Beyond stationarity: characterizing drought in the context of changing baselines traditional drought assessment methods are based on the assumption that the distributions of surface water parameters…

The National Oceanic and Atmospheric Administration (NOAA) awarded a $500,000 Cooperative Agreement to Isciences LLC on July 1, 2026, under the Climate and Atmospheric Research program (CFDA 11.431) to develop novel statistical methods for characterizing drought under non-stationary conditions.

Isciences will develop and evaluate structural time series decomposition (STSD) methods to parse hydrologic time series and isolate long-term trends, changes in seasonality, and effects of major ocean-atmosphere oscillations. The work will generate dynamic normals—expected daily drought parameter values that account for non-stationarity in both long-term trends and seasonal patterns—against which current drought conditions can be compared. Isciences will evaluate multiple STSD approaches, develop principled hyperparameter selection methods, refine drought indicators based on STSD results, assess drought rarity using parametric and non-parametric methods, control for ocean-atmosphere oscillations, and identify spatial patterns. The research will begin with geographic focus on Vermont, expand to the Northeast DEWS region, and ultimately extend to the Continental U.S., starting with data from active GHCND stations.

Isciences will collaborate with state climatologists, drought coordinators, NIDIS program office staff, and Drought.gov throughout the project. Performance of work occurs in Burlington, Vermont. The award runs through June 30, 2028.

Beyond stationarity: characterizing drought in the context of changing baselines traditional drought assessment methods are based on the assumption that the distributions of surface water parameters are stationary over long periods of time. However, we can now clearly observe trends in surface water parameters that violate this assumption. Distortions due to erroneous assumptions of stationarity compromise the effectiveness of drought planning, mitigation, and adaptation efforts by a broad spectrum of government, business, and civil society stakeholders throughout the united states. This requires a fundamental rethink about how to characterize drought without assuming stationarity. Our proposed research will address this critical need by developing novel methods using structural time series decomposition (stsd) to parse hydrologic time series and isolate long-term trends, changes in seasonality, and the effects of major ocean-atmosphere oscillations. The results of these statistical models will be used to project dynamic normals - expected daily drought parameter values that diagnose and account for non-stationarity in both long-term trends and seasonal patterns and to which current drought conditions can be compared. In close collaboration with state climatologists, drought coordinators, nidis program office staff, and drought.GOV, we will: (1) evaluate multiple stsd methods for their ability to characterize drought during periods of change; (2) develop principled approaches for selecting stsd hyperparameters; (3) refine indicators based on stsd results that effectively characterize drought conditions; (4) assess drought rarity using both parametric and non-parametric methods; (5) control for ocean-atmosphere oscillations; and (6) identify and characterize spatial patterns. Our work will begin with a geographic focus on vermont, expand to the northeast dews region, and ultimately extend to CONUS. Similarly, we will start our research with data from active ghcnd stations that have the longest periods of record and then incrementally expand to stations with shorter periods of record, and then to other sources of station data for stream discharge and snow water equivalent. During the second project year, we will expand the work to analyze raster reanalysis data such as nclimgrid daily, NLDAS2, NWM, and ERA5. This phasing ensures that useful results are produced at each step along the way and that lessons learned with relatively straightforward data for specific geographic areas can be applied as the geographic and data complexity scopings expand over the course of the project. This research directly addresses nidis's goal of improving drought indicator performance to account for non-stationarity, enabling more accurate drought assessments that support communities in preparing for, mitigating, and responding to drought. Our results will advance NOAA's mission to build a climate-ready nation by developing methodological breakthroughs in drought assessment that will enhance drought early warning capabilities, improve understanding of drought severity and impacts, and deliver science-based information reflecting regional differences in drought conditions. Our CO-production approach with state climate offices ensures the research outcomes will be operationally viable and provide lasting benefits through improved integration of multiple data sources, temporal and spatial scales, and drought types within a coherent framework.