MONITORING AND ASSESSMENT OF LAGUNA DE BAY WATERS FROM SPACE OBSERVATIONS AND OPTICAL MODELING

Enrico C. Paringit

JSPS Research Fellow Tokyo Institute of Technology

Assistant Professor, Department of Geodetic Engineering University of the Philippines Diliman

Kazuo Nadaoka

Professor, Department of Mechanical and Environmental Informatics Tokyo Institute of Technology

Abstract: Laguna de Bay is one of the largest inland freshwater bodies in Asia and is currently host to a number of industrial, aquaculture and other livelihood activities. Protection of the lake from environmental degradation requires constant and reliable monitoring at synoptic scales. Time-series satellite imagery from ASTER AM-1 sensor, were used to monitor the Laguna de Bay water quality condition. A bio-optical model was first developed specific for the waters of Laguna de Bay, which is shallow and often turbid. Special studies were carried out to estimate absorption and scattering properties as well as backscattering probability of suspended matter. The inversion of bio-optical model allows for direct retrieval of turbidity and chlorophyll-a from the visible-near infrared (VNIR) range sensor. Spatial distribution of temperature for the lake was extracted from the thermal infrared (TIR) sensor. Corresponding field surveys were conducted to parameterize the bio-optical model. In-situ measurements include suspended particle and chlorophyll-a concentrations profiles from nephelometric devices and water sampling and processing. Hyperspectral measurements were used to validate results of the bio-optical model and satellite-based estimation. This study provides a theoretical basis and a practical illustration in carrying out space-based measurements that can be performed on an operational basis. Key Words: Laguna de Bay, water quality, remote sensing, bio-optical model 1. INTRODUCTION With over 900 km2 surface area, Laguna de Bay ranks notably as one of the largest freshwater lakes in Asia. In the face of massive environmental stresses imposed by adjacent land area and within the lake itself, the sheer size of the Bay requires constant and dedicated monitoring system for the sustainable management of its resources (Nauta et. al., 2003, Delos Reyes and Martens 1994). Natural conditions (topography, presence of tributaries, wind forcing, circulation etc.) coupled with different activities (aquaculture, power generation, irrigation etc.) in and around the lake contribute to the significant variability, both at temporal and spatial scales, in the lake water quality conditions. While previous studies established distinctive seasonal variations in water quality (Barril et al., 2001; Santiago, 1991), critical data to deduct further information on the state of shallow lake, which is often turbid, are still piecemeal and intermittent since the primary mode of data collection are still point-specific, costly and labor-intensive. Although remote sensing techniques possess considerable potential for limnological monitoring on an operational basis due to synoptic data acquisition mode and periodic onsite revisit capability, its application to assess water quality conditions particularly for Laguna de Bay remains

scarce or untapped. For instance, only the work by Mancebo et al. (1997) to determine the correlation between the total suspended solids (TSS) in-situ and the reflectance in the Landsat TM of Laguna de Bay, exist. Nonetheless, methods such as these are based mainly on statistical analysis of remotely sensed and field data, leading to algorithms with limited applicability and accuracy. Our objective therefore in this study is to demonstrate the capability of time-series satellite image analysis, particularly that from ASTER data to generate seasonal patterns of spatially distributed water quality parameters through the use of physically-based bio-optical model. 2. METHODS AND MATERIALS First, we present the background theory and model to extract water quality parameters from spectral data. Next, the model is applied to a specific satellite data acquired at different time periods. Spatial patterns, temporal trends and other features peculiar to Laguna de Bay are analyzed according to the resultant calculations. 2.1 Bio-optical Modeling for Turbid Waters and Inversion Technique Bio-optical modeling in natural waters exploits the fact that various optically active constituents such as suspended sediments and chlorophyll-a respond differently to visible portions of the electromagnetic spectrum. The basic remote sensing equation relating the inherent optical properties (IOPs) of natural waters with measured reflectance is (Gordon et al., 1975)

( )0bw S S

w S S CHL CHL CDOM bw S S

b b BCR k

a a C a C a b b Cλ

−

+=+ + + + +

(1)

where ( )0R λ − : irradiance reflectance just below the water surface (%) at wavelength λ

k : a constant factor (unitless) 0.544(0.629cos 0.975)sθ= +

wa : absorption coefficient for pure water ( )1m−

Sa : absorption coefficient for suspended sediments ( )1m−

CDOMa : absorption coefficient for colored dissolved organic matter (CDOM; unitless)

wb : scattering coefficient for pure water ( )1m−

Sb : backscatter coefficient for suspended particles ( )1m−

B : backscattering ratio (unitless) SC : concentration of suspended sediments (mg/L)

