Estimation of River Runoff Loadings by Classification of Land Cover Use Using QuickBird Remote Sensing Data and Stepwise Multiple Regression Analysis between Land Cover Use and Discharges of Nutrients and Organic Carbon

Abstract

We estimated the runoff loadings of nutrients and organic carbon to the Koise River from its catchment area, a major river emptying into Lake Kasumigaura. The catchment was divided into several subwatersheds on the basis of digital elevation data, and extensive surveys of water quality and quantity were conducted in those defined subwatersheds. The runoff loadings were estimated for each subwatershed by multiple regression analysis using land-cover use distributions derived from high-resolution satellite (Quickbird) imagery and the measured loadings. The stepwise method, which cases a nonlinear model requiring no initial values, was introduced in the regression analysis to improve reliability in the estimation of the loadings.
During the nonirrigated period, the unit loading factor from crop lands was highest for nitrogen, phosphorus, and organic carbon in the dissolved phase. The loading factor in the particulate phase from deciduous forests was constant for nitrogen and organic carbon, while that from crop lands was constant for phosphorus. The dominant unit loading factors of total nitrogen, total phosphorus, and total organic carbon were found to differ between the irrigated and nonirrigated periods.