Prediction of pesticide runoff at the global scale and its key influencing factors.

The prediction of pesticide loss in runoff water is a critical step in quantifying pesticide pollution potential and risks. Herein, we compiled a global database and developed a machine learning model to predict the runoff loss of 92 widely used pesticides at the global scale. We found that the pesticide runoff loss rates were mostly influenced by soil properties and rainfall volume. The predicted runoff loss rate of very mobile (VM) and nonmobile (NM) pesticides varied with latitude. Moreover, 2.30 % and 0.55 % of the global agricultural area were classified as "High potential" for pollution caused by VM and slightly mobile (SM) pesticides, respectively, according to the high water risk and high runoff loss of pesticides. The pollutions potential of mobile (M), moderately mobile (MM), and NM pesticides were classified as "Medium and low potential" worldwide. Among the "High potential" areas of VM and SM pesticide pollution, there were 24.36 % and 42.42 % area in low-income and lower middle-income nations, which can cause more serious pesticide pollution problem due to their backward agricultural management strategies and agricultural infrastructure construction. We identified eastern and southern Asia (China, India, Pakistan, and Turkmenistan) and southern Europe (mainly Ukraine, Spain, and Italy) as high-risk regions of pesticide contamination. This study is the first to predict the runoff loss of pesticides at a global scale.