Crop yield forecasts allow policy makers to anticipate market behaviour and regulate prices. Annual updates on which crops are grown where can improve crop yield forecast accuracy. Existing efforts to map crops across the European Union resulted in late-season map availability or short time series that do not meet forecasting requirements. We propose a new approach to retrieve annual winter crop maps and improve forecasting efforts by identifying pixels with dominant winter crop signals using moderate resolution imagery. These pixels are distinguished from summer crop signals based on their senescence date. When this date precedes the theoretical maturity date of a winter crop, expressed in GDD, the pixel is labelled as having a dominant winter crop signal. Our 2018 map accurately identified 77% and 83% of dominantly winter-crop area, when compared to farmers’ declaration data and a high-resolution crop map for Europe, respectively. While the resulting annual winter crop maps underestimated winter crop area, derived region-specific annual NDVI profiles better described winter crop phenology as compared to the use of static maps. Regression analysis between these regional NDVI profiles and statistical wheat yield data indicates that our annual maps help explain more yield variability than static maps, with an RMSE reduction of 3% for the EU27 as whole. The proposed approach is applicable to long historical timeseries and provides maps before the end of the agricultural season. Those maps positively impact crop yield description, notably in eastern, northern, and northeastern European regions.

中文翻译：

---

MODIS NDVI 时间序列的年度冬季作物分布可改善欧洲的产量预测


农作物产量预测使政策制定者能够预测市场行为并调节价格。每年更新哪些作物在哪里种植可以提高作物产量预测的准确性。现有的欧盟农作物地图绘制工作导致地图可用性晚或时间序列短，无法满足预测要求。我们提出了一种新方法来检索年度冬季作物地图，并通过使用中等分辨率图像识别具有主导冬季作物信号的像素来改进预测工作。这些像素根据其衰老日期与夏季作物信号区分开来。当该日期早于冬季作物的理论成熟日期（以 GDD 表示）时，该像素被标记为具有主导冬季作物信号。与欧洲农民申报的数据和高分辨率作物地图相比，我们的 2018 年地图分别准确识别了 77% 和 83% 的冬季作物占主导地位的面积。虽然由此产生的年度冬季作物地图低估了冬季作物面积，但与使用静态地图相比，得出的特定区域年度 NDVI 剖面更好地描述了冬季作物物候。这些区域 NDVI 概况和统计小麦产量数据之间的回归分析表明，我们的年度地图比静态地图有助于解释更多的产量变异性，整个欧盟 27 国的 RMSE 降低了 3%。所提出的方法适用于长期历史时间序列，并在农业季节结束之前提供地图。这些地图对作物产量描述产生积极影响，特别是在东欧、北欧和东北欧地区。