DroughtCast: A Machine Learning Forecast of the United States Drought Monitor.

Colin Brust,Rolf H Reichle,Marco P Maneta,John S Kimball,Kelsey Jencso

doi:10.3389/fdata.2021.773478

Colin Brust, Rolf H Reichle + Show 3 more

Open Access

https://doi.org/10.3389/fdata.2021.773478

Copy DOI

Abstract

Drought is one of the most ecologically and economically devastating natural phenomena affecting the United States, causing the U.S. economy billions of dollars in damage, and driving widespread degradation of ecosystem health. Many drought indices are implemented to monitor the current extent and status of drought so stakeholders such as farmers and local governments can appropriately respond. Methods to forecast drought conditions weeks to months in advance are less common but would provide a more effective early warning system to enhance drought response, mitigation, and adaptation planning. To resolve this issue, we introduce DroughtCast, a machine learning framework for forecasting the United States Drought Monitor (USDM). DroughtCast operates on the knowledge that recent anomalies in hydrology and meteorology drive future changes in drought conditions. We use simulated meteorology and satellite observed soil moisture as inputs into a recurrent neural network to accurately forecast the USDM between 1 and 12 weeks into the future. Our analysis shows that precipitation, soil moisture, and temperature are the most important input variables when forecasting future drought conditions. Additionally, a case study of the 2017 Northern Plains Flash Drought shows that DroughtCast was able to forecast a very extreme drought event up to 12 weeks before its onset. Given the favorable forecasting skill of the model, DroughtCast may provide a promising tool for land managers and local governments in preparing for and mitigating the effects of drought.

Highlights

IntroductionA single drought event can cause more than one billion dollars in damages and lead to the shift or degradation of entire ecological regimes (Crausbay et al, 2017; Smith 2020)
Drought is one of the most pervasive natural disasters affecting the United States
As in the mean squared error (MSE) and R2 performance seen in Table 2 and Figure 2, the confusion matrices show that model performance is best at smaller lead times in the training set and spatial holdouts, and that misclassifications are more common in the temporal holdouts and as the lead time increases

Summary

Introduction

A single drought event can cause more than one billion dollars in damages and lead to the shift or degradation of entire ecological regimes (Crausbay et al, 2017; Smith 2020). Meteorological, ecological, agricultural, hydrologic, and socio-economic droughts are all caused by a different combination of environmental and economic factors, making it difficult to create a single holistic definition of drought (Wilhite and Glantz 1985; IPCC, 2021). Flash droughts are characterized by their rapid onset, which tends to be driven by anomalously high temperatures, high evapotranspiration (ET), low precipitation, and low soil moisture (Otkin et al, 2018). Less common than typical droughts, flash droughts can pose a significant risk, as they have driven widespread crop and livestock losses leading to notable economic and ecological damage (Otkin et al, 2018; He et al, 2019)

Objectives

Methods

Results

Discussion

Conclusion