Moyna block in Purba Medinipur district (West Bengal) experiences drainage cramming and waterlogged because of substantial rainfall during the monsoon season and resultant standing water in low laying agricultural land (Sahu et al., 2009). This causes failure of crop cultivation, emergence of vector borne disease (e.g., malaria) and water-borne disease (e.g., cholera) and communication problems etc. consequently, the development of river canal during the period between 1985 and 1986, local people transferred low laying agricultural lands into pisciculture to earn more money and furtherance of economic situations (Sahu, 2014).
Hence, the spectral indices used in this study (NDWI, MNDWI, AEWI, WRI and WI) may effectively and precisely distinguish water pixels from non-water pixels by determining threshold value between water and non-water surface. Although, the determination of threshold is carried out manually. Another benefit of these remotely sensed spectral indices as these are simple and straightforward technique for demarcating more precise information and evaluating of the product output.
The projected method has the advantage of instantaneously automatic distinguishing the surface waterbodies and non-water surface using multi-temporal images through applying remotely sensed spectral indices. The results indicate that the performance of MNDWI and MNDWI closely matches of surface water detection efficiency in the study area.
The accuracy derived through RMSE also showed maximum accuracy was calculated through MNDWI index. Yang et al., (2011) and Rokni et al., (2014) are conducted similar experiment using remote sensing data to simulate the natural condition of water bodies. Consequently, the MNDWI could be convenient in identifying the dwindling or increasing vicissitudes in any open surface water in the world, as the behaviour of water is about analogous in different regions and dissimilar satellite images with similar band width. Our results also corroborated with the earlier study (Sarp and Ozcelik, 2016; Haibo et al., 2011).
The accuracy problem of waterbody extraction is perceptible where the background land cover includes low albedo surface, like cloud, buildings, shadows etc (Yang et al., 2018). As such, the minor differences of spectral variability between the pixels cannot be determined manually. Earlier study also suggests that NDWI output provide less accurate information for small waterbodies (Haibo et al., 2011); and the output derived through NWI, is not good for separating residential and waterbodies for medium resolution satellite data (Haibo et al., 2011).
The presence of shadows in the satellite data may cause misclassification because of analogous spectral reflectance patterns of the areas covered with waterbodies and this resemblance may lessen the accurateness of extracted surface water areas and vary the investigation between specific time intervals. Present study revealed that MNDWI is more appropriate for differentiating water in many built-up areas compared to using other spectral indices (Elsahabi et al., 2016).
The applied threshold value of the MNDWI used to obtain the best water extraction spectral index. Using ‘zero’ as a default threshold value, MNDWI can produce better waterbodies separation accuracy and provides more detailed information regarding open water than does the other spectral indices (Du et al., 2016). The information also useful for detection of water quality differences in specified time intervals.