Newswise — Scientists at the National Center for Atmospheric Research (NCAR) have devised a novel computer modeling method that has the capacity to produce drought predictions for the upcoming summer months in the Western United States. This technique holds the promise of distinguishing between arid circumstances at nearby sites separated by only a few miles.
Utilizing statistical approaches and machine learning, the methodology examines crucial indicators of drought during the winter and spring. It establishes correlations between these indicators and the probability of dry conditions across the entire landscape in the subsequent summer. The researchers suggest that if this method is adopted by forecasters, it could offer valuable insights for managing water resources, wildland fire and fuels, as well as agriculture, addressing pertinent priorities.
"The technique enables the prediction of drought conditions prior to their most significant impact," stated Ronnie Abolafia-Rosenzweig, a scientist at NCAR and the primary author of a recent paper outlining the methodology. "It provides managers with an extra resource to prepare for and inform their decision-making processes."
Abolafia-Rosenzweig and colleagues discovered that forecasts made between one to three months ahead were able to accurately predict summer drought occurrence with an accuracy ranging from 81% to 94%. These forecasts were conducted at a spatial resolution of 4 kilometers (2.5 miles) in the arid and rugged western third of the United States. The highest level of accuracy was observed in regions characterized by persistent drought, illustrating the ability to differentiate upcoming dry conditions between cultivated fields, nearby mountainsides, and forested areas. However, the predictions were less precise in areas where periods of heavy summer precipitation intermittently interrupted dry spells.
The scientists documented their research findings in a recent publication in Water Resources Research, a scientific journal published by the American Geophysical Union. The study received financial support from various organizations, including the National Oceanic and Atmospheric Administration (NOAA), the U.S. Geological Survey, and the U.S. National Science Foundation, which serves as the sponsor for NCAR.
Strengthening societal resilience
Droughts can have severe consequences on both public health and the economy, with the United States incurring a minimum cost of $249 billion since 1980 and serving as a catalyst for extensive wildfires. In the Western region, tree ring data reveals that the period from 2000 to 2021 marked the driest 22-year interval since at least the year 800. The year 2021 witnessed a significant toll, with the drought and accompanying heatwaves contributing to the loss of hundreds of lives in the affected region.
In order to enhance societal resilience, scientists are actively engaged in enhancing computer modeling techniques for generating drought predictions several months in advance. Presently, existing drought forecasts suffer from a limitation in resolution, typically around 10 kilometers at best. This coarse resolution fails to capture the diverse levels of drying experienced across different landscape features in the Western region. As a result, efforts are being made to refine these modeling techniques and improve their ability to provide more precise and localized drought forecasts.
The collaboration between NCAR scientists and the U.S. Geological Survey has resulted in the development of a new dataset called CONUS404, which has played a pivotal role in enabling more detailed drought forecasts. This dataset encompasses simulations of hydrological and climate conditions at a resolution of 4 kilometers across the entire continental United States (CONUS) for a span of over 40 years. In addition, the researchers utilized the high-resolution meteorological observations from the U.S. Department of Agriculture's PRISM (Parameter-elevation Regression on Independent Slopes Model) dataset. These comprehensive and finely-grained datasets have provided valuable inputs for improving the accuracy and resolution of drought predictions.
Thanks to the availability of the CONUS404 and PRISM datasets, the scientists were able to explore intricate connections, with a resolution as fine as 4 kilometers, between climate patterns and drought conditions in the late fall and winter, and the subsequent extent of drying during the summer. To uncover these relationships, the researchers employed machine learning techniques, training specialized statistical models to analyze the data and extract meaningful insights. These advanced methodologies allowed them to uncover valuable patterns and correlations that contribute to more accurate and detailed drought predictions.
The scientists directed their attention towards pre-summer climate variables, including temperature, precipitation, and humidity, in addition to distant ocean-atmosphere patterns such as the Pacific Multidecadal Oscillation, which exert widespread influence on climate patterns. They observed that widely utilized drought indicators, such as the Palmer Drought Severity Index and Soil Moisture Percentiles, exhibit significant persistence from the winter and spring seasons into the summer. As a result, the severity of drought conditions prior to the summer season emerges as a particularly crucial factor in predicting summer drought conditions accurately. This finding underscores the importance of considering pre-summer drought severity as a key predictor for forecasting drought outcomes during the summer months.
Abolafia-Rosenzweig highlighted that the newly developed drought forecasting method can complement a fire prediction technique previously developed by him and his co-authors. By combining these two models, there is a possibility of obtaining a highly detailed assessment of fire hazard throughout the Western region. This integration of drought and fire models holds promise for providing comprehensive insights into the risk and potential impact of wildfires, contributing to more effective fire management and mitigation strategies.
Abolafia-Rosenzweig emphasized the exceptional nature of the current drought and fire situation in the Western region, surpassing historical records dating back thousands of years. Climate projections indicate that drought conditions will further intensify in the future. Consequently, the significance of having enhanced tools that can provide more accurate information for effective management is becoming increasingly crucial. In light of these challenges, the development and utilization of advanced forecasting techniques are imperative for informed decision-making and proactive measures to mitigate the impacts of prolonged drought and increased fire risks.
RSS