal ET, Rne

is net radiation for the soil/plant surface at

equilibrium temperature, p is atmospheric

pressure and ∆e is slope of saturation vapour

pressure curve at equilibrium temperature, b1 and

b2 are empirical coefficients.

(b)Chen and Dudhiya (2001): This was

developed as a part of a coupled land surface-

hydrology model in the Penn state-NCAR fifth-

generation Mesoscale Model (MM5). The PET is

computed using Penman-based energy balance

approach. AET is computed as sum of direct

evaporation from soil, wet canopy evaporation of

intercepted water and canopy transpiration. These

components are derived as a fraction of PET

based on the vegetation fraction, canopy

resistance and few other parameters.

𝐸𝑑𝑖𝑟 = (1 − 𝜎𝑓)𝛽𝐸𝑃

β =θ − θw

θref − θw

𝐸𝐶 = 𝜎𝑓𝐸𝑝 (𝑤𝐶

𝑠)

𝑛

𝐸𝑡 = 𝜎𝑓𝐸𝑝𝐵𝐶 (1 − (𝑤𝐶

𝑆)

𝑛)

Where, Edir is the direct evaporation from soil, σf

is the vegetation fraction, Ep is potential ET, β

determines the availability of water for

evaporation as a function of soil moisture θ, θref

and θw are field capacity and wilting point for that

soil type. Ec is wet canopy evaporation

(intercepted water), Et is canopy

evapotranspiration, Wc is intercepted canopy

water content, S is maximum canopy capacity, Bc

is a function of canopy resistance.

(c)FAO-56: FAO-56 method (Allen et al 1998)

for estimation of ET in agricultural areas involves

the computation of PET using Penman-Monteith

approach and then experimentally determined

ratios of ETc/ETo, called crop coefficients (Kc),

are used to relate AET to PET.

𝐸𝑇𝑐 = 𝐾𝑐𝐸𝑇0

Where, ETc is the crop ET, Kc is crop coefficient

and ETo is potential ET. Due to variations in the

crop characteristics throughout its growing

season, Kc for a given crop changes from sowing

till harvest. The effects of characteristics that

distinguish field crops from the reference grass

crop are integrated into the crop coefficient Kc.

Instead of using one value of Kc, dual crop

coefficients can also be used to distinguish soil

and crop ET. FAO-56 provides a handbook

detailing the entire procedure and experimental

values of Kc for different crops at various stages

of growth. This is the most widely used procedure

for estimating ET throughout the world.

Other notable models for AET include Granger-

Gray model (1989) and Szilagyi-Jozsa model

(2008). For more details on the above mentioned

models, including solved examples, please refer

to MacMahon et al 2013 (main paper and

supplementary material).

3. Remote Sensing and ET

In the previous section, many widely used models

were discussed in brief. When we wish to apply

such models over a large area, say a state or a

country, it becomes impractical to use

meteorological data from ground stations for ET

estimation as they are not well distributed over

the country. Remote sensing satellites orbiting

around Earth can provide a holistic view of an

entire region and can be used to estimate various

meteorological parameters. These satellite

observations are well distributed over space and

time and hence prove to be a viable tool for

estimating ET for large regions.

Satellites DO NOT provide any direct measure of

ET. Instead, they measure different

meteorological variables and land surface /

vegetation parameters that are then used in

different models for estimation of ET.

Space-based remote sensing satellites are

categorized based on their orbits into two types:

polar orbiting and geostationary. Polar-orbiting

satellites (e.g. ResrouceSAT-2, LANDSAT

series, Aqua/Terra-MODIS etc.) are placed

usually at ~500-800 km altitude and move around

the earth to capture images of the whole earth

surface. But the revisit time (i.e. the time interval

between successive observations of same region

on Earth’s surface) of such polar satellites is

anywhere from 2 days to 30 days depending on

the swath covered. These satellites provide

variables like leaf area index, snow cover, land

use-land cover etc. which can be used in the

models mentioned in previous section.

Geostationary satellites stay stationary relative to

a fixed point on earth surface at an approximate

altitude of 36,000 km. These type of satellites

provide continuous data recording over a fixed

coverage area at frequent time intervals.

Geostationary Indian weather satellites such as

Kalpana-1, INSAT 3D and INSAT 3DR estimate

parameters like rainfall and earth’s radiation

budget using its various spectral channels and

provide data every 30 minutes. This data can be

used to estimate parameters like land surface

temperature, down-welling shortwave radiation,

upwelling longwave radiation, cloud cover etc. to

determine the surface energy budget and then be

implemented in the PET estimation models.

A few of the notable remote sensing based ET

estimation methodologies are the MOD16

MODIS ET algorithm (Mu et al 2007,2011),

SEBEL (Bastiaanssen et al 1998, Bastiaanssen

2000), METRIC (Allen et al 2007) and

methodologies by Batra et al 2006, Cleugh et al

2007, Yao et al 2013 etc. For further study on

remote sensing based ET estimation please refer

to the above papers. Following are two case

studies of operational ET estimation methods

using satellite data.

3.1 Case Study: MOD16 Algorithm

Estimation of PET/AET from space based

platforms can be best understood with the help of

a case study. NTSG employs a standard

methodology to compute ET using MODIS

(Moderate-resolution Imaging Sensor) onboard

Terra/Aqua satellites (Mu et al 2007, 2011).

This methodology uses the following set of

inputs:

(a) Remote-Sensing Derived Inputs: Land cover,

Leaf Area Index (for vegetation fraction),

Albedo, FPAR (Fraction of absorbed

Photosynthetically Active Radiation). These

products are available at 8/16 day intervals

and derived from MODIS.

(b) Meteorological Inputs: Air pressure, air

temperature, humidity, solar radiation. These

inputs are taken at daily basis from global

weather forecasting models calibrated against

ground stations.

The first step involves partitioning of incoming

solar radiation into net radiation available to

plants and net radiation to soil. This is done by

computing the vegetation fraction using the LAI

parameter. Potential soil evaporation is computed

using the Penman-Monteith approach which

utilizes meteorological parameters.

Plant transpiration is computed by first estimating

canopy conductance, water cover fraction,

aerodynamic resistance and plant intercepted

radiation. Then, plant transpiration is computed

using a Penman-Monteith based approach.

𝜆𝐸𝑡𝑟𝑎𝑛𝑠

=(𝑠 𝐴𝑐𝐹𝑐 + 𝜌 𝐶𝑝(𝑒𝑠 − 𝑒)

𝐹𝑐𝑟𝑎

) ∗ (1 − 𝐹𝑤𝑒𝑡)

𝑠 + 𝛾 (1 + 𝑟𝑠

𝑟𝑎⁄ )

Where, s is slope of saturated vapour pressure

curve, Ac is energy available to plant canopy, Fc

is vegetation fraction, Cp is specific heat of air, es

is saturated vapour pressure, rs is surface

resistance and ra is aerodynamic resistance, Fwet is

wet surface fraction and is a function of relative

humidity.

