On Oct. 2, the NDMC published a new experimental CDI as part of a cooperative agreement with the U.S. Department of Agriculture’s Office of the Chief Economist.By Emily Case-Buskirk, Communications Specialist
A new iteration of the “objective blend” composite drought indicators (CDI) uses machine learning techniques and new metadata inputs to help authors make decisions about the U.S. Drought Monitor (USDM).
On Oct. 2, the National Drought Mitigation Center (NDMC) published a new experimental CDI. This builds upon objective blend tools that have been developed and refined over the years. This effort is funded through a cooperative agreement with the U.S. Department of Agriculture’s Office of the Chief Economist.
The objective blends are a useful tool for drought monitor authors and assessment teams working on weekly timeframes to produce updates to the USDM every week, said Brian Fuchs, NDMC climatologist. Through the updated CDI, USDM authors and others making drought-related decisions will have a new lens to identify rapidly-changing situations like flash drought.
“For USDM authors, one of our biggest challenges is determining short-term and long-term drought. It really helped when we started using objective blends,” he said. “Now, bringing in machine learning will improve the blends map, especially in classifying short-term and long-term drought.”
Through the NDMC Composite Drought Indicators website, users can access the NDMC blends maps, archive, data, inputs and metadata. A legacy section shows the objective blends as they were previously calculated.
The weekly maps are created using a combination of the Standardized Precipitation Index, Standardized Precipitation Evapotranspiration Index (SPEI) and soil moisture data. Datasets for the blends are computed every week using weighted sums of the ranked inputs. The results are then ranked to provide normalized 0.0-1.0 values for comparison.
Inputs added to the new CDI include the Standardized Net Moisture Balance Index — the NDMC’s experimental expansion of the SPEI that incorporates snow water equivalent data. The short-term blend also includes the Normalized-Difference Vegetation Index and Enhanced Vegetation Index.
The machine learning component uses a semi-guided iterative process involving two methods of supervised scoring and multi-pass linear regression. Historical drought monitor data was used to train the model.
In 2000, USDM authors started using a weighted model blending several sets of climate division data together to create the short- and long-term objective blends. The National Oceanic and Atmospheric Administration’s Climate Prediction Center developed the product and produces the blends every Monday.
The NDMC published an updated gridded dataset system in 2022 to provide a more detailed picture of drought conditions in the U.S.
Adjustments may be made to the new CDIs to best serve the drought community. To share questions or feedback, email DroughtMonitor@unl.edu.