Study on extraction methods for water information in Nantong city, China using Landsat ETM+ data

Abstract

To select an optimal method for Nantong region, the study compared normalized difference water index (NDWI), modified normalized difference water index (MNDWI) and relation of spectrum to extract water information with Landsat ETM+ data. Supervised classification that needs training site was used to evaluate the results. Relation of spectrum could accurately extract wide water body such as Tong-Lu canal, but it was difficult to extract a smaller water body. Moreover, relation of spectrum could not extract water information covered by sparse vegetation. In contrast, NDWI and MNDWI did better. The water information by NDWI always mixed with built-up land noise and thus the area of a water body was overestimated. MNDWI could reveal more subtle features such as the distribution of suspended sediment, water quality changes than NDWI. MNDWI could significantly highlight the water information, especially Haohe river that surrounds the city center. MNDWI could depress the buildings' effect while highlighting water information and accurately extract the water information. MNDWI revealed successfully significant non-pollution of the water. Also, MNDWI could easily distinguish shadow and water body and solve the problem to eliminate shadow in extracting water information. MNDWI was a better method for extracting water information in Nantong city.