Wet canopy evaporation is computed by using

LAI and wet fraction to estimate wet canopy

aerodynamic resistance. Following the Biome-

BGC model λEwet_can is computed.

The soil surface is divided into saturated surface

and moist surface by the parameter Fwet. The

potential soil evaporation is computed as sum of

evaporation from saturated and moist soil

surfaces. The actual soil ET is computed using the

complementary hypothesis by Bouchet (1963).

𝜆𝐸𝑠𝑜𝑖𝑙 = 𝜆𝐸𝑤𝑒𝑡_𝑠𝑜𝑖𝑙 + 𝜆𝐸𝑝𝑜𝑡_𝑠𝑜𝑖𝑙(𝑅𝐻

100)𝑉𝑃𝐷/𝛽

Where, 𝜆𝐸𝑠𝑜𝑖𝑙 is total actual soil ET, 𝜆𝐸𝑤𝑒𝑡_𝑠𝑜𝑖𝑙 is

AET due to wet soil, 𝜆𝐸𝑝𝑜𝑡_𝑠𝑜𝑖𝑙is the potential ET

from moist soil surface (unsaturated), RH is

relative humidity in %age, VPD is vapour

pressure deficit and β is set as 200.

The total daily ET (λE) is computed as sum of

evaporation from wet canopy surface, the

transpiration from dry canopy surface and

evaporation from soil surface.

𝜆𝐸 = 𝜆𝐸𝑤𝑒𝑡_𝑐𝑎𝑛 + 𝜆𝐸𝑡𝑟𝑎𝑛𝑠 + 𝜆𝐸𝑠𝑜𝑖𝑙

This MODIS ET product has been extensively

validated by using Eddy covariance towers from

FLUXNET (Mu et al 2011). The following global

map of mean annual ET during 2000-2006 was

produced by the NTSG using the above

mentioned algorithm.

Fig.1. Mean annual ET during 2000-2006

(Adapted from Mu et al 2011).

Fig 2. Flowchart of MODIS ET algorithm

(Adapted from Mu et al 2011).

3.2 Case study: Indian Perspective

Bhattacharya et al (2010) developed a simplified

single-source energy balance scheme to estimate

ET. Indian geostationary satellite Kalpana-1’s

Very-High Resolution Radiometer (VHRR) data

was used to obtain the major inputs for the ET

model, namely Land Surface Temperature (LST),

surface albedo, insolation and air temperature.

This methodology is based on an energy balance

approach. The available energy at surface is given

as sum of latent and sensible heat fluxes. A

parameter called evaporative fraction (Λ) is

introduced as the ratio of latent flux and total

available energy.

Λ = 𝜆𝐸

(𝜆𝐸 + 𝐻)

Subsequently, λE is estimated by multiplying the

net energy with this evaporative fraction term.

𝜆𝐸 = (𝑅𝑛 − 𝐺) 𝛬

Λ term is computed using the relationship

between LST and albedo and their relationship

with soil moisture. For wet soil, its albedo is 2-3

times lower compared to dry soil. Surface albedo

determines the outgoing shortwave radiation.

This method uses a third order polynomial

relationship between LST, Soil moisture and

albedo as stated in Bastiaanssen et al (1998).

Fig 3. A concept figure of LST (TS) vs surface

albedo curve. At a given albedo, Dry edge (DC)

and wet edge (EF) represent maximum (TH) and

minimum (TE) LST lines. (Adapted from

Bhattacharya et al 2010).

The evaporative fraction at a given time in day

(noontime) is approximated by:

𝛬 =𝑇𝐻 − 𝑇𝑆

𝑇𝐻 − 𝑇𝐸

This term is multiplied to the net radiation to

obtain actual ET over Indian region. Figure 4

shows the output of this methodology at 0.08o

resolution.

Figure 4. Estimated AET in mm/day for India for

two 8-day periods in Nov and Dec 2005. Adapted

from Bhattacharya et al 2010.

4. Future Prospects

Most models for ET estimation rely on other

proxy parameters to solve one of the two

fundamental equations. This results in significant

errors in estimated ET.

The state-of-art techniques in the field of ET, for

e.g. Jasechko et al 2013, use the stable isotope

ratios of oxygen (18O/16O) and hydrogen (2H/1H)

to separate transpiration from evaporation. It

relies on the fact that evaporation process results

in enrichment of heavy isotopes of O and H in the

leftover water. However, transpiration process

does not produce fractionation of these isotopes.

Their analysis of catchments of some major lakes

of the world has shown that terrestrial water flux

is dominated by transpiration and not

evaporation.

ET plays a major feedback role in the land-

atmosphere system, thus affecting the global

climate. As the climate warms up, it is expected

that global evaporation losses will increase.

However, in a recent study by Jung et al 2010, the

authors have shown that there is indeed a decline

in global land ET owing to reduced moisture

supply. They computed and analyzed global ET

data for 27 years using integrated flux tower

measurements and remote sensing inputs. Also,

they related the impact of major El-Nino to the

changes in spatio-temporal behavior of ET.

Upcoming ISRO mission, GISAT (Geostationary

Imaging SATellite), is an advanced earth-

observation satellite. As the name suggests,

GISAT will be a geostationary satellite providing

high resolution multi-spectral and hyper-spectral

observations over India in optical, near-infrared

and thermal wavelengths, multiple times in a day.

This satellite will be capable of providing all the

necessary parameters required for estimation of

ET over India on a daily basis. This will greatly

improve the evaporation estimates over Indian

region and will help us better understand ET and

its associated processes.

Evapotranspiration is not easy to measure and

even today it has huge potential for

improvements. Symons in 1867 best described

estimating evaporation as “...the most desperate

art of the desperate science of meteorology”

(Monteith, 1997). With the little we know and the

vast unknowns left to explore, ET is definitely

one of the most challenging and exciting

branches of Hydrology.

5. References

1. Allen, R. G. and Pruitt, W. O. (1986).

Rational use of the FAO Blaney-Criddle

Formula. J. Irrig. Drain. E. ASCE, 112, 139–

155.

2. Allen, R. G., Pereira, L. S., Raes, D., and

Smith, M. (1998). Crop evapotranspiration

Guidelines for computing crop water

requirements, FAO Irrigation and Drainage

Paper 56, Food and Agriculture Organization

of the United Nations.

3. Allen, R. G., Tasumi, M., and Trezza, R.

(2007). Satellite-based Energy Balance for

Mapping Evapotranspiration with

Internalized Calibration (METRIC)—Model.

Journal of irrigation and drainage

engineering, 133(4), 380-394.

