Abstract:  

Limited water resources is the major factor affecting sustainable development in arid areas,and the water resources in arid areas are mostly used for agricultural irrigation.Rapidly and accurately mapping cropping structure in arid areas can provide an important basis for optimizing water use in agriculture.In this study,the Northern Tianshan Economic Belt,Xinjiang,China was chosen as the study area,and a method to map the cropping structure in this region with multi-source remote sensing data based on the Google Earth Engine (GEE) cloud platform was proposed.The Sentinel-2 and Landsat 7-8 data were chosen as the remote sensing data sources to extract the cropping structure in the study area through the following steps. First, in order to simplify the cropping structure extraction process and minimize the impacts from non-crop vegetation, a cropland mask was constructed by using the maximum NDVI value and slope information throughout the year in the study area.Second,according to the phenology calendars of the main crops in the study area, the time-series data of the maximum NDVI value and the corresponding date was calculated with remote sensing data. Then, the 10 feature bands were constructed. Third, the 10 feature bands were masked with the cropland mask. Based on these data together with the field samples, the random forest classifier was applied to cropping structure extraction. The accuracy evaluation results showed the overall accuracy of the classification results in 2018 was 92.19%,and the Kappa coefficient was 0.883.In order to further verify the accuracy of the classification algorithm, the crop structure in the study area in 2017 was also extracted, the classification results showed that the planted area of cotton, corn and wheat in the study area were 5 270 km²,2 000 km² and 2 340 km² respectively in 2017, and then compared it with the results of statistical yearbook data in 2017.The relative accuracy of cotton, corn and wheat planted area were 86.53%,77.54% and 86.19%,respectively.

Key words: remote sensing in agriculture')">remote sensing in agriculture, Google Earth Engine, random forest classification, crop structure extraction, oasis agriculture