Predicting agricultural drought indicators

The research compared seven machine learning and deep learning models using three data arrangement methods to predict soil moisture and the Palmer drought severity index (PDSI) in Sweden. The study found that the temporal ARIMA model was most effective for local-scale soil moisture prediction, while the multi-features Random Forest (RF) model performed best for national-scale predictions. Overall, soil moisture was easier to predict than PDSI, with previous-year average monthly soil moisture identified as the most critical feature for prediction and mapping.

• The temporal ARIMA model was the most suitable for local scale soil moisture prediction, while the multi-features Random Forest (RF) model was more suitable for national-scale prediction, achieving Mean Absolute Errors (MAE) of 9.1 and 11.95 and R2 values of 0.79 and 0.59, respectively.
• Predictive models generally showed higher performance when predicting soil moisture compared to the Palmer drought severity index (PDSI).
• The most significant feature for predicting and mapping drought indicators was the average monthly soil moisture from the previous year, which was most effective when combined with elevation, precipitation, evapotranspiration, and PDSI.

APA

Stockholm Environment Institute (2023). Predicting agricultural drought indicators. https://www.sei.org/publications/predicting-agricultural-drought-indicators/

Chicago

Stockholm Environment Institute. Predicting agricultural drought indicators. 2023. https://www.sei.org/publications/predicting-agricultural-drought-indicators/.

Wikipedia

{{cite report |author=Stockholm Environment Institute |title=Predicting agricultural drought indicators |date=27 September 2023 |url=https://www.sei.org/publications/predicting-agricultural-drought-indicators/ |access-date=17 August 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{stockholmenvironmentinstitute2023predicting, author = {{Stockholm Environment Institute}}, title = {{Predicting agricultural drought indicators}}, institution = {Stockholm Environment Institute}, year = {2023}, month = sep, url = {https://www.sei.org/publications/predicting-agricultural-drought-indicators/}, urldate = {2026-08-17}, note = {Indexed by Climate Insights Directory} }