CHLC : concentration of chlorophyll-a (mg/L) while the concentration of total suspended matter for the same volume is T S CHLC C C= + . Equation (1) holds when the vertical distribution of water constituents is uniform over the water column and that optical thickness through depth z on a portion of the lake is large

enough making bottom influence on reflectance negligible. Use of an auxiliary variable, ( )0 1

1x k R λ −= − renders Equation (1) as

( ) ( )w CDOM bw CHL S S CHL S Sa a b x a a b x C b BC− + + = + − + (2)

For m number of wavelength bands, Equation (2) can be rewritten in matrix form as

= +z AC n (3)

where z : ( )1m× containing ( )0.5 w CDOM bwa a b x− + + for all m wavelength ranges available

A : 2m× , ( )CHL S S

S

a a b x

b

+ − =

A

C : ( )2 1× the water constituents[ ]TCHL SC BC to be estimated

n : variations in z not explained by the model

The ordinary least squares solution to simultaneously estimate the water constituents, C is

( ) 1ˆ T T−=C A A A z . By setting the partial derivatives ( ) 0T∂ ∂ =n n C and 0F λ∂ ∂ = , we

can invert Equation (3)

( ) 1ˆ Tα α −= + −C Uj U UJU A N z � � � � � � � � � � (4)

where U : ( )m m× -1 -1TU = A N A

N : ( )m m× the dispersion or covariance matrix of the residuals

j : ( ) [ ]2 1 1,1T× =

2.2 Surface Temperature Measurements from Satellite Radiance Data The radiation emitted in the thermal infrared part of the spectrum can be related to surface temperature so that an image of lake surface temperature may be produced. A physics-based approach can be used which couples a surface temperature and emissivity model with a radiative transfer model. The spectral radiance of a blackbody at temperature T and wavelength λ , is given by the Planck function.

( ) ( )1

5 1 12exp 1BB

cL

c Tλ

λ π λ− −=

− (5)

where ( )BBL λ : blackbody radiance ( )-2 -1 1Wm sr mµ −

T : temperature ( )K�

of the target as it were a blackbody

1c : first radiation constant ( )-3 13.74151x10 -22 Wm mµ −

2c : second radiation constant ( )0.0143879 m °K Equation (5) is integrated over the instrument response function ( )S λ to calculate the radiance that corresponds to a brightness temperature for a particular instrument channel or band i within the[ ]1 2,x x range.

( )( ) ( )

( )

2

1

2

1

BB

S

BB

S L b dL b

S d

λ

λλ

λ

λ λ

λ λ=

∫∫

(6)

where ( )SL b is the radiance measured by the sensor while ( )BB

L b is the radiance at 370°K

at band number b . For ( )S λ of finite (not infinitesimal) width, Equation (6) cannot be inverted explicitly. Instead, we employ an iterative method to obtain brightness temperature T and emissivity ε . 2.3 ASTER Data The ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) is an imaging instrument onboard Terra, an earth observation satellite launched in 1999 through a cooperative effort between the United States National Aeronautic and Space Administration (NASA) and Japan's Ministry of Economy, Trade and Industry (METI) Earth Remote Sensing Data Analysis Center (ERSDAC). ASTER is composed of three instruments to detect 14 wavelength ranges in the visible infrared (VNIR), shortwave infrared (SWIR) and thermal infrared (TIR) spectrum (See Table 1 for summary of instrument characteristics). As it is, images over water-laden areas produced by ASTER will be sensitive to shorter wavelengths but the extended wavelength ranges are useful for determining the thermal properties of surface waters.