4. Bastiaanssen, W. G. M. (2000). SEBAL-

based Sensible and Latent Heat Fluxes in the

Irrigated Gediz Basin, Turkey. Journal of

hydrology, 229(1), 87-100.

5. Bastiaanssen, W. G. M., Menenti, M.,

Feddes, R. A., and Holtslag, A. A. M. (1998).

A Remote Sensing Surface Energy Balance

Algorithm for Land (SEBAL): Part 1:

Formulation. Journal of hydrology, 212, 198-

212.

6. Batra, N., Islam, S., Venturini, V., Bisht, G.,

and Jiang, L. E. (2006). Estimation and

Comparison of Evapotranspiration from

MODIS and AVHRR Sensors for Clear Sky

Days over the Southern Great Plains. Remote

Sensing of Environment, 103(1), 1-15.

7. Bhattacharya, B.K., Mallick, K., Patel, N.K.,

and Parihar, J.S. (2010). Regional clear sky

evapotranspiration over agricultural land

using remote sensing data from Indian

geostationary meteorological satellite.

Journal of Hydrology, 387, 65-80.

8. Bouchet, R. J. (1963). Evapotranspiration

reelle et potentielle, signification climatique,

International Association of Hydrological

Sciences Publ., 62, 134–142.

9. Chen, F. and Dudhiya, J. (2001). Coupling an

Advanced Land Surface-Hydrology Model

with the Penn State-NCAR MM5 Modeling

System. Part I: Model implementation and

sensitivity. Monthly Weather Review, 129,

569-585.

10. Cleugh, H. A., Leuning, R., Mu, Q., and

Running, S. W. (2007). Regional evaporation

estimates from flux tower and MODIS

satellite data. Remote Sensing of

Environment, 106(3), 285-304.

11. Dingman, S. L. (1992). Physical Hydrology,

Prentice Hall, Upper Savage, New Jersey.

12. Granger, R. J. and Gray, D. M. (1989).

Evaporation from natural non-saturated

surfaces, Journal of Hydrology. 111, 21–29.

13. Hargreaves, G. H. and Samani, Z. A. (1985).

Reference crop evapotranspiration from

temperature, Appl. Eng. Agric., 1, 96–99.

14. Jasechko, S., Sharp, Z.D., Gibson, J.J., Birks,

S.J., Yi, Y., and Fawcett, P.J. (2013).

Terrestrial water fluxes dominated by

transpiration. Nature Letter. 496.

doi:10.1038/nature11983.

15. Jung, M., Reichstein, M., Ciais, P.,

Seneviratne, S.I., Sheffield, J., Goulden,

M.L. et al (2010). Recent decline in the

global land evapotranspiration trend due to

limited moisture supply. Nature Letter. 467,

951-954.

16. McMahon, T.A., Peel, M.C., Lowe, L.,

Srikanthan, R., and McVicar, T.R. (2013).

Estimating actual, potential, reference crop

and pan evaporation using standard

meteorological data: a pragmatic synthesis.

Hydrol. Earth Syst. Sci., 17, 1331-1363.

17. Monteith, L.J. (1997). Evaporation Models.

In: Agricultural Systems Modelling and

Simulation. Edited by Robert M. Peart and R.

Bruce Curry. University of Florida,

Gainesville, Florida: 197-234.

18. Monteith, L.J. (1981). Evaporation and

surface temperature, Q. J. Roy. Meteor. Soc.,

107, 1–27.

19. Morton, F. I., Richard, F., and Fogarasi, S.

(1985). Operational estimates of areal

evapotranspiration and lake evaporation –

Program WREVAP, NHRI Paper 24, Inland

Waters Directorate, Environment Canada,

Ottawa.

20. Mu, Q., Heinsch, F. A., Zhao, M., and

Running, S. W. (2007). Development of a

Global Evapotranspiration Algorithm Based

On MODIS and Global Meteorology Data.

Remote Sensing of Environment. 111(4),

519-536.

21. Mu, Q., Zhao, M., and Running, S. W.

(2011). Improvements to a MODIS global

terrestrial evapotranspiration algorithm.

Remote Sensing of Environment. 115(8),

1781-1800.

22. Penman, H. L. (1948). Natural evaporation

from open water, bare soil and grass, Proc. R.

Soc. Lond. A. 193, 120–145.

23. Priestley, C.H.B. and Taylor, R.J. (1972). On

the assessment of the surfaces heat and

evapotranspiration using large-scale

parameters. Monthly Water Review.

100(2):81-92.

24. Szilagyi, J. and Jozsa, J. (2008). New

findings about the complementary

relationship-based evaporation estimation

methods. Journal of Hydrology. 354, 171–

186.

25. Thornthwaite, C. W. (1948). An approach

toward a rational classification of climate.

Geogr. Rev., 38, 55–94.

26. Turc, L. (1961). Estimation of irrigation

water requirements, potential

evapotranspiration: A simple climatic

formula evolved up to date, Ann. Agronomy,

12, 13–49.

27. Yao, Y., Liang, S., Cheng, J., Liu, S., Fisher,

J.B., Zhang, X., Jia, K. et al. (2013). MODIS-

driven Estimation of Terrestrial Latent Heat

Flux in China Based on a Modified Priestley-

Taylor Algorithm. Agricultural and Forest

Meteorology. 171, 187-202.

WATER QUALITY MONITORING FROM SPACE

ASHWIN GUJRATI

Land Hydrology Division

Geosciences, Hydrology, Cryosphere Sciences and Applications Group (EPSA)

Space Applications Centre, ISRO

Ahmedabad-380015

Water provides basic resources for various human uses and diverse habitat ecosystem services, supporting

high levels of biodiversity. Thus, it is paramount that water quality is assessed every so often to determine

suitability and safety for varying purposes. Monitoring and understanding the water quality i.e. physical,

chemical and biological status of global water is immensely important to scientists and policy makers.

Key Words: Water Quality, water color, turbidity, radiative transfer.

1. Introduction

Water quality of any system can be measured by

the following physical, chemical and biological

parameters. The physical data includes, pH,

temperature, dissolved oxygen, turbidity, Secchi

depth, specific conductivity. Chemical analysis

includes concentrations of Silicate, Nitrate,

Nitrite, Ammonium, total Nitrogen and total

Phosphorus. Biological parameter includes

chlorophyll, dissolved organic matter and

suspended particulate matter.

The in-situ measurements of water quality are

often very scarce because of large areas to

monitor. Furthermore, these measurements do

not represent the actual water quality at a large

scale since measurements are restricted to

specific regions. Consequently, one may consider

measurement techniques so as to get relevant

information especially at large scales and to be

able to characterize water quality over a whole

region. In this context, remote sensing from space

is a perfect tool to get the required information.

Satellite data may be able to provide a greater

amount of spatial information at an improved cost

compared to spot sample grabs.

