A crop type dataset for consistent land cover classification in Central Asia

Land cover is a key variable in the context of climate change. In particular, crop type information is essential to understand the spatial distribution of water usage and anticipate the risk of water scarcity and the consequent danger of food insecurity. This applies to arid regions such as the Aral Sea Basin (ASB), Central Asia, where agriculture relies heavily on irrigation. Here, remote sensing is valuable to map crop types, but its quality depends on consistent ground-truth data. Yet, in the ASB, such data is missing. Addressing this issue, we collected thousands of polygons on crop types, 97.7% of which in Uzbekistan and the remaining in Tajikistan. We collected 8,196 samples between 2015 and 2018, 213 in 2011 and 26 in 2008. Our data compiles samples for 40 crop types and is dominated by “cotton” (40%) and “wheat”, (25%). These data were meticulously validated using expert knowledge and remote sensing data and relied on transferable, open-source workflows that will assure the consistency of future sampling campaigns.<br>

土地覆被是气候变化研究中的关键变量。具体而言，作物类型信息对于明晰水资源消耗的空间分布、预判水资源短缺风险及其引发的粮食不安全隐患至关重要。这一点在中亚咸海流域（Aral Sea Basin, ASB）等干旱地区尤为适用，该区域农业高度依赖灌溉。在此场景下，遥感技术可有效实现作物类型制图，但其制图质量依赖于一致可靠的地面实测数据（ground-truth data）。但咸海流域目前尚缺乏此类数据。为解决这一问题，我们采集了数千份作物类型矢量多边形样本，其中97.7%的样本采集自乌兹别克斯坦，剩余样本取自塔吉克斯坦。我们在2015至2018年间采集了8196份样本，另包含2011年的213份与2008年的26份样本。本数据集涵盖40种作物类型的样本，其中占比最高的两类为棉花（40%）与小麦（25%）。所有数据均通过专家知识与遥感数据完成严谨校验，并依托可推广、可复用的开源工作流进行采集，可保障后续采样工作的一致性。