NOAA GFDL - Geophysical Fluid Dynamics LaboratoryGFDL - Geophysical Fluid Dynamics Laboratory

Identifying Sources of Predictability for Atmospheric Dryness in the Southwestern United States

February 27th, 2026

Key Findings

  • Vapor pressure deficit (VPD) measures how dry the air is and is strongly linked to wildfire activity in the western United States.
  • Month-to-month changes in VPD were predicted in several ways, including a physics-based Earth System model (SPEAR) and a data-driven model-analog forecasts.
  • Both methods show useful predictions for month-to-month dryness at short lead times. In many cases, the lower-cost data-driven method often performs comparably to the SPEAR model, highlighting the potential for accessible forecasting tools where computing resources are limited.

Jiale Lou, Youngji Joh, Thomas L. Delworth, and Liwei Jia. npj Climate and Atmospheric Science DOI: 10.1038/s41612-025-01028-6

Fire seasons in the western United States are becoming longer, drier, and more intense. As rising temperatures and lower humidity are increasing the atmosphere’s evaporative demand, seasonal forecasts of atmospheric dryness are gaining importance for wildfire preparedness. Vapor pressure deficit (VPD), a measure based on temperature and humidity, serves as a useful indicator of fuel drying and fire-weather risk.

The authors evaluate monthly VPD forecasts over the U.S. Southwest using two complementary approaches. The first uses NOAA GFDL’s Seamless System for Prediction and Earth System Research (SPEAR), a physics-based, fully coupled Earth system model. The second applies a data-driven analog method that generates forecasts by identifying similar past conditions from archived model output and forming a weighted combination of those initial states. Despite differences in complexity and computational cost, both approaches demonstrate useful forecast skill at short lead times. In many cases, the data-driven method performs comparably to SPEAR, offering a viable and efficient alternative for operational contexts with limited resources.

Timely VPD outlooks can help local and regional decision-makers anticipate elevated fire risk and take early action—from staffing and fuel management to emergency planning. Together, these tools offer flexible forecasting strategies tailored to diverse user needs. This work supports the delivery of timely, actionable information to help protect against weather and environmental hazards, such as wildfires, in a changing planet.

Forecast Skill of Monthly Vapor Pressure Deficit Anomalies in the U.S. Southwest
Forecast skill of monthly vapor pressure deficit anomalies in the U.S. Southwest
Forecast skill of monthly vapor pressure deficit (VPD) anomalies across different methods. The black line shows results from the Seamless System for Prediction and Earth System Research (SPEAR) model, with shading indicating 5–95% confidence intervals. The weighted model-analog method (red) is trained by identifying initialization-dependent analog weights that maximize forecast skill at each lead time over a training period, and by applying the averaged optimal weights to an independent validation period that overlaps with the SPEAR seasonal forecasting system. For comparison, unweighted analogs from SPEAR preindustrial control runs (pink, open diamonds) and a multi-model ensemble of CMIP6 simulations (blue, open triangles) are also shown. Both the physics-based and data-driven approaches demonstrate skill, particularly at shorter lead times.