1.1 Remote sensing of water

Conventional monitoring approaches tend to be

limited in terms of spatial coverage and temporal

frequency. Remote sensing has the potential to

provide an invaluable complementary source of

data at local to global scales. But remote sensing

of water can measure only those water quality

parameters that have optically active constituents.

Constitutes that interact with light and changing

the energy spectra of reflected solar radiation

emitted from surface waters (Ritchie et al., 2003).

These include phytoplankton pigments

(chlorophylls, carotenoids, phycocyanin, etc.),

colored dissolved organic matter (CDOM), and

inorganic and non-living suspended matter,

which coincide well with the previously

mentioned parameters determining the majority

of water quality issues in inland waters.

The measured radiance originates from sunlight

that passes through the atmosphere, is reflected,

absorbed, and scattered by constituents in the

water bodies, and is transmitted back through the

atmosphere to the satellite-based sensor (Fig. 1)

(http://www2.dmu.dk/resc-

oman/project/Backgrounds/challenges.htm). The

processes of scattering and absorption by

optically active constituents in the water affect

the spectrum and radiance distribution (light

field) of the light emerging from the water – the

so called water-leaving radiance. The scattering

and absorption characteristics of water and its

constituents are described as the inherent optical

properties (IOPs). The spectral quality and

quantity of the water-leaving radiance is largely

determined by the inherent optical properties.

The modification/alteration of the radiance has

been used to determine water constituents,

typically the desired parameter has been the

chlorophyll-a concentration, colored dissolved

organic matter (CDOM), and inorganic and non-

living suspended matter. In essence, water colour

is determined by inherent optical properties.

Fig.1: Schematic diagram of Remote sensing of

water

Inherent optical properties (IOP's) are those

properties that depend only upon the medium,

and therefore are independent of the ambient light

field within the medium. The two fundamental

IOP's are the absorption coefficient and the

volume scattering function. Other IOP's include

the index of refraction, the beam attenuation

coefficient and the single-scattering albedo.

Apparent optical properties (AOP's) are those

properties that depend both on the medium (the

IOP's) and on the geometric (directional)

structure of the ambient light field, and that

display enough regular features and stability to be

useful descriptors of the water body. Commonly

used AOP's are the irradiance reflectance, the

average cosines, and the various diffuse

attenuation coefficients.

Remote sensing of water is broadly divided into

retrieval over Case 1 waters and Case 2 waters.

Case 1 waters are waters in which the

concentration of phytoplankton is high compared

to non-biogenic particles. Absorption by

chlorophyll and related pigments therefore plays

a major role in determining the total absorption

coefficient in such waters, although detritus and

dissolved organic matter derived from the

phytoplankton also contribute to absorption in

case 1 waters. Case 1 water can range from very

clear (oligotrophic) water to very turbid

(eutrophic) water, depending on the

phytoplankton concentration. Case 2 waters are

"everything else," namely waters where inorganic

particles or dissolved organic matter from land

drainage dominate, so that absorption by

pigments is relatively less important in

determining the total absorption.

2. Literature Review

Two types of methods are commonly used for

interpreting water quality from remotely sensed

data: empirical and analytical approach (Bhatti et

al. 2010; Cannizzaro and Carder 2006; Giardino

et al. 2007; Kallio 2000; Knaeps et al. 2010;

Ritchie et al. 2003). The empirical based

approaches are most commonly used method

which are determined through statistical

relationships between measured spectral

properties (i.e. radiance or reflectance) versus the

measured water quality parameter of interest (e.g.

Lee et al. 1996; Garver & Siegel, 1997; Hoge &

Lyon, 1996, 2005; Le et al., 2009a; Ritchie et al.

2003; Wang et al., 2005; Bhatti et al. 2010).

Usually algorithm development searches for a

combination of radiance signals at several

wavelengths to find ratio, or other combination,

that relates radiance at particular band

empirically to the desired water quality

parameter. The coefficients contained in these

algorithms are generally derived by pooling data

collected at various spatial and temporal scales.

Empirical approaches are region dependent, that

works better for one site but may fail on other site.

On the other hand, analytical algorithms are

based on radiative transfer equations works

equally well for different water bodies and

usually perform better than the empirical

algorithm (L. Li el al.2013).

Recently many analytical and semi-analytical

algorithms for inland waters are developed for

retrieving inherent optical properties from remote

sensing reflectance. Gege (2012) developed an

analytic model for the direct and diffuse

components of the down-welling irradiance in the

water column. Giardino et al. (2012) developed a

software package incorporating their Bio-Optical

Model Based tool for Estimating water quality

and bottom properties from Remote sensing

images (BOMBER). Brando et al. (2012) present

an adaptive implementation of the linear matrix

inversion (LMI) method which accounts for

variability in both IOPs and mass-specific IOPs

(SIOPs) over space and time in wide-ranging

optically-complex waters. More sophisticated

neural network and physics-based inversion

methods have also been used to estimate in-water

inherent optical properties (IOPs) (Odermatt et

al., 2012). Salama and Verhoef (2014) present a

new, forward model analytical inversion solution

(“2SeaColor”) for the retrieval of the depth

profile of the down-welling diffuse attenuation

coefficient.

Remote sensing of inland water bodies poses a

challenge due to its highly complex optical nature

as compared to clear marine waters. Simulated

remote sensing reflectance spectra of water with

different water quality parameters plotted with

their true color is shown in figure 2. The optical

complexity of inland waters stems from the fact

that these waters are typically characterized by

high concentrations of phytoplankton biomass

(typically on the order of between 1 and 100 mg

m−3 chlorophyll-a (chl-a), and up to 350 mg m−3

(Gitelson et al., 1993), mineral particles, detritus

and CDOM that typically do not co-vary over

space and time. Moreover, their optical properties

are highly variable between and even within

water bodies.

Fig.2: Remote sensing reflectance spectra of

different water color.

Table 1: Variables that affect the water quality

parameters

Variables that

can affect

remote sensing

of physical

water-quality

characteristics

Variable Explanation

Time of year

The Earth receives 7 per

cent more energy from the

sun on 1 January than on 1

July because of an oval

orbit.

Sun-elevation

angle

More solar energy is

specularly reflected from

water surfaces at low sun-

elevation angles than at

high angles. Also,- the

path length of solar energy

through the atmosphere is

longer at low sun-

elevation angles, and more

solar energy is absorbed

and scattered.

Aerosol and

molecular content

of atmosphere

These constituents

determine the amount of

solar energy absorbed and

scattered by the

atmosphere. Some energy,

received by a satellite, is

backscattered before

reaching the water

surface.

Water-vapour

content of the

atmosphere

Water vapour affects

energy absorption at near

infrared and thermal

infrared wavelengths.

Specular

reflection of

skylight from

water surface

Specularly reflected

skylight is received by a

satellite. The intensity and

wavelength distribution of

this energy depends on

atmospheric scattering,

which produces skylight.

