1

Directional Difference of Satellite Directional Difference of Satellite Land Surface TemperatureLand Surface Temperature

Yunyue YuYunyue Yu

NOAA/NESDIS/STARNOAA/NESDIS/STAR

2

OutlineOutline

Research MotivationResearch Motivation Concept of the Directional EffectConcept of the Directional Effect Directional Effect Observed in Polar-orbiting Directional Effect Observed in Polar-orbiting

Satellite DataSatellite Data Directional Effect Observed in Geostationary Directional Effect Observed in Geostationary

Satellite DataSatellite Data Possible Solutions?Possible Solutions?

3

Research MotivationResearch Motivation

Is Directional LST Difference Significant ?Is Directional LST Difference Significant ? Affect to LST validation processAffect to LST validation process Affect to produce climate data recordAffect to produce climate data record Affect to LST applications, e.g. data assimilation Affect to LST applications, e.g. data assimilation

for forecast modelfor forecast model

4

Concept of the Concept of the Directional EffectDirectional Effect

One of directional effect described by Modified Geometric Projection Model

4/1

1

4 ),(),(

1),(

N

kkkk XTT

N

kkk X

1

),(),(

k=N endmembers X: fraction of the cover probability : emissivity of the endmember T: temperature of the endmember

Four endmembers: Sunlit Crown, Shaded Crown Sunlit background, Shaded Background

Sunlit Crown

Shaded Crown

Shaded BackgroundSunlit Background

5

Concept of the Directional Concept of the Directional Effect Effect MGP Model Run ExampleMGP Model Run Example

Examples of the surface temperature distributions and the mean emissivity distributions (solid line) along with the satellite view zenith angle. The temperature and emissivity distributions are calculated from the MGP model temperature settings, for solar zenith angle at 0, 30 and 60 degrees, respectively; the LAI value is 1.

The vegetation coverage is 60%.

The vegetation coverage is 30%.

Apparent Emissivity Apparent LST

6

Directional Effect Directional Effect Observed in AVHRR DataObserved in AVHRR Data

Ghanzi (21.7 S; 21.6 E)

300

305

310

315

320

325

330

212 216 220 224 228 232 236 240 244 248 252 256

Julian Day (1995)

LS

T (

K)

Daily daytime AVHRR a) Land surface temperature and b) View zenith angle. [Pinheiro et al., Remote Sensing of Environment, 103 (2006)]

Ghanzi (21.7 S; 21.6 E)

-70

-50

-30

-10

10

30

50

70

212 216 220 224 228 232 236 240 244 248 252 256

Julian Day (1995)

Vie

w z

en

ith

an

gle

Daytime AVHRR, LST observations, at Ghanzi, Botswana

Demonstrated LST variability is a combination of:a) residual atmospheric effects b) real aggregate temperature differences c) emissivity angular variability

7

Directional Effect Observed in Directional Effect Observed in Geostationary Satellite DataGeostationary Satellite Data

No.No. Site LocationSite Location Lat/LonLat/Lon Surface Type*Surface Type*

11 Pennsylvania State University, PAPennsylvania State University, PA 40.72/77.9340.72/77.93 Mixed ForestMixed Forest

22 Bondeville, ILBondeville, IL 40.05/88.3740.05/88.37 Crop LandCrop Land

33 Goodwin Creek, MSGoodwin Creek, MS 34.25/89.8734.25/89.87Evergreen Evergreen

Needle Leaf Needle Leaf ForestForest

44 Fort Peck, MTFort Peck, MT 48.31/105.1048.31/105.10 Grass LandGrass Land

55 Boulder, COBoulder, CO 40.13/105.2440.13/105.24 Crop LandCrop Land

66 Desert Rock, NVDesert Rock, NV 36.63/116.0236.63/116.02 Open Shrub Open Shrub LandLand

Down-looking PIR at 8 meter height from the ground

UP-looking PIR

Diffuse Radiometer

Down-looking PIR on the towerAt 8-m from ground

Thermometer

Anemometer

Duration of Data: Jan 1 – Dec 31, 2001

GOES-8 and GOES-10 Imager data were aqpplied in validating LST algorithm using ground data from SURFace RADiation (SURFRAD) budget network stations

8

Two-directions from GOES Two-directions from GOES SatellitesSatellites

135° W 75°W

9

Difference of LSTs observed by GOES-10 and GOES-8 imager at the same location of SURFRAD station Desert Rock, NV, 36.63ºN, 116.02ºW. The simultaneous observation pairs are about 2096.