Table 1. Characteristics of ASTER Instruments

Characteristic VNIR SWIR TIRSpectral Range Band 1: 0.52 – 0.60 µm Band 4: 1.600 - 1.700 µm Band 10: 8.125 - 8.475 µm

Band 2: 0.63 - 0.69 µm Band 5: 2.145 - 2.185 µm Band 11: 8.475 - 8.825 µmBand 3: 0.76 - 0.86 µm Band 6: 2.185 - 2.225 µm Band 12: 8.925 - 9.275 µm

Band 3*: 0.76 - 0.86 µm Band 7: 2.235 - 2.285 µm Band 13: 10.25 - 10.95 µmBand 8: 2.295 - 2.365 µm Band 14: 10.95 - 11.65 µmBand 9: 2.360 - 2.430 µm

Ground Resolution 15 m 30m 90m* Backward looking

The application of optical imagery on surface with high absorption require precise spectral calibration. Radiance calculation from L1BDN based on the results for the trend analysis of the on-board calibrator and/or the vicarious calibration

( ) ( ) ( ) ( )L1B

L1B

, , , 1 ,

( , )trend

C b m DN b m v K b vL b

K b d

− = (7)

where ( ),C b m : Unit conversion coefficients in band b and gain mode m R(b, v):� Optical calibration coefficients for sensor degradation in band b and

calibration version v. ( , )trendK b d : Degradation function in band b and the days since launch estimated from the

sensor trend analysis due sensor degradation in band b and calibration version v.

For this study, ASTER datasets acquired at seven different dates in a period of two years (2001/01/03; 2001/04/25; 2001/11/03; 2002/03/27; 2002/09/03; 2002/12/08; 2003/01/16) have been compiled. All images were acquired approximately 10:35 AM at –8.59° viewing angle.

3. FIELD DATA AND PROCESSING TECHNIQUES Field surveys were conducted on October 2001, February 2002, and March 2003 covering the west, central and south bay portions of Laguna de Bay. Chlorophyll-a, temperature, turbidity and salinity values were obtained using three modes of measurement. The first is by deployment of data-logging sensors, designed to capture the temporal variability of the water quality parameters at diurnal to monthly time scales at a fixed point in the lake. The second method is done through lowering an STD (Alec Instruments Inc.) instrument at desired locations to obtain profile of the water quality, and third, by direct water sampling, particularly for precise determination of chlorophyll-a and suspended particle concentrations through laboratory processing. To characterize the optical properties of Laguna Lake waters, particularly absorption and scattering, an Ocean Optics S2000 underwater spectroradiometer was used to acquire spectral vertical profiles in the 350-1015 nm range. The spectrometer was also towed along a line transect in the west bay to detect spatial variability of in-water reflectance. The bio-optical modeling is implemented in two parts. The first part computes for the inherent optical properties (IOPs) of water ( , , ,S CHL Sa a b B and )bwb at known

CHLC and SC based on in-situ spectral data known as the forward model. The second part implements the inversion scheme, which in turn, utilizes the IOPs obtained from the forward model to estimate water quality from the satellite data. For the first part,

( )0R λ − computations considers the chlorophyll and turbidity profiles instead of assuming a single value or depth-averaged for turbidity and chlorophyll-a values.

4. RESULTS AND DISCUSSION Figure 1 shows the vertical profile of water quality parameters. It can be seen that although there is significant variability through depth, values do not exceed 1 order of magnitude, hence an average value at each depth segment is used for the forward model. Note that the profile for water temperature is not applicable for the retrieval algorithm since TIR sensors only detect surface kinetic temperature with effective penetration down to 30 cm or less depending on the wind condition.

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

5.0 10.0 15.0Chlorophyll-a (µ g/L)

Dep

th (

m)

L1 L2

L3 L4

90 100 110 120

Turbidity (mg/L)

L1 L2

L3 L4

24 26 28 30 32

T emperature ( � C)

L1 L2L3 L4

14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'14°•15'

14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'14°30'

��� � � �� � � ��

� ����

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"��������������� �!�#"�$���%�&���'�(� �!�#"�$���%�&���'�(� �!�#"�$���%�&���'�(� �!�#"�$���%�&���'�(� �!�#"�$���%�&���'�(� �!�#"�$���%�&���'�(� �!�#"�$���%�&���'�(� �!�#"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%�&���'�(� �*�+"���)�%���������,�*�+"���)�%���������,�*�+"���)�%���������,�*�+"���)�%���������,�*�+"���)�%���������,�*�+"���)�%���������,�*�+"���)�%���������,�*�+"���������������,�*�#"���������������,�*�#"���������������,�*�#"���������������,�*�#"���������������,�*�#"���������������,�*�#"���������������,�*�#"���)�%�&���'�(� �*�+"

-. / 0 1324 5

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15'