Roughness of

water surface

A rough surface may

produce more or less

specular reflection than a

smooth surface. At high

sun-elevation angles, the

area of sun glint may be

within the satellite fields

of view.

Film, foam,

debris, or floating

plants on water

surface

These features may not be

resolved on a satellite

image, but they contribute

to the spectral

characteristics of the

measured signal.

Water colour

Dissolved, coloured

materials increase

absorption of solar energy

in water.

Water turbidity

The concentration, size,

shape, and refractive

index of suspended

particles determine

turbidity and increase the

amount of energy

backscattered in water

bodies.

Reflectance and

absorptance

characteristics of

suspended

particles

Particles may be inorganic

sediments, phytoplankton,

zooplankton, or a

combination. When

present in high

concentrations, particles

affect the spectral

distribution of

backscattered energy.

Multiple

reflections and

scattering of solar

energy in water

The spectral results of

these processes are

difficult to predict, but

may not be important.

Depth of water

and reflectance of

bottom sediments

Water clarity determines

the importance of bottom

reflectance. Solar energy

may not reach bottom in a

turbid water.

Submerged or

emergent

vegetation

Vegetation may change

bottom reflectance,

obscure water surface, or

contribute to the spectral

characteristics of the

measured signal.

3. Radiative Transfer Model

In optically shallow waters, the upwelling

irradiance just below the surface, Eu(0), results

from adding the flux backscattered by the water

column and the flux reflected by the bottom

substrate and then transmitted through the

column as shown in fig 3 below,

𝐸𝑢(0) = [𝐸𝑢(0)]𝐶 + [𝐸𝑢(0)]𝐵

The subscripts C and B stand for water column

and bottom. The first component corresponds to

the photons that have never interacted with the

bottom, whereas those that have interacted with

the bottom at least once form the second

component.

Fig.3: Schematic diagram of radiative transfer

model

To estimate the first component on RHS, we

consider an infinitely thin layer of uniform

thickness dZ at depth Z. At this level, the down-

welling irradiance is Ed(Z). The backscattering

coefficient (or reflectance function) for the down-

welling light stream is denoted bbd; the fraction of

upwelling irradiance created by this layer is:

𝑑𝐸𝑢(𝑍) = 𝑏𝑏𝑑𝐸𝑑(𝑍)𝑑𝑍

Ed(Z) can be expressed as

𝐸𝑑(𝑍) = 𝐸𝑑(0)exp(−𝐾𝑑𝑍)

Ed(0) is the downwelling irradiance at null depth,

and Kd is the diffuse attenuation for downwelling

irradiance. Before it reaches the surface, dEu(Z)

suffers an attenuation along the path from Z up to

0, expressed by exp(-kZ), where k is the vertical

diffuse attenuation coefficient for upward flux.

This coefficient refers to an exponential

attenuation (with distance travelled upward) of

the upward flux while travelling upward

originating from any thin layer. The contribution

of the considered layer to the upwelling

irradiance just below the surface is denoted

dEu(Z0); this term is expressed as

𝑑𝐸𝑢(𝑍 → 0) = 𝑏𝑏𝑑𝐸𝑑(0) exp[−(𝐾𝑑 + 𝑘)𝑍] 𝑑𝑍

The contributions of each layer between Z and 0

in forming Eu(0, Z) can be summed, so that

𝐸𝑢(0, 𝑍) = 𝑏𝑏𝑑𝐸𝑑(0)

× ∫ exp[−(𝐾𝑑 + 𝑘)𝑍] 𝑑𝑍𝑍

0

𝐸𝑢(0, 𝑍) = (𝐾𝑑 + 𝑘)−1𝑏𝑏𝑑𝐸𝑑(0) × [1− exp−(𝐾𝑑 + 𝑘)𝑍]

For an infinite water depth (Z=∞), above equation

reduces to

𝐸𝑢(0,∞) = (𝐾𝑑 + 𝑘)−1𝑏𝑏𝑑𝐸𝑑(0)

𝐸𝑢(0, 𝑍) = 𝑅(0,∞)𝐸𝑑(0)

𝑅(0,∞) =𝑏𝑏𝑑

(𝐾𝑑 + 𝑘)

R(0,∞)represents the reflectance at null depth of

the deep ocean, hereafter denoted R∞. For a

column limited by the presence of a perfectly

absorbing bottom at a depth H,

𝐸𝑢(0, 𝐻) = 𝑅∞𝐸𝑑(0)× [1 − exp−(𝐾𝑑 + 𝑘)𝑍]= [𝐸𝑢(0)]𝐶

and thus provides the first term in first equation.

If the bottom is a Lambertian reflector with an

albedo A, the reflected flux at level H (i.e.

immediately above the bottom) is

[𝐸𝑢(𝐻)]𝐵 = 𝐴 ×𝐸𝑑(𝐻)= 𝐴 × 𝐸𝑑(0)exp(−𝐾𝑑𝐻)

This contribution of the bottom to the upwelling

irradiance will be attenuated from H up to the

surface. If we suppose that this upward flux is

attenuated with the same K as above, the

contribution of the bottom to the upward

irradiance reaching the surface becomes

[𝐸𝑢(𝐻)]𝐵 = 𝐴 × 𝐸𝑑(0)exp[(−𝐾𝑑 + 𝑘)𝐻]

By adding, we obtain

𝐸𝑢(0) = 𝐸𝑑(0)(𝑅∞ × [1 − exp−(𝐾𝑑 + 𝑘)𝑍]+ 𝐴 × exp[(−𝐾𝑑 + 𝑘)𝐻])

Dividing by Ed(0) and rearranging, the

reflectance, R(0, H), below the surface of a

homogeneous ocean bounded below by a

reflecting bottom at depth H, is

𝑅(0,𝐻) = 𝑅∞ + (𝐴 − 𝑅∞)exp[(−𝐾𝑑 + 𝑘)𝐻]

To the extent that the two kinds of upward fluxes,

either scattered by the series of thin layers or

reflected by the bottom, do not have the same

geometrical structure, they are not attenuated in

the same way. If KB and KC denote the attenuation

coefficients for the upward streams originating

from the bottom and from the water column

respectively, Equation must be written as

𝑅(0,𝐻) = 𝑅∞ + exp(−𝐾𝑑H) ×[Aexp(−𝐾𝐵𝐻)− 𝑅∞exp(−𝐾𝐶𝐻)]

Authors have simplified the above equation for

implementation with various approximations

based on their study area and practice difficulties

of measurement. The coefficients were derived

using either Monte Carlo or hydro-light

simulations.

4. References

1. Bhatti, A.M., Schalles, J., Rundquist, D.,

Ramirez, L & Nasu, S. (2010). Accuracy

2010 Symposium, July20-23, Leicester, UK

2. Brando, V.E., Dekker, A.G., Park, Y.J., &

Schroeder, T. (2012). Adaptive

semianalytical inversion of ocean color

radiometry in optically complex waters.

