Вы находитесь на странице: 1из 12

Mohsin Jamil and Majid Nazeer, 2009

Proceedings of the Global Conference on Microbial Contaminants in Drinking Water,


5-8 October, Singapore.

Assessment of Water Quality Parameters of Rawal Lake Using Satellite Remote Sensing

Mohsin Jamil and Majid Nazeer


Climate Snow and Hydrology Research Group (CSHRG)
Department of Meteorology
COMSATS Institute of Information Technology
Islamabad-44000
Email: mohsin_jamil@comsats.edu.pk
________________________________________________________________________
Abstract
Satellite remote sensing provides a synoptic view of the ground features and a better
mean of updating natural resources information. Under this fact, the satellite remote
sensing is very useful for surface water monitoring because there is a strong empirical
relationship between satellite data and regular ground observation data of secchi disc
transparency (SDT). In this study, the usability of satellite imagery for estimating water
quality patterns is evaluated for the water quality of Rawal Lake. Four Landsat Thematic
Mapper (TM) image of the years 1992, 1998, 1999 and 2000 are used. For water quality
assessment images for the months of May to September have been preferred due to stable
water conditions and reliable water clarity. Multiple Regression Model is used for the
prediction of secchi disc transparency from satellite data. The independent variables
TM1/TM3 and TM1 in the regression model are strong predictors of SDT. The time
window between the collection of physical data of secchi disc and satellite data affects
the value of regression coefficient (r2). In this study the value of regression coefficient
(r2) ranges from 0.51-0.87 and the standard error of estimate ranges from 0.041-0.146 and
the Carlson’s trophic state index showed that Rawal Lake is an eutrophic lake. A wider
time window decreases the value of r2. By predicting the values of SDT with the help of
regression model we are able to make a retrospective analysis on the past conditions of
lake water.
________________________________________________________________________

1. Introduction

Satellite remote sensing is a better mean of updating natural resources information,


monitoring land use changes and its environmental impacts near real time and providing
a historical profile of the area under study. Satellite remote sensing has become a
powerful tool for surface water monitoring as well. It has well documented that the
satellite imagery is useful for monitoring water clarity due to the strong relationship
between satellite data and regular ground observation data [1].

1
Mohsin Jamil and Majid Nazeer, 2009

The lake water clarity and the trophic conditions of a lake can be significantly measured
through satellite imagery and many satellite systems can be used for this purpose like
ASTER, IKONOS, Quick bird and SPOT. The objectives of the study are to relate the
ground observation data taken by a 20 cm diameter secchi disc with the radiance data
taken by Landsat Thematic Mapper of the Rawal Lake [1,2,3,4,5] and to find the trophic
state of the Rawal Lake with the help of Carlson’s trophic state indices.

Trophic state is basically the influence of biomass on a water body at a certain time and
site which is classified into a continuum ranging from 0-100, mainly there are three
trophic states, oligotrophic, mesotrophic and eutrophic [6]. The values of the trophic state
index are calculated from secchi disc transparency (meters), chlorophyll concentration
(µg/l-1) and total phosphorus (µg/l-1) [4]. The trophic state index can be calculated by
using these three parameters or any one of these. For a single lake the value of trophic
state will be the same either it is calculated from secchi depth, chlorophyll concentration
or from phosphorus concentration. The TSI associated with secchi depth can be
calculated from the equation

(1)

Similarly the equations for chlorophyll and phosphorus are given below respectively;

(2)

(3)

In equations 1, 2 and 3, secchi depth is in meters while chlorophyll and phosphorus


values are in micrograms per liter (µg/l-1) [6]. These are the standard equations used to
calculate TSI values from secchi depth, chlorophyll-a and total phosphorus
independently. Table 1 shows the different classes of lakes depending on the values of
SDT, chlorophyll-a and phosphorus;

Table1: Lake trophic states and classification ranges for trophic state index, total
phosphorus, Secchi disc transparency and chlorophyll-a. Adopted from USGS fact sheet
2007-3022 [11].
Lake Trophic Carlson SDT Chl-a TP
Condition TSI (ft) (µg/l-1) (µg/l-1)
Oligotrophic <38 >15 <2.2 <10
Mesotrophic 38-48 7.5-15 2.2-6 10-20
Eutrophic 49-61 3-7.4 6.1-22 20.1-50
Hypereutrophic >61 <3 >22 >50

2
Mohsin Jamil and Majid Nazeer, 2009

Remote sensing of water quality monitoring for the open water bodies, mainly involves
the estimation of chlorophyll and the suspended particles from the radiance data received
from the sensor. Chlorophyll-a is primary index for the presence of phytoplankton and is
used to estimate the productivity of the biomass in the water body. On the other hand
suspended sediments like total phosphorus and yellow substance are the common types of
pollutants both in terms of weight and volume in the surface of water system and are used
to estimate the contamination of the water body [7].