121°

•15' 6 6�77

8:9 ;=< > ?@A ? BC CEDD

FHGI J K G L

Figure 1. Vertical Profile of a) chlorophyll; b) turbidity concentrations and c) temperature in the west bay portion of Laguna de Bay observed March 2003 by STD measurements

Figure 2 shows the time series images of Laguna de Bay as depicted by visible band 1 sensor (0.52-0.60 µm). It can be immediately noticed from the varying intensity of water reflectance that significant changes in water composition take place through time. The shorter wavelength range of the VNIR sensor detects the scattering of the suspended matter present in the lake waters, thus giving rise the appearance of brighter reflectance returns from the water body in contrast to the darker land area.

(a) 2001/01/03 (b) 2001/04/25 (c) 2001/11/03 (d) 2002/03/27

(e) 2002/09/03 (f) 2002/12/08 (g) 2003/01/16

Figure 2. ASTER AMS-1 Band 1 Image of Laguna de Bay at Different Acquisition Dates Calibrated and Co-registered

4.1 Results of Water Quality Parameter Inversion Temporal trends deducted from Figure 4a and b indicate that it is much more turbid during the dry season (November to May) that it is during the wet season (May to October). Based on the computations, the east bay turbidity levels remained almost constant all throughout the year. Productivity is seen to be much lower in the west bay than in the central bay while chlorophyll-a values remain consistently lower in narrow strait between the south and central bay portions relative to other areas all throughout the period of study. Strong presence of chlorophyll-a is noticeable during the rainy season. This closely corresponds to growth of aquatic biota (phytoplanktons and zooplankton) with the onset and towards the end of the rainy season. This is evident from local observation of shore areas especially around Talim island where higher productivity is attributed to lower circulation activity in the area restricted by the presence of a land mass. Although the bay is shallow, influence of bottom cover reflectance is considered to be practically negligible.

2001/01/03

2001/04/25

2001/11/03

2002/03/27

2002/09/03

Figure 3. Results of the Chlorophyll-a, Suspended Particle and Temperature

Estimations Based on ASTER Data

2002/12/08

2003/01/16

0 5 10 >15 0 28 29 29.5 30 3>120

0 28 29 29.5 30 31 >32

(a) chlorophyll-a (µg/L) (b) suspended particle(mg/L) temperature (°C) Figure 4 cont’d. Results of the Chlorophyll-a, Suspended Particle and Temperature

Estimations Based on ASTER Data

4.2 Results for Lake Surface Temperature Estimation The resulting lake surface temperature estimates are shown in Figure 4c. Maximum temperature is observed in the height of the dry season but seem to be uniform throughout the bay. Considering the limits of standard deviation, the resulting temperature estimates from the satellite data is comparable to the field data (See Figure 5).

MONP$QR$STOUTWVTWX

Y Z [ \ ] ^ ^ \ _ ` a b c d e f e c c f g h i b c d ej,k�lnmpoq rsrtuvw xury z r{|} ~����p� ���E�:�3���:�3��� �����E�����H�)�H�3�E���!�

� �3���p�3�=�3���:�������n�����p�3�=�����3�3�)���

Figure 5. Lake Surface Temperature at the West Bay of the Lake

4.3 Discussion Research gaps include size distribution for suspended particles matter and determination of absorption due to CDOM. The particle size of the suspended sediment and chlorophyll-a are likewise key determining factors for computing backscattering coefficients (Risovic, 2002). Particle sizes were deducted from bottom sediment samples (Suzuki et al., 2003) but do not necessarily represent suspended particle dimensions. CDOM may likewise dominate in the lake waters since most suspended materials consist of diatoms (Chrysophyceae C.) and organic matter. Results of the chlorophyll-a estimation can be used for providing input or validation data to number of physical nutrient and plankton growth models (Baldia et al., 2003; Bocci, 1999) and for correlation with tributary runoffs and suspended solid loads (NHRC, 1999; Zimmer and Bendoricchio, 2001). Another benefit of satellite-based processing is that assessment

