Analysis of modified water index (MWI) for extraction of water bodies in landsat-8 imagery

Abstract

Remote sensing techniques play an important role in exploring and management of water resources on the surface of the earth. Water bodies' identification from multispectral imagery is mainly done using several water indices. Water indices evolved based on the reflectance variations in multispectral imagery from different water bodies. Normalized Difference Water Index (NDWI) is mostly popularly used water index for detection of water bodies. The NDWI water index is used to classify the clear water bodies from non-water bodies and mixed water bodies. Modified Normalized Difference Water Index (MNDWI) is used to classify the clear water bodies along with mixed water bodies from non-water bodies. The selection of appropriate indices significantly affects the performance accuracy in extraction of water bodies. This paper aims in proposing simple and effective water index based on multi bands, Blue, Green, NIR and SWIR2.The objective of Modified Water Index (MWI) is better handling classification of the mixed water pixels. The Blue, Green bands has high reflectance values and Near Infrared (NIR), Short Wave Infrared (SWIR2) bands has low reflectance from water bodies. The significant reflectance variance feature of SWIR2 from clear water to mixed water bodies is useful in the classification of the mixed water pixels in the MWI. The performance of the Modified Water Index is analyzed and compared with the existing several water indices. The performance metrics such as Water Spread Area, mean and standard deviation are used to compare the effectiveness of water body indices.