View zenith of GOES-8: 60.140

View zenith of GOES-10: 46.810

LST Directional Effect in LST Directional Effect in GOES-8 and -10 ImagerGOES-8 and -10 Imager

10

Goodwin Creek, MS, observation pairs are about 510. View Zenith of GOES-8/-10: 42.680/61.890

LST Directional Effect in LST Directional Effect in GOES-8 and -10 Imager (2)GOES-8 and -10 Imager (2)

11

Boulder, CO, observation pairs are about 510. View Zenith of GOES-8/-10: 42.680/61.890

LST Directional Effect in LST Directional Effect in GOES-8 and -10 Imager (3)GOES-8 and -10 Imager (3)

12

LST Directional Effect in LST Directional Effect in GOES-8 and -10 Imager (4)GOES-8 and -10 Imager (4)

Bondville, IL. Data pairs: 710 Fort Peck, MT. Data pairs: 912

View Zenith of GOES-8: 48.120

View Zenith of GOES-10: 66.140

View Zenith of GOES-8: 62.420

View Zenith of GOES-10: 62.360

Note the difference of the two sites

13

SummarySummary

SummarySummary» LST directional effect were observed from Polar-orbiting LST directional effect were observed from Polar-orbiting

satellite data (NOAA/AVHRR) and Goestationary satellite satellite data (NOAA/AVHRR) and Goestationary satellite data (GOES/Imager)data (GOES/Imager)

» LST difference due to the viewing angle difference LST difference due to the viewing angle difference changes diurnally; the effect during daytime is changes diurnally; the effect during daytime is considerable bigger than that is during nighttime. considerable bigger than that is during nighttime.

» The satellite LST uncertainty due to the directional effect The satellite LST uncertainty due to the directional effect is considerably larger comparing to the requirement, and is considerably larger comparing to the requirement, and cannot be ignored (particularly during the daytime).cannot be ignored (particularly during the daytime).

» VIIRS/LST should provide correction/complimentary VIIRS/LST should provide correction/complimentary information on it after the launch (do work from now). information on it after the launch (do work from now).

14

Possible Solution?Possible Solution?

Sample of common area observed by GOES-E, GOES-W and POES satellites. LSTs derived from those satellite data will be used to develop an unified LST

algorithm.

Different satellite observations over common areas can be calibrated each other for the data consistency.

15

Back-upsBack-ups

16

Fort PeckFort Peck

17

Goodwin CreekGoodwin Creek

18

Desert RockDesert Rock

19

BoulderBoulder

20

Mean differences Δi,j between nighttime observed LST.

Let us consider that nighttime LST observation at five selected SURFRAD stations is unbiased. In such assumption all GOES-8 observed LST should be corrected by adding constant bias ~1.0 ºC, and all GOES-10 observed LST should be corrected by adding constant bias ~1.3 ºC. This table will be recomputed!

WE CAN TRY TO USE NIGHTTIME OBSERVATION TO EVALUATE SYSTEMATIC ERRORS IN LST

SURFRAD Network Systematic Differences, i,j , ºC

GOES -8 - SURFRAD GOES -10 - SURFRAD GOES -10 – GOES -8 Station Name

2,1 3,1 3,2

Goodwin Creek, MS 0.4 -0.3 -0.8

Desert Rock, NV -3.2 -2.3 0.9

Bondville, IL -0.1 -1.4 -1.3

Boulder, CO -1.1 -1.2 -0.2

F ort Peck, MT -0.8 -1.1 -0.3

AVERAGE -1.0 -1.3 -0.3