Global Crop-Specific Fertilization Dataset from 1961-2019

Abstract

As global fertilizer application rates increase, high-quality datasets are paramount for comprehensive analyses to support informed decision-making and policy formulation in crucial areas such as food security or climate change. This study aims to fill existing data gaps by employing two machine learning models, eXtreme Gradient Boosting and HistGradientBoosting algorithms to produce precise country-level predictions of nitrogen ($N$), phosphorus pentoxide ($P_2O_5$), and potassium oxide ($K_2O$) application rates. Subsequently, we created a comprehensive dataset of 5-arcmin resolution maps depicting the application rates of each fertilizer for 13 major crop groups from 1961 to 2019. The predictions were validated by both comparing with existing databases and by assessing the drivers of fertilizer application rates using the model's SHapley Additive exPlanations. This extensive dataset is poised to be a valuable resource for assessing fertilization trends, identifying the socioeconomic, agricultural, and environmental drivers of fertilizer application rates, and serving as an input for various applications, including environmental modeling, causal analysis, fertilizer price predictions, and forecasting.

Figures (7)

• Figure 1: Outline of the process for generating the gridded crop-specific fertilizer dataset.
• Figure 2: SHapley Additive eXplanation (SHAP) values of the top 10 most important features in the prediction of, respectively, the crop N (a,d), P2O5 (b,e) and K2O (c,f) application rates using Histogram-based Gradient Boosted regression. (a,b,c) The top plots present the average feature importance, determined by the mean absolute SHAP value of each feature. (d,e,f) The bottom plots depict a SHAP value for each prediction and show the local feature importance and the feature effect. The color of a dot represents the value of the feature in that instance - red indicating relatively high, blue indicating relatively low values. A dot with a high SHAP value for a feature suggests a positive contribution to the prediction, whereas a negative SHAP value leads to a lower prediction. The features are ranked in order of descending average importance and the blue, green and orange squares indicate whether the feature is an environmental, agrological or socioeconomic characteristic.
• Figure 3: Comparison of the application rates per ha per year for various crops between our predicted model output and the data reported by the United States Department of Agriculture (USDA) for the USA.
• Figure 4: Comparison of the application rates per ha per year for various crops between our predicted model output and the data reported by the Department for Environment, Food & Rural Affairs (DEFRA) for the UK.
• Figure 5: Spatial pattern of crop-specific fertilizer (N) kg ha -1 consumed by each 0.05° grid cell for the following: a) average for the 1960s decade across all 13 crop classes, b) average for the 1960s decade for wheat, c) average for the 1960s decade for rice, d) average for the 1960s decade for maize, e) average for the 1960s decade for other cereals, f) average for the 1960s decade for all oil crops, g) average for the 1960s decade for vegetables and fruits, h) average for the 1960s decade for roots and tubers, sugar crops, fiber crops, and other crop classes, i) average for the 2010s decade across all 13 crop classes, j) average for the 2010s decade for wheat, k) average for the 2010s decade for rice, l) average for the 2010s decade for maize, m) average for the 2010s decade for other cereals, n) average for the 2010s decade for all oil crops, o) average for the 2010s decade for vegetables and fruits, p) average for the 2010s decade for roots and tubers, sugar crops, fiber crops, and other crop classes. The 1960s decade includes the years 1961-1969, and the 2010s decade includes the years 2010-2019