of the surrounding watershed regions can be carried out with the same dataset covering these areas. This feature allows synchronized monitoring of the Laguna de Bay ecological system as a whole at a cost-effective manner because the same set of hardware and software resources can be utilized. Aside from monitoring of water quality, high spatial resolution of the 15-m ASTER VNIR sensor also allows for discernment of minute features in the lake waters and shores. For example, fishponds and fish cages (average area: 50 has per cell) can be identified by their distinct linear-block shapes. This would have immediate application to enforcement as lake waters are being leased for aquaculture purposes. Lake shoreline development due to depth levels could be distinguished through the season. This information would guide shore property delineation and proper use of accrued agricultural plots. Build-up of fast growing water hyacinths (Eichornia crassipes) are likewise discernable in the images and can serve as basis for their control and disposal. Environmentally, water hyacinth is considered an environmental and health problem since depleted oxygen conditions develop beneath water hyacinth mats and the dense floating mats impede water flow and create conditions conducive for mosquitoes. 5. CONCLUSIONS A method to extract various water quality parameters for Laguna de Bay waters based on ASTER AM-1 data analysis is presented. The magnitudes of the estimates are well within the field measurement values. The technique provides a feasible alternative but richly informative tool for assessment and monitoring water quality conditions

ACKNOWLEDGMENTS The ASTER data used in the research was provided by the ERSDAC ARO Program (ARO 94). The authors are likewise grateful to Prof Fernando Siringan of the University of the Philippines National Institute of Geological Sciences (UP-NIGS) for sharing the Laguna Lake shoreline data.

REFERENCES Baldia S.F., Conaco M.C.G., Nishijima T., Imanishi S., and Harada K.I. (2003) Microcystin production during algal bloom occurrence in Laguna de Bay, the Philippines, Fisheries Science, Vol. 69, No. 1, 110-116. Barril C.R., Tumlos E.T., and Moraga W.C. (2001) Seasonal variations in water quality of Laguna de Bay, Philippines: Trends and implications, Philippine Agricultural Science, Vol. 84, No. 1, 19-25. Bocci, M. (1999) Modelling the growth of Nile Tilapia (Oreochromis niloticus) feeding on natural resources in enclosures in Laguna de Bay (Philippines), Ecological Modelling, Vol. 119, No. 2-3, 135-148.

Delos Reyes, M. and R. Martens (1993) Geoecology of Laguna de Bay, Philippines I. Techno-commercial impact on the trophic level structure of the Laguna de Bay aquatic ecosystem 1968-1980, Ecological Modelling, Vol. 75-76, 497 509. Gordon, H.R., Brown, O.B., and Jacobs, M. M. (1975) Computed relations between the inherent and apparent optical properties of flat homogenous ocean, Applied Optics, Vol. 14, 417-427. Mancebo, F.F., Bruce, R.C., Catalan, Z.B., Lim suan, Medel P., and Malayang, B.S., III (1997) Correlation of Total Suspended Sediments and Reflectance of Landsat TM in Laguna De Bay, Philippines, Proceedings of the 18th Asian Conference on Remote Sensing, Kuala Lumpur, Malaysia. Accessed online: http://www.gisdevelopment.net/aars/acrs/1997/ts2/ts2003.shtml (17 June 2003). Nauta, T. A., Bongco A. E. and Santos-Borja, A.C. (2003) Set-up of a decision support system to support sustainable development of the Laguna de Bay, Philippines, Marine Pollution Bulletin, Vol. 47, No. 1-6, 211-218. National Hydraulic Research Center, (1999) Development of Laguna de Bay Water Management Simulation Model, Report No. 117, National Hydraulic Research Center, University of the Philippines Diliman. Risovic, D. (2002) Effect of Suspended Particulate-Size Distribution on the Backscattering Ratio in the Remote Sensing of Seawater, Applied Optics-LP, Vol. 41 Issue 33, 7092. Santiago, A.E. (1991). Turbidity and seawater intrusion in Laguna de Bay, Environmental Monitoring and Assessment, Vol. 16, 85-95. Suzuki, Y., Urase, T., Nadaoka, K., Tamura, H., and Siringan, F., The Distribution of Metal Concentrations in Sediments in Manila Bay, Pasig River and Laguna Lake, Philippines and Lead as Indicators of Pollution by Runoff, Proceedings of the 4th Regional Symposium on Infrastructure Development in Civil Engineering, Bangkok, Thailand, (in CD-ROM). Zimmer, V. and Bendoricchio (2001) Nutrient and suspended solid loads in the Laguna de Bay, Philippines, Nutrient and suspended solid loads in the Laguna de Bay, Philippines, Water Science and Technology, Vol. 44, No. 7, 77-86.