Applied Optics, 51(15), 2808–2833.

3. Cannizzaro, J. P., and K. L. Carder. 2006.

Estimating chlorophyll a concentration from

remote-sensing reflectance in optically

shallow waters. Remote Sensing of

Environment 101 (1): 13-24.

4. Garver, S. A., & Siegel, D. A. (1997).

Inherent optical property inversion of ocean

color spectra and its biogeochemical

interpretation .1. Time series from the

Sargasso Sea. Journal of Geophysical

Research-Oceans, 102, 18607–18625.

5. Gege, P. (2012). Analytic model for the

direct and diffuse components of

downwelling spectral irradiance in water.

Applied Optics, 51(9), 1407–14019.

6. Giardino, C., V. E. Brando, A. G. Dekker, N.

Strombeck, and G. Candiani. 2007.

Assessment of water quality in Lake Garda

(Italy) using Hyperion. Remote Sensing of

Environment 109 (2): 183-195.

7. Gitelson, A., Garbuzov, G., Szilgyi, F.,

Mittenzwey, K. -H., Karnieli, A., & Kaiser,

A. (1993). Quantitative remote sensing

methods for real-time monitoring of inland

waters quality. International Journal of

Remote Sensing, 14, 1269–1295.

8. Hoge, F. E., & Lyon, P. E. (1996). Satellite

retrieval of inherent optical properties by

linear matrix inversion of oceanic radiance

models: An analysis of model and radiance

measurement errors. Journal of Geophysical

Research-Oceans, 101, 16631–16648.

9. Kallio, K. 2000. Remote sensing as a tool for

monitoring lake water quality. In

Hydrological and limnological aspects of

lake monitoring, ed. P. Heinonen, G. Ziglio,

and A. van der Beken, 237-245. Chichester,

England: John Wiley & Sons, Ltd.

10. Knaeps, E., D. Raymaekers, S. Sterckx, and

D. Odermatt. 2010. An intercomparison of

analytical inversion approaches to retrieve

water quality for two distinct inland waters.

In Proceedings of the Hyperspectral

Workshop. Frascati, Italy.

11. Le, C. F., Li, Y. M., Zha, Y., Sun, D. Y., &

Yin, B. (2009a). Validation of a quasi-

analytical algorithm for highly turbid

eutrophic water of Meiliang Bay in Taihu

Lake, China. IEEE Transactions on

Geoscience and Remote Sensing, 47, 2492–

2500.

12. Lee, Z. P., Carder, K. L., Peacock, T. G.,

Davis, C. O., & Mueller, J. L. (1996). Method

to derive ocean absorption coefficients from

remote-sensing reflectance. Applied Optics,

35, 453–462.

13. Odermatt, D., Gitelson, A., Brando, V.E., &

Schaepman, M. (2012). Review of

constituent retrieval in optically deep and

complex waters from satellite imagery.

Remote Sensing of Environment, 118, 116–

126.

14. Ritchie, J. C., P. V. Zimba, and J. H. Everitt.

2003. Remote sensing techniques to assess

water quality. Photogrammetric Engineering

& Remote Sensing 69 (6): 695-704.

15. Salama, M. S., & Verhoef, W. (2015). Two-

stream remote sensing model for water

quality mapping: 2SeaColor. Remote

Sensing of Environment, 157, 111-122.

16. Wang, P., Boss, E. S., & Roesler, C. (2005).

Uncertainties of inherent optical properties

obtained from semianalytical inversions of

ocean color. Applied Optics, 44, 4074–4085.

Digital Image Processing

Vibhuti Bhushan JhaSpace Applications Centre, Ahmedabad

In this lecture we will try to look into the domain of image processing and the mathematicalformulation of some image processsing tools. We look into different statistical image analysismethods and classification tools.

Introduction

Images obtained from space based observations openthe door for data and interpretations, the key to whichlies in the proper deciphering of the underlying processesand techniques. Image processing is a tool to extract thenot so obvious data. The objective of this lecture is tofamiliarize the audience with the mathematical foundationsof image processing and some basic image operations whichcan be utilised as of when needed in further lectures. Wewill look into different image operations like, sharpening,zooming, shrinking, filtering, histograms, classificationsetc. We will also look into the different types of imagesused in hydrological applications and how to extract somemeaningful information from them which will be followedby hands on tutorial on the use of different softwares andtechniques.This chapter has several objectives: (1) to definethe scope of the field that we call image processing; (2) togive a historical perspective of the origins of this field; (3)to give an idea of the state of the art in image processing byexamining some of the principal areas in which it is applied;(4) to discuss briefly the principal approaches used in digitalimage processing; (5) to give an overview of the componentscontained in a typical, general-purpose image processingsystem.

What is an image?

An image can be considered as a 2 dimensional functionf (x, y) where x, y are the spatial coordinates and f is a scalarvalues real function ∈ R which is dependent on the source ofillumination. One of the benefits of data acquired from spacebased platforms is availability in digital format. Spatially thedata is composed of discrete picture elements called pixels.From the data handling and analysis point of view, the prop-erties of image data of significance are the number and lo-cation of the spectral measurements provided by a particularsensor, the spatial resolution as described by the pixel size,and the radiometric resolution. The later describes the rangeand the discernible number of discrete brightness values. It isalso sometimes called as dynamic range. It is also expressed

in terms of the number of bits required to represent the rangeof available brightness values. Hence, data with 8 bit ra-diometric resolution has 256 levels of brightness values. Asimilar situation applies when using microwave image data:viz, several transmission wavelengths can be used to assist inidentification of cover types by reason of their different scat-tering behaviours with wavelength. A useful property whichcan be utilised for the image analysis is polarisation. Thereare different techniques for analysis of SAR imagery, namely,interferometry, polarimetry etc.

Types of Spatial Data

Before we embark on this journey of processing, we needto know the types of spatial data which we encounter in satel-lite imagery. The data must be available in discrete formspatially that is corresponding to the pixels, with each pixeldescribing the properties at the ground. Secondly,the imagemust be georeferenced that is should correspond to proper latlong values. It has to be mutually registered and referencedto a a base map like UTM.

