Surface water such as in lakes, dams forms the major resource for agriculture, drinking purpose, industries, fishing and many other application and nowadays human activity has been found to be polluting this resource. Inspite of expensive and time consuming conventional technique of studying water quality (Khorram et al., 1991), remote sensing provides spatial, spectral and temporal information about water which paves a way for effective monitoring , mapping and assessing water quality (Youichi Oyama et al., 2009). Remote sensing technique also yields high correlation coefficient ranging from .86 to .96 with the In-situ field data (K. Kallio et al.,2001). Concentration of chlorophyll, fluorescence, dyes, suspended solids, yellow substance, chemicals are some of the optical parameters that determines the water quality (G.Corsini et al.,). Many empirical (O'Reilly et al., 1998) and semi-analytical algorithms (Carder et al.,1999; Ciotti et al.,1999) are used for retrieval of water quality. Empirical algorithms with cubic or power polynomials are relatively simple and take less time for calculating. But many empirical algorithms has difficulty in estimation of water quality and its physio- chemical characteristics, because it has to be taken as a nonlinear, multiple variable subject (Bukata et al; Keiner and Yan, 1998). Hence neural networks have been widely applied to the nonlinear transfer function approximation. A primary use for NN's has been in the area of classification and pattern recognition (Atkinson and Tatnall 1997). The network's response is relatively insensitive to small variations in the input after training it. This allows the network to generalize, i.e. to disregard noise and detect the underlying pattern, an ability that is vital in remote sensing (L . E. Keiner and C. W. Brown., 1999). The information content from individual bands of remote sensing data is indeed very low sometimes and discrimination between subtle changes of, e.g. turbidity and sediment deposition. By using ratios and other combinations of spectral bands the ability to detect changes in the surface water can be increased, since more information is included in the retrieval process (C. Ostlund et al., 2001). In order to decide optimum band combinations of available bands the artificial neural network is integrated with evolutionary algorithm since they are well fit for optimization.
Some oceanic observation satellite sensors such as CZCS (Coastal Zone Color Scanner) and SeaWiFS (Sea-viewing Wide Field Sensor) have been used to measure water quality. However those data cannot provide detailed information about the spatial distribution of water quality in lakes which at a small- or medium-scale (LANDSAT, SPOT and IRS) having coarse spatial resolution (Liu et al., 2003). WorldView-2 launched 2009, the first commercial satellite to offer 8-band capability with unsurpassed accuracy, agility, capacity and spectral diversity expected. The four additional color bands: Coastal Blue, Yellow, Red Edge and Near-Infrared 2 enable the various effective applications and paves the way for analysts to discriminate features more accurately and increase the scope of remote sensing applications.
Therefore the objective of this proposal is 1) To design an optimized network to map and retrieve water quality using world view -2 data 2) To improve the retrieval accuracy by using evolutionary algorithm amd texture information using (GLCM) from newly introduced bands 3) To find the impact of water quality on surface sub-surface waters of surrounding area.
Study area:
Tirupur is an Indian textile town which constitutes many polluting industries like textile dyeing, bleaching situated in upstream of Orathupalayam dam. Tirupur serves as one of the major exporters of textiles. The industrial pollution have affected not only the surface water but also the soils and ground water. In Tirupur 1981, only 68 textile processing units were functioning. The number rapidly increased to 450 in 1991 and600 in 1997. At present over 900 textile processing unitsare functioning. About 75 to 100 million liters of effluents are released every day, which carry considerable volume of chemicals used at the wet processing stage. The continuoes discharge of untreated effluents for more thana decade has accumulated in the soil, ground water, etc, at Tirupur and Surroundings. A systematic study has been carried out to assess the surface water as well as the underground water contamination and the effect of textile effluents in and Orathupalayam dam.
The orathupalayanDam is located on the border between Kangayamand Perundurai Taluks in Erode District. The dam was constructed bybuilding a barrier across the Noyyal River. The Latitude is 11deg 24 'N and the longitude of the dam is 77deg 45'E. the Orathupalayam dam is chosen for water quality analyses and the surrounding area for impact studies as study area for the proposal. The water spread area is 1049 acres and it is used to irrigate 500 acress in Erode district and 9875 acres in Karur district. It was built in order to utilize the water in the river during the monsoon season and thisdam is the focus of the study.
Need for study: thewater in theorathupalayam dam has destroyed the soil structure. Seeds did not germinate after irrigation with polluted water. Agricultural farmers lost their working oppurtunities due to the pollution.It hasdrastic effects on the surface and ground water quality. The polluted water alsohas a serousimpacton aquatic life. Animals that drank the contaminated water had poor health.the release of untreated effluents from the textile industries in the orathupalayam dam had moved from theaffected hamlets and started to work in textile industries, coconut mills and tanneries in near by cities.Local people need to get updated information about the quality of water. Key environmental issues associated with textile manufacture are water use treatment and disposal of aqueous effluent. Treatment for colour removal can increase the risk of pollution. Treating azo-dyes results in production of amines which could be a greater environmental risk than the dye itself. Also, dye baths could have high level of BOD,COD,Colour,toxicity,surfactants, fibres and turbidity, and the may contain heavy metals. Textile effluents are high coloured and saline, contain non- biogedreadble compounds and are high in Biochemical and Chemical Oxygen demand. Dying process usuallycontributes chromium, lead, zinc and copper to the water. There was ahigh number of plankton and a higher plankton and fish diversity in the downstream of the dam because of the diluting effect of seepage water from the canal. Indicators such as liver glycogen and blood glucose level in fish confirmed that the water in the dam is more toxic to fish than water .
Advantages of WV-2 : world view -2 with 0.46 m spatial resolution and eight spectral bands allows for various environmental application. It is of importance to make use of the features that are being extracted to maximize the utility of additional spectral bands of WV-2.
The additional bands contains maximum information contents that are used for estimating water quality. Aquatic vegetation/turbidity due to various dyes and bleaching agents is very sensitive to response around 700 nm and the Red Edge band can be used to detect low density aquatic vegetation. The red edge is often used for detection of algae.
It has also been observed (Dekker, 1993; Gitelson et al., 1993. that the existence of inorganic suspended sediment and yellow substance have not been studied with existing data sets. WV-2 make it easier by giving reliable texture information of yellowness in the yellow edge (585 to 625nm) and inorganic sediments in (545.3_554.2) nm.
The Coastal Blue (400- 450 nm) in combination with Blue band allows penetration of clear water up to 45 feet. The Coastal and Blue bands is used for mapping of turbidity, suspended sediments, and depth of suspended sediments.
Using band combinations of the available bands enhance the discrimination of aquatic life and its condition. For instance, R 605/R760, is used to map different stress levels of agriculture around the study area and the ratio of WV-2 Yellow Edge to Red Edge will enable the mapping of vegetation health to analyse the major impact of the study.
Initialization of bands
Mutation
Cross over
Band Combinations
Individual Bands (1, 2, 3, 4, 5, 6, 7,8)
Spectral Image
Texture Image (GLCM)
Training Sample Extraction
Training Sample Extraction
ANN Training based on field data collected
Evidence Integration
Accuracy Assessment of Water Quality
Assessment of Water quality and mapping its spatial distribution
If previous band combination's accuracy is less than present accuracy
Step-1: Differential Evolution for optimization
of band combinations
+
+
Spectral Image
Texture Image (GLCM)
Training Sample Extraction
Training Sample Extraction
ANN Training based on field data collected
Evidence Integration
Accuracy Assessment of Water Quality
No
Assessment of Water quality and mapping its spatial distribution
If previous band combination's accuracy is less than present accuracy
yes
Yes
