2000-2020年玛纳斯河流域的作物种植结构与需水满足度

变化环境条件下干旱绿洲水土资源的高效利用一直广受关注。玛纳斯河流域分布着新疆最大、最典型的绿洲农耕区，其水土资源高效利用无疑应基于对种植结构和需水满足度（供水量/需水量）时空演变规律的了解。基于谷歌地球引擎（Google Earth Engine，GEE）平台并通过区域调查和调研，该研究利用面向对象的随机森林分类，建立流域地物遥感识别模型，分析当地2000-2020年种植结构的变化过程，探讨种植结构变化与膜下滴灌棉花水分供应状况和需水满足度间的关系。结果表明：基于GEE平台，融合简单非迭代聚类图像分割算法和随机森林分类算法可快速、准确识别流域地物，总体精度约90%；近20年，流域种植作物始终以棉花为主，占耕地总面积的80%以上，受益于膜下滴灌技术的节水、抑盐等功效，中、下游盐碱荒地不断被开垦为棉田，致使其面积以每年约101 km2的速度增长，但棉田面积增长与灌溉水资源供给的矛盾日益突出，棉花需水满足度显著下降，尤其是水资源相对匮乏的下游灌区及需水旺盛的夏灌期，2020年流域下游棉花夏灌期需水满足度已降至约46%，种植结构的调整和优化已势在必行。研究可为玛纳斯河流域水资源配置及农业种植结构优化提供科学依据。

Crop planting structure and water demand satisfaction degree in Manas River Basin from 2000 to 2020

Abstract: Efficient utilization of water and soil resources has been critical under the changing environmental conditions in the arid oasis. It is very necessary to understand the temporal and spatial evolution of crop planting structure and crop water demand satisfaction degree (i.e. the ratio of supplied water amount to water demand) for the efficient use of water and soil resources and sustainable development of agriculture in the arid Manas River Basin, including the largest typically agricultural oasis in Xinjiang of western China. In this study, a classification model was established for the surface features via the remote sensing inversion using the Google Earth Engine (GEE) cloud computing platform, together with the coupled Simple Non-Iterative Clustering super pixel image segmentation and Random Forest classifier. The classification model was verified using a regional survey at the fixing sampling positions with the GPS. Then, the model was applied to the crop planting structure in the Manas River Basin over the past 20 years from 2000 to 2020. The Penman-Monteith formula was used to calculate the effective precipitation in the irrigation schedule from the regional survey and the water demand of the main crop (drip-irrigated cotton under film mulch). A systematic investigation was made to explore the effect of the crop planting structure on the irrigation water supply and water demand satisfaction degree for the drip-irrigated cotton under film mulch. Results showed that the new remote sensing inversion model was reliable for the surface feature classification in the Manas River Basin, with an annual average overall accuracy of 90%. The planting crop of the basin was dominated by drip-irrigated and film-mulched cotton, accounting for more than 80% of the total planting area in all the last 21 years, and mainly distributed in the midstream and downstream regions, where the climate and hydrological conditions were more suitable for the cotton growth. Beneficial from the water-saving and salinization-alleviating characteristics of drip irrigation under the film mulch, a lot of mildly- and moderately- or even severely-salinized wastelands were continuously reclaimed into the cotton fields, resulting in an increased rate of about 101 km2 yr-1 in the area. There was an ever-increasing prominent contradiction between the expansion of cotton fields and the limited supply of irrigation water resources, particularly with a significant declining trend in the water demand satisfaction degree for cotton. The irrigation amount and water demand satisfaction degree in downstream were generally lower than those in the midstream, due to the low irrigation and drainage conditions, especially when the crop was in the peak water demand during the summer irrigation period. The average water demand satisfaction degree decreased even to less than 53% over the entire basin in the whole irrigation period by 2020. It is a high demand to optimize the crop planting structure in the arid Manas River Basin.