Image registration

It is important to georegister the image with some baseimage or map. This tries to establish a mathematical rela-tionships between the addresses of a pixel in an image andthe corresponding coordinates of those points on the groundvia a map. But this has an inherent assumption that a basemap is available of the desired location. A mapping functionf and g is required such that :

u = f (x, y), v = g(x, y) (1)

If the functions are known then we can locate a point on theimage knowing it’s position on the map and this process is in-vertible also. What we use for this is a mapping polynomialdefined by:

u = a0 + a1x + a2y + a3xy + a4x2 + a5y2 (2)v = b0 + b1x + b2y + b3xy + b4x2 + b5y2 (3)

Often the value of the coefficients is unknown andis assigned a value based on the knowledge of

2 VIBHUTI BHUSHAN JHA

Ground control Points(G.C.Ps). From the image it iseasy to identify the GCPs as they are some time invariantfeatures, such as road intersections, airport runway, bendsin rivers etc. For hydrological applications it is sometimeseasier to identify meandering features and bends in therivers, as will discussed in the tutorial. From the equations,1.2, 3 it is trivial that we should identify atleast 6 groundcontrol points for the mapping to be unique for a secondorder polynomial registration.The coefficients are evaluatedusing the least square estimation method. The three mostused methods for resampling and interpolation of the pixelvalues are discussed below:

• Nearest Neighbour : This method simply choses theactual pixel that has its centre nearest to the point lo-cated in the image, which is then transferred to the dis-play grid location. This is advantageous in the sensethat the originality of the base images’ brightness isretained in terms of pixel values in the new image.

• Bilinear Interpolation :Let B be the pixel brightnessand (i, j), (i, j+1), (i+1, j), (i+1, j+1) be the locationsof the 4 pixels. Let j′ be the horizontal offset of themap point from the (i + 1, j)th pixel. Then the bilinearinterpolation technique uses three linear interpolationsover the four pixels surrounding the point correspond-ing to the given grid.

• Cubic Convolution: CC method uses the 16 surround-ing pixels along the 4 lines of the 4 pixels. The ac-tual form of the cubic polynomial employs techniquesfrom sampling theory and is far from the scope ofthe lecture. Since the interpolation uses convolutiontechniques, hence the name . The advantage with thismethod is that the image is generally smooth and usedfor photointerpretation, but is not used when our aimis to classify the data as the brightness values may befar off from the radiance obtained from the satellite.

Choice of Control points and georeferencing

As we have seen in the previous sections, it is necessaryto define enough control points so that the accurate mappingpolynomial is generated. As to where the points have tobe, it is advised that the points be distributed well along theedges of the image so that the mapping polynomials are wellbehaved over the image. A key point to note is that higherorder polynomials are effective in capturing the image nearthe control points but they differ markedly outside the GCPrange. Thus it is advised to use nearest neighbour or bilinearinterpolation techniques.

It is often the case that a scene is acquired over different datesand has to be processed simultaneously. As will be discussedin the tutorial we will georegister the images and it is usefulin preforming change analysis. One image can be kept asthe master image which is the base for georegistering andthe other as the slave image. Image to image registering hasthe advantage of saving time in registering to the base mapfor both the images. The algorithm used in the accurate colocation of points is called Sequential similarity detection.For hydrological applications, we take points which arejust near the land heads to minimise the error due to waterboundary shifting due to rainfall etc.

Image interpretation techniques

In this section we will look into different operationsrequired for image processing, we will look into thehistogram equalisation techniques, image sharpening, image

LHD 3

classification etc to name a few.Pattern recognition comes toaid when we want to label and identify the pixels becausethe image comes in the form of spectral classes distributedover the image, our ultimate aim to extract the labels and doclassification, for this we have to first do some operations onthe image so that the classes can be visualised. After this weperform different classification techniques. The pixel vectorscontain the sets of brightness values for the pixels arrangedin the column form as:

x1x2...xn

Where x1, x2......xn are the pixel brightness value in thebands 1 to N respectively.

Histogram

Consider an image, the plot between the number of pixelsvs the brightness value gives us a 2D plot known as the his-togram. An image which spans the brightness values, the his-togram is equally distributed over the region. In this section,we will see some histogram modification techniques whichcan be used to change the contrast, brightness etc.

Contrast Modification

Suppose an image has a poor contrast and it is needed toenhance the image, then one of the easiest ways to do thisis to change the histogram using some mathematical oper-ations. For example an image may occupy the histogramrange between 20 − 60, but for clarity of the image , wewould like to increase the dynamic range to span the entiregray scale 0−255.While the number of bars in the histogramis not altered in this process, the location of the bars is re-specified favourably. The contrast modification process canbe specified as:

y = f (x) (4)

where x is the old brightness of a particular bar in the his-togram and y is the new value and f is the contrast modifica-tion function. The most common contrast modification oper-ation is the one in which the old and new values are relatedby a linear operation, which can be expressed in the form:

y = ax + b (5)

Sometimes a particular region in the image may occupy arestricted limit of values thus saturating linear contrast en-hancement increases the dynamic range in that region to userdefined Bmax and Bmin. Usually the remotely sensed data islow in brightness and poor in contrast, thus there is a need for

automatic contrast stretching in which the mean and stan-dard deviation of the brightness and adjusting the dynamicrange to µ ± 3σ range. Exponential and logarithmic contrastmodification is used for enhancing light and dark featuresrespectively. Another contrast modification method used isthe piecewise modification in which there are break points,where the function changes, it is upto the user to define thepoint and the number of points.

Histogram Equalization

While the previous sections have taken care of basic op-erations on the histograms, sometimes it is desirable that wematch or modify the histogram according to some base im-age. The number of pixels in the range y to y + δy in themodified histogram to match the base image in the range x tox + δx and hi(x) and ho(y) are density functions, which gives,

hi(x)δx = ho(y)δy (6)

Thus if we want to know the shape of the revised histogram,we get after performing some trivial inverse and differentialoperators,

ho(y) = hi( f −1(y))d f −1(y)

dy(7)

But, what will happen if the histogram is a discrete one in-stead of a continuous function. A way in is to find the cumu-lative histogram C(x) which is the sum of the preceding bars.The modified brightness is given by the relation:

y = λC(x) (8)

where λ is a constant. The range of values of y is from 0 toL − 1 for L values of brightness. Thus the constant’s value isL−1N . Following image shows the process of histogram equal-

isation according to a base image.

Image domain analysis

In the previous sections we have seen the pixel wise oper-ations whereas in the present section we will directly work in

4 VIBHUTI BHUSHAN JHA

the image domain as whole, i.e, we will look into neighbour-hood operations. Sometimes it is necessary to assign valuesto a pixel based on the statistical properties of the neighbour-hood. Image analysis often requires methods like smoothing,edge detection and enhancement. Most of these methods re-quire defining a window or a box over a region which canthen be used to cover the entire image. This generates a ra-diometrically modified image. The response for any imagepixel is given by:

r(i, j) = ΣMm=1ΣN

n=1φ(m, n)t(m, n) (9)

Where φ(m, n) is the pixel brightness value and t(m, n) is thetemplate entry at that location. This can be viewed as a con-volution operator.