There is a strong relationship between the satellite radiometric data and the ground
observation data collected with the help of a secchi disc. Through secchi disc
transparency we can estimate the clarity level of a lake, reservoir or a river. The water
body appearing in shades of green, will have highly polluted water, the water reflecting in
shades of dark blue shows very low concentration of algal mass and the sedimentation of
water reflects in shades of brown [5].

The usability of remote sensing for the lake water clarity can be described by the
following points;

First is the time window influencing the relationship of satellite and ground observations.
A wide time window is useful in remote areas where ground observations can not be
taken easily. A wide time window between collection of ground and satellite data would
allow more flexibility i.e. it may use a calibration data for remote areas. And in contrast a
narrow time window requires the routinely monitoring of ground observations and
satellite data for assessment of water clarity.

Second factor is the size of the area of interest (AOI). A best AOI is the area whose
spectral characteristics best match with the ground observation data. It is commonly
assumed that a best site which provides a representative sample of the whole lake is the
centre of the lake. During the data extraction process from the satellite imagery it should
be noted that the areas with vegetation and terrain features other then water should be
excluded from the AOI otherwise they will introduce false results in the final outcome.
These features could be excluded from the AOI by analyzing their spectral reflectance.

Third is the use of a regression equation to relate the ground observations and satellite
imagery. The use of regression equation allows the easy comparison of the results from
different images [2].

3
Mohsin Jamil and Majid Nazeer, 2009

2. Study Area

Our study area is Rawal Lake which is an artificial reservoir providing the water needs of
Rawalpindi and Islamabad. Rawal Lake is located within the Margalla Hills National
Park as shown in figure 1 and covers an area of 8.8 km².

The catchment area is 275 km² and maximum depth is 31 meters. It not only provides the
water needs of Rawalpindi and Islamabad but also gives a recreational site to the visitors.
The main source of water for Rawal Lake is Kurrang River which generates 84,000 acre
feet of water in an average rainfall year. The water of Rawal Lake is being continuously
polluted from many years by agricultural activities, residential areas near lake,
sedimentation of the water and the wastage of the poultry forms. It is important to take
adequate measures for the conservation of Rawal Lake, protection of environment of its
surrounding areas and improving the quality of water [8].

Figure 1: Location map of Rawal Lake showing the extent of lake.

The landscape of this area is rugged, with numerous valleys and many steep slopes. The
area falls in the monsoon belt and experiences two rainy seasons. The mean annual
rainfall is 951 mm and mean monthly maximum and minimum temperatures range from
16.9 C to 40.1 C and from 3.1 C to 24.7 C respectively [9].

4
Mohsin Jamil and Majid Nazeer, 2009

3. Methodology
The methodology consists of several steps as shown in the flow chart. The first step is to
collect the ground observation data of secchi disc transparency second is the satellite data
acquisition which should be cloud free third is the AOI creation of the SDT data from the
satellite image and finally a regression model is developed to find out the relationship
between the ground observation data and satellite data.

Study of Rawal
Lake

Landsat TM Secchi Disc Data


Images

Pre-processing
TM Imagery

Creation of AOI

Suitable TM
bands and band
ratios
Regression Model Carlson Trophic
Development State Index

Calculating r2 and Water Clarity


SEE Map of Rawal
Lake

Figure 2: Flow chart showing the steps of the methodology of the study.

3.1- Ground observation data collection

The secchi disc data for the Rawal Lake was collected by several ground surveys of the
lake to increase the positional accuracy of the points Garmin GPS 12CX was used. These

5
Mohsin Jamil and Majid Nazeer, 2009

observation points were used to create an AOI from the satellite image. In the mean while
the samples of lake water were collected to find out the concentration of chlorophyll-a
and phosphorus to determine the trophic condition of the lake. To avoid the distortion,
cloud free satellite imageries were selected.

3.2- Satellite data acquisition and pre-processing

The images used for the study have cloud cover of 10% or less. The images taken in
summer from May to September show better results because they have less cloud cover
and stable lake conditions. All the images selected for analysis were registered to
Universal Traverse Mercator (UTM) Zone 43 and WGS84 projection system. The
registration process used approximately 40 ground control points (GCP’s) for each
image. Nearest-neighbor resampling technique was used to maintain the original
brightness values of the image. After stacking the image a subset of the lake area was
made to reduce the file size of the whole image and to increase the processing speed of
the software. For this process ERDAS Imagine 9.1 was used.

