Projected Global Fertilizers Consumption Datasets during 2020-2100 under SSP scenarios

1. Background

Accurate projections of future global fertilizer consumption are critical for advancing research in earth system modeling, agricultural sustainability, and fertilizer industry planning. However, existing datasets often lack long-term temporal coverage and high spatial resolution. To address this gap, we present the Projected Global Fertilizers Consumption Datasets (PGFCD), which provide spatially explicit estimates of nitrogen (N), phosphorus (P), and potassium (K) fertilizer consumption from 2020 to 2100 under three Shared Socioeconomic Pathway (SSP) scenarios: SSP1 (sustainable development), SSP2 (intermediate development), and SSP3 (regional rivalry).

2. Methodology

2.1 Ensemble Machine Learning (EML) Framework

Algorithms: Integrated six machine learning models:

Multiple Linear Regression (MLR)

Decision Trees (DT)

Autoregressive Integrated Moving Average (ARIMA)

Multi-Layer Perceptron (MLP)

Radial Basis Function (RBF)

Random Forests (RF)

Training Data: Historical national/regional fertilizer consumption (FAOSTAT, 1961–2015).

Validation Metrics:

Nash-Sutcliffe Efficiency (NSE): 0.93

Kling-Gupta Efficiency (KGE): 0.89

Mean Absolute Percentage Error (MAPE): 10.97%

2.2 Spatial Downscaling

Baseline: FAO 2000 fertilizer data combined with gridded nutrient application maps for major crops in 2000.

Dynamic Projection: Annual change rates applied to 5′ × 5′ grids under SSP-specific socioeconomic drivers.

3. Dataset Overview

3.1 Key Features

Temporal Coverage: 2020–2100 (10-year intervals).

Spatial Resolution: 5-arcminute (≈10 km at the equator).

Scenarios: SSP1, SSP2, SSP3.

Variables: N, P, and K fertilizer consumption (tonnes/year).

3.2 Dataset Structure

The dataset is provided as a compressed archive (PGFCD.rar), containing:

1. Fertilization_Consumption_GeoTiff/

Subfolders:

N_fer/: Nitrogen fertilizer projections

P_fer/: Phosphorus fertilizer projections

K_fer/: Potassium fertilizer projections

File Format: GeoTIFF (81 files total).

Naming Convention:[FertilizerType]_fer_con_[SSP]_[Year].tif

Example: K_fer_con_SSP1_2020.tif

2. Country_region_based/

Shapefiles: National/regional fertilizer consumption for 26 prediction units (2020–2100).

Excel File: Global Fertilizer Consumption (2020-2100)_version3.xlsx.

3. Technical Annex.docx

Detailed methodology, validation, and workflow documentation.

4. Applications

This dataset supports:

Earth System Modeling: Improved parameterization of fertilization impacts on biogeochemical cycles.

Agricultural Policy: Scenario-based planning for sustainable fertilizer use.

Industry Strategy: Long-term market analysis under diverse socioeconomic pathways.

Note:

Global aggregated totals of fertilizer consumption derived from the 26 prediction units (country/regional scale) may exhibit minor discrepancies compared to sums calculated from the 5-arcminute gridded data (≈10 km resolution). Such differences stem from variations in spatial aggregation methods, file formats (vector vs. raster), and underlying data processing frameworks. Users may select the dataset best aligned with their analytical objectives:

The country/region-based data (26 units) is recommended for national-scale analyses or policy evaluations requiring administrative boundaries.

The 5-arcminute gridded data is preferable for spatially explicit modeling or subnational assessments.

Both datasets maintain equivalent quality and methodological rigor; the choice depends on the desired spatial granularity and application context.

We are profoundly indebted to Dr. Andreas Gericke at Section II 2.3 Protection of the Seas and Polar Regions, German Environment Agency, and Dr. Veronika Schlosser at Chair of Sustainability Assessment of Food and Agricultural Systems, Technical University of Munich for their diligent review and insightful feedback on the previously submitted data. Their expertise has enabled us to thoroughly correct the identified inaccuracies, strengthening the integrity of our research.

We hold Dr. Andreas Gericke and Dr. Veronika Schlosser in the highest esteem and sincerely apologize for any oversights that may have marred our work.