Mean value image smoothing

Satellite data many a times contain random noise super-imposed on the pixel brightness owing to noise generated bythe sensors. This shows up as salt and pepper. It can beremoved by the process of filtering and especially low passfilter. Mean value smoothing uses the simple average of pixelbrightness values inside the template, the template taken is:

t(m, n) =1

MN(10)

thus applying equatiion, 1.9,we get the response for the im-age.

r(i, j) =1

MNΣM

m=1ΣNn=1φ(m, n) (11)

Thus the pixel at the centre of the template is thus repre-sented by the average brightness level in a neighbourhooddefined by the template dimensions. The problem with thisapproach is that the information about edges which are highfrequency information is lost in this averaging method. Thiscan be taken care of by defining a threshold T such that:

ρi, j =1

MNΣM

m=1ΣNn=1φ(m, n) (12)

r(i, j) = ρ(i, j) : |φ(i, j) − ρ(i, j)| < T (13)= φ(i, j), otherwise (14)

Median filtering

Median filtering is a way to avoid the problems in meanmethod for the edges. The center pixel is assigned a medianvalue of the all the pixels covered in the template. It is fre-quently used in removing impulse related noises as they havean abrupt change for some pixels and their behaviour is wellcaptured if we look for the median values in the neighbour-hood,rather than the mean. Now since the median filtering isnot a linear function, it cannot be described by the convolu-tion operation.

Edge detection

One of the most important tasks in hydrological applica-tions is to find the edges or boundaries. Edge detection high-lights the edges by some gradient or masking methods. Inthis section, we will discuss the edge detection method.

Spatial derivatives

The vector gradient for an image can be defined for thebrightness function as:

∇φ(x, y) =∂

∂xφ(x, y)i +

∂

∂yφ(x, y)j (15)

In making DEMs, we are interested in the magnitude of thegradient,

|∇| =

√∇2

1 + ∇22 (16)

But we need to discretize the gradient operators so that wecan apply template wise processing.

∇1 = φ(i, j) − φ(i + 1, j + 1) (17)∇2 = φ(i + 1, j) − φ(i, j + 1) (18)

An advantage of defining the gradient operators in this wayis that we can detect horizontal, vertical as well as diago-nal edges. This operator is called the Roberts Operator asthe derivatives are defined for the point (i + 1

2 , j + 12 ) in the

diagonal derivatives. Other edge detection operator is theSobel operator, which is computationally more time con-suming because it calculated both for the horizontal as wellas vertical pixels, thus it is like computing a forward differ-ence scheme.

∇1 = ∇1a − ∇1b (19)

and∇2 = ∇2a − ∇2b (20)

where ∇1a = φ(i−1, j+1)+2φ(i−1, j)+φ(i−1, j−1),∇1b =

φ(i + 1, j + 1) + 2φ(i + 1, j) + φ(i + 1, j − 1) and similarlyfor ∇2a,b. Sobel operator can be thought of as the imple-mentation of following template in the image. Sometimes it

becomes necessary to extract information in the fourier do-main, i.e, the frequency domain. This is done by means offourier transformation. For example, a periodic noise can beremoved by working in the fourier domain rather the spatialdomain as it captures the periodicity pretty well. Familiarity

LHD 5

with basic complex analysis and integral calculus is assumedin this section. The fourier transformation can be written as:

F(ω) =

∫ ∞−∞

f (t)e−iωtdt (21)

The impulse response relationship is given a convolution op-eration as:

y(t) =

∫ ∞−∞

x(τ)h(t − τ)dτ ∼ x(t) ∗ h(t) (22)

But to extend it to an image requires the use of double inte-grals as the image is two variable function.

Supervised Classification

The analysis of the image requires the extraction of sim-ilar features in the image, for example, a class of vegeta-tion, water bodies, forests etc. For this we need the conceptof classification, supervised and unsupervised in which weclassify the clusters according to some mathematical modelsespecially probabilistic models. Supervised classification in-volves probabilistic models in which the user labels the classand the total number of classes has to be determined. Usuallythe steps involved are:

• Decide the number of classes to be classified, an infor-mation about the segments of the image for example,water, urban regions, croplands.

• Training data is formed in which representative pixelsof a particular class are chosen.

• Estimate the probability models from the training dataand these equations will decide the signature of theclass.

• Finding the accuracy of the final product.

Bayes’ classification rule

The spectral classes of the image are ωi, i = 1, 2, ....,M,where M is the number of classes. We need to find the con-ditional probability of a pixel vector x, which is given by:

p(ωi|x), i = 1, 2, ...,M (23)

Now the pixel vector signifies the brightness values. Thisgives us the likelihood that the correct class is ωi correspond-ing to the pixel brightness vector x. The schema is:

p(ωi|x) > p(ω j|x)∀x ∈ ωi (24)

This figure illustrates the class assignment:

Maximum likelihood

The conditional probability distribution is given by a gaus-sian distribution in multivariate form:

p(x|ωi) = (2π)−N2 |Σ|−

12 e−

12 (x-mi)tΣ−1

i (x-mi)(25)

Where mi and Σi are class mean and covariance matrix re-spectively. The discriminant function is given by gi(x) =

ln(p(x|ωi)p(wi)). Thus the condition for belonging to a classis x ∈ ωi if gi(x) > g j(x)∀ j , i. Now the implementationof the maximum likelihood is to calculate the discriminantfunction and then label the classes based on it,

gi(x) = −ln|Σi| − (x-mi)tΣ−1i (x-mi)(26)

Unsupervised classification

In general it is not possible to discern the number of spec-tral classes present in an image. For this clustering can beused for unsupervised classification. In unsupervised clas-sification, pixels in an image are assigned to spectral classeswithout the user having beforehand knowledge of the classes.In this section we will discuss many clustering techniques.

Clustering criteria

Grouping of the pixels in a multispectral space is calledclustering. The idea is to generate some similarity mea-sure for clustering, this is done by making distance measures

6 VIBHUTI BHUSHAN JHA

which in a way signify the distance between the pixels. It isa L1 measure and is given by:

d(x1, x2) = ΣNi=1|x1i − x2i | (27)

where x1i,2i are the ith component of the pixel vectors respec-tively. It is thus easy to form clusters based on this distancemeasure. After this we find the sum of sqaured errors (SSE)defined as:

S S E = ΣCiΣx∈Ci ||x −mi||2(28)

Migrating Means clustering

In this method, following steps are followed:

• N points are selected in multispectral space as a samplecluster centres.

• The location of each pixel is assigned to the nearestcluster based on a distance measure.

• The new means are computed, if this matches the ini-tial taken mean ,then the process terminates otherwiseis repeated again.

The most common algorithm for supervised classification ismaximum likelihood. The limitation of this model is that theclasses must be represented by a multivariate normal distri-bution. But most of the times the data is multimodal and thusclustering algorithms are needed for classification.

References

• Richards Jia, Remote sensing digital image analysis.

• Gonzalaez and Woods, Digital Image Processing.

• The images have been taken from Richards Jia, Re-mote Sensing digital image analysis.