Using remote sensing to extract and correct irrigation data during early cotton growth stage

Vegetation index is affected by background soil, especially in the early stage of crop growth. When vegetation cover is low, the effect of background soil is very obvious. In order to improve the precision of remote sensing (RS) monitoring on crop growth in the early growth stage, it is necessary to eliminate the effect of background soil moisture due to irrigation on normalized difference vegetation index (NDVI). Agricultural irrigation districts have failed to develop an effective method to eliminate difference in NDVI change, which has in turn hindered efforts to limit the effect of irrigation on NDVI. Thus, in order to increase the accuracy of RS monitoring of crop growth at early stage, this study explored the effects of difference in soil moisture information between irrigated and non-irrigated cotton field on NDVI. Two cotton plots in San Joaquin Valley in California (US) were selected as the research area. Day of Year (DOY) 174 was determined as the critical phase at early growth stage of cotton for the extract of irrigation data through band reflectance, NDVI analysis of cotton field for 2002. Based on RS images, NDVI, normalized difference water index (NDWI), soil adjusted vegetation index (SAVI) and modified soil adjusted vegetation index (MSAVI) of irrigated and non-irrigated pixels were calculated. Also the relationships between NDWI and different vegetation indexes (VIs) were analyzed, and the two methodsthe standard deviation of the NDWI method (STD_WI) and irrigation line extraction method (based on relationship between NDVI and NDWI of irrigation and non-irrigation pixels, IR_L) were used to extract the irrigation data. Then the accuracies of different methods were compared to determine the optimum extraction method of irrigation information. The IR_L method was next used to extract irrigation data and correct the NDVI of irrigation pixels in the early stage of cotton to improve monitoring accuracy of cotton growth. The results showed that difference in NDVI between irrigation and non-irrigation pixels was as high as 12% in the early growth of cotton. There was an extremely significant linear correlation between NDVI and NDWI of both irrigation and non-irrigation pixels, with coefficients of determination greater than 0.80. Compared with STD_WI method, IR_L method had a higher accuracy and with a precision greater than 88%. Through IR_L model correction, the accuracy of irrigation linear regression model was as high as 0.95. With this, correction effect of irrigation was obvious and the difference in NDVI between irrigated and non-irrigated pixels dropped to 2%. Thus in this study, NDVI with irrigation data was corrected, the effect of irrigation on NDVI eliminated while the effect of background soil moisture reduced. Finally, the study reflected the true vegetation data, obtained accurate remote sensing monitoring of cotton growth at the early growth stage and provided convenient monitoring method of crop growth via remote sensing. Moreover, it promoted accurate irrigation towards saving water resources.