Image classification refers to the automatic categorization of all the pixels in an image
into land cover classes or themes. There are two classification techniques, supervised and
unsupervised. In supervised classification the computer defines the classes based on the
training sites provided by the analyst. Supervised classification needs the prior
knowledge about the land cover type and its location. While the unsupervised
classification requires few inputs in classification process and defines classes by itself
into natural spectral clusters. Then the analyst compares the classified data with some
reference data to determine the information about the spectral classes. The advantage of
unsupervised classification is that it makes the clusters of the image pixels which sort out
the data into classes such that the between-class variance of the specified number of
groups is capitalized [10].

Here unsupervised classification method is applied to extract the information from the
satellite. In this process ten classes were made for the study site. The most representative
sample of the whole lake is provided by the centre of the lake [2]. The differentiating
property of water is that it absorbs the energy at near infrared and outside of that range.
The presence of suspended or dissolved materials in a water body effect the reflectance of
water, highly contaminated water signature shows high reflectance in visible portion of
the spectrum in comparison with the clear water signature while the increase in
chlorophyll concentration in the water shifts the signature from blue towards green [10].
The signatures obtained from different classes of the TM image are shown in figure 3.

6
Mohsin Jamil and Majid Nazeer, 2009

3.3- Creation of area of interest

The points of secchi disc readings were located on the satellite image and for each point
an AOI was created. The smallest AOI created from secchi disc reading contained 8
pixels and largest was of 1000 pixels. The size of AOI depends on the number of pixels
in a group having similar reflectance values. The values of band1 and band1/band3 from
these signature classes were used in the regression model.

Figure 3: Normalized mean brightness values from unsupervised classification for TM


scene.

3.4- Regression Model Development

Regression analyses are performed on Landsat TM images (between the years 1992-
2000) to calculate the relationship between the natural log of SDT and satellite radiance
data of band 1 and band1:band3. Many studies proved that the ratio band1/band3 is a
strong predictor of secchi disc transparency. The increase in the time window between
the collection of secchi disc data and satellite over pass decreased the value of regression
coefficient r2 and increased the standard error of estimate (SEE) [1,3,4].

The numbers of pixels included in the area of interest (AOI) affect the strength of
relationship between the collection of ground observation data and satellite data. If the
ground observations are contemporaneous with the satellite image then a small cluster of
pixels containing the location of the ground observation would provide the best
correlation between satellite and ground based data.

7
Mohsin Jamil and Majid Nazeer, 2009

Table 2: Regression equations for r2 and SEE determination from Landsat TM data.

Image Date Equation r2 Standard


Error
20 September 1992 ln (SDT) = - 0.41 (TM1) + 9.94 (TM1: TM3) + 12.13 0.82 0.041

17 June 1998 ln (SDT) = - 0.04 (TM1) + 0.23 (TM1:TM3) + 2.84 0.87 0.045

17 April 1999 ln (SDT) = - 0.03 (TM1) – 3.34 (TM1:TM3) + 10.44 0.51 0.146

5 May 2000 ln (SDT) = - 0.071 (TM1) – 0.14 (TM1:TM3) + 5.87 0.86 0.048

4. Results and Analysis:

Regression analyses were performed to observe the correlation between the water quality
parameters and TM bands and band ratio reflectance values. Regression model was used
to determine the values of regression coefficient r2 and standard error of estimate. The
regression analysis was performed on all the four images (1992, 1998, 1999 and 2000). In
our study the value of regression coefficient ranges from 0.51-0.87 and standard error of
estimate ranges from 0.041-0.146. The correlation strength was affected by the increase
in the ground observations. The regression analysis showed that there is a strong
relationship between the ground observation values and the Landsat TM brightness
values. The decrease in the regression coefficient value for the year 1999 is due to the
increase in the values of secchi disc transparency. The trophic condition of Rawal Lake
determined with the help of Carlson’s trophic state index showed that the Rawal Lake
falls in the category of eutrophic lakes. The trophic state was determined by the mean
secchi disc values and shown in Table 3;

Table 3: Trophic State Index Calculation from mean Secchi disc data, for different years.

Year SDT (m) TSI (SDT)


1992 0.73 64.5
1998 0.72 64.7
1999 0.66 66
2000 1.3 56.2

The eutrophic lakes are highly productive water bodies, supporting high levels of algae,
plant growth and dissolved materials. This type of water body cannot support the cold
water fisheries. The water clarity for the subjected years (1992, 1998, 1999 and 2000)
based upon the secchi disc transparency is shown in the images, which shows the clarity
level of Rawal lake in different colors.

8
Mohsin Jamil and Majid Nazeer, 2009

Scale Turbid Clear

Figure 4: Water clarity maps of Rawal Lake based on Secchi disc transparency for the years
1992, 1998, 1999 and 2000.

