Rice paddy inventory in a few provinces of China using AVHRR data

Abstract

Times series Normalized Difference Vegetation Index (NDVI) data, computed from Advanced Very High Resolution Radiometer (AVHRR) data, were used in an attempt to locate areas of rice cultivation in China. NDVI dynamics were examined using 16 km global area coverage satellite data from 1988 composited into 12 monthly images. Unsupervised classification was used to categorize the data with the hope of distinguishing rice cultivation areas from other vegetation cover types. There was no ground truthing. Results from the classification were compared to a digital vegetation map of China and an elevation map. On an acreage basis, between 11% of the rice area in Hunan (described as such on the vegetation map) and 92% of the rice area in Sichuan coincided with classes obtained by the unsupervised classification routine. The rice cultivation signal, low at planting and flooding, high at peak vegetative growth, low at harvest, was calculated as the average from all the pixels belonging to a particular class, (i.e. n = 150 to 800 pixels), and as a result, was dramatically dampened from the original signal.