9
Mohsin Jamil and Majid Nazeer, 2009

The Landsat TM reflectance and ground observation data of secchi disc depth are plotted
(Fig; 3-a and 3-b) and a linear regression line was drawn which shows the SDT values
stretch out very close to the regression line.

Figure 5: Plot of TM bands vs Secchi disc depth. Showing the relationship between SDT and
brightness values of TM 1 and TM 1: TM 3 (a)- Plot of band1 and Secchi depth. (b)- Plot of
band1:band3 and Secchi depth.

5. Conclusions:

• The temporal coverage, spatial resolution and the data availability of Landsat
makes it very useful for the monitoring of inland water bodies. It is observed that
Landsat TM data has a strong correlation with the ground observation data.
• Through remote sensing techniques we can estimate the chlorophyll and
suspended sediments spread by interpreting the radiance data received through the
sensor.
• The water of Rawal Lake is highly polluted with algal mass and suspended
sediments which is the main reason for the eutrophication of the Rawal Lake.
• The phosphorus and chlorophyll-a are loaded in the lake water due to the urban
growth and agricultural activities in the lake vicinity. The phosphorus is mainly
contaminated in the lake water by the use of fertilizers, used for better yield of
crop in the lake vicinity.
• The water quality of Rawal Lake can be improved by using herbicides and
algaecides to stop the growth of aquatic plants.
• The reduction of the nutrient loading in a lake improves the water clarity and
increases its aesthetic and recreational value.

10
Mohsin Jamil and Majid Nazeer, 2009

References:

1. Cox, R. M., Forsythe, R. D., Vaughan, G. E. and Olmsted, L. L. (1998).


Assessing water quality in the Catawba River reservoirs using Landsat Thematic
Mapper satellite data. Lake and Reservoir Management, 14, pp 405– 416.

2. Steven M. Kloiber, Patrick L. Brezonik, Leif G. Olmanson and Marvin E. Bauer


(2002). A procedure for regional lake water Clarity assessment using Landsat
multispectral data. Remote Sensing of Environment, 82, pp 38-47.

3. Lathrop, R. G. (1992). Landsat thematic mapper monitoring of turbid inland water


quality. Photogrammetric Engineering and Remote Sensing, 58, pp 465– 470.

4. Lillesand, T. M., Johnson, W. L., Deuell, R. L., Lindstrom, O. M. and Meisner, D.


E. (1983). Use of Landsat data to predict the trophic state of Minnesota lakes.
Photogrammetric Engineering and Remote Sensing, 49, pp 219– 229.

5. Ruiqiu Li and Jonathan Li (2004). Satellite Remote Sensing Technology for lake
water clarity monitoring: An overview. Environmental Informatics Archives, 2, pp
893-901.

6. Carlson, R.E. (1977). A trophic state index for lakes. Limnology and
Oceanography, 22, pp 361-369.

7. Ritchie, J.C., Copper, C. M. and Schiebe, F.R. (1990). The relationship of MSS
and TM digital data with suspended sediments, chlorophyll, and temperature in
Moon Lake, Mississippi. Remote Sensing of Environment, 33, pp 137-148.

8. Report on Rawal Lake catchment area monitoring operation (2004). Pakistan


Environment Protection Agency. Ministry of Environment, Islamabad.

9. Michael John Beverley Green; World Conservation Monitoring Centre.; IUCN


Commission on National Parks and Protected Areas (1990). IUCN Directory of
South Asian Protected Areas. Cambridge University press.
Accessed June 2009 at URL;
http://books.google.com.pk/books?id=bs_t9aOtIk0C&dq=directory+of+south+asi
an+protected+areas&printsec=frontcover&source=bl&ots=_RKMVK61Xs&sig=
Mm3RefL7B2KxFWKCSKefRGpYQkI&hl=en&ei=V56JStOsKpiWkQX9uNSp
CA&sa=X&oi=book_result&ct=result&resnum=1

10. Lillesand, T.M., Kiefer, R.W. and Chipman, J.W. (2004). Remote Sensing and
Image Interpretation, Fifth Edition. John Wiley and Sons.

11. Fuller, L.M. and Minnerick, R.J. (2007). Predicting water quality by relating
secchi-disc transparency and chlorophyll a measurements to Landsat satellite

11
Mohsin Jamil and Majid Nazeer, 2009

imagery for Michigan inland lakes, 2001-2006. U.S. Geological Survey, fact sheet
no. 2007-3022.

12. Steven M. Kloiber, Patrick L. Brezonik and Marvin E. Bauer (2002). Application
of Landsat imagery to regional-scale assessments of lake clarity. Water Research,
36, pp 4330-4340.

12

Вам также может понравиться