Download PDF
Version 3 Metadata
The new Composite Drought Indicators were developed by using machine learning to compare the U.S. Drought Monitor data over 12 years to various drought indicators and indices.
- The input “mixes” were developed by focusing the ML on the relative accuracies of the different input aggregate timeframes (e.g., SPI-1 vs. SPI-2 vs. SPI-3).
- The transition between “standard” and “winter” scenarios was determined by comparing the accuracies of the vegetation, potential evapotranspiration, and surface water equivalent inputs against the mean 1-month temperature values.
- The aridity-vegetation factor is then applied by the following steps:
- Each week, a short-term factor using 3-month aggregates is calculated, along with a mid-term factor using 6-month aggregates and a long-term factor using 12-month aggregates.
- 5-year means are generated using the same week of the year
- Three different factor classifications are defined as: “arid” if factor < 0.2, “sub-humid” if 0.2 < factor < 0.4, and “humid” if factor > 0.4. Each input mix has three weight sets based on the aforementioned classifications
- Since the aridity-vegetation factors are computed every week, they can adapt both to seasonality and climate non-stationarity.
An improved fit-curve for converting the USDM categories to estimated percentiles was used in the ML process. This resulted in more accurate correlation scores of the indices with the USDM for weight generation.
- The transition between “standard” and “winter” scenarios was determined by comparing the accuracies of the vegetation, potential evapotranspiration, and surface water equivalent inputs against the mean 1-month temperature values.
- The inputs to test were chosen to cover the drought dynamics of water supply and storage, evaporative demand, and vegetation health.
Ranking of the inputs and indicators was determined using the Gaussian percentiles associated with a standardized distribution, as opposed to a manual calculation based on the minimum/maximum values in the history. However, unlike V2, in V3, no scaling is performed to align intervals closer to the rankings from V1. Finally, because the CDIs are calculated on a per-cell basis at 4km grid spacing, a minor smoothing algorithm is applied to make localized D4/W4 values more apparent.
Flash Drought Conditions Indicator
The Flash Drought Conditions Indicator (FDCI) is an automated assessment of the rate at which conditions are changing toward drought or abnormal wetness. The FDCI is calculated using a weighted average of four drought indicators, along with a weighted average of their strongest category changes. The “delta” values are determined using the greatest change between the valid date and the previous four weeks. The following indices and indicators are used along with their respective deltas:
- Standardized Precipitation Index (SPI) using 1-month aggregates
- Standardized Evapotranspiration (SPEI) using 1-month aggregates
- Standardized 10cm Soil Moisture anomalies using 1-week aggregates
- Standardized 40cm Soil Moisture anomalies using 1-week aggregates
Composite Drought Indicators (CDIs)
The three CDIs (short-term, mid-term, and long-term) are automated assessments of drought and abnormal wetness conditions. The CDIs are calculated using weighted averages of indices and indicators for both “standard” and “winter” scenarios. The scenario is determined per cell using the 1-month mean temperature.
NDMC Short-Term Composite Drought Index
- Standardized Precipitation Index (SPI) short-term mix comprised of 1-3 months. PRISM SPI (calculated using the Pearson-3 distribution) are now also included.
- Standardized Precipitation-Evapotranspiration Index (SPEI) short-term mix comprised of 1-3 months.
- Standardized Potential Evapotranspiration (PET) short-term mix comprised of 1-3 months.
- NOAH Soil Moisture short-term mix is comprised of 40 cm values out to 1 month and 100 cm values out to 2 weeks.
- Vegetation short-term mix is comprised of Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) values out to 2 weeks.
- Standardized Snow-Water-Equivalent (SWE) short-term mix comprised of 1-3 months.
NDMC Mid-Term Composite Drought Index
- Standardized Precipitation Index (SPI) short-term mix comprised of 4-6 months. 6-month PRISM SPI (calculated using the Pearson-3 distribution) is now also included.
- Standardized Precipitation-Evapotranspiration Index (SPEI) short-term mix comprised of 4-6 months.
- NOAH Soil Moisture short-term mix is comprised of 100 cm and 200 cm values out to 2 months.
- Vegetation short-term mix is comprised of Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) values out to 2 weeks
- Standardized Snow-Water-Equivalent (SWE) short-term mix comprised of 4-6 months.
NDMC Long-Term Composite Drought Index
- Standardized Precipitation Index (SPI) short-term mix comprised of 7-24 months. PRISM SPI (calculated using the Pearson-3 distribution) at time scales of 9, 12, 15, 18 and 24 months is now also included.
- Standardized Precipitation-Evapotranspiration Index (SPEI) short-term mix comprised of 7-24 months.
- NOAH Soil Moisture short-term mix is comprised of 200 cm values out to 12 months.
- Vegetation short-term mix is comprised of Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) values out to 3 months.
- Standardized Snow-Water-Equivalent (SWE) short-term mix comprised of 7-24 months.
Notes
- “Winter” CDI scenarios occur when the 1-month mean temperature is less than 2.25C.
- Standardized SWE is included to account for above-ground water storage that is not always captured by traditional precipitation or soil moisture indices.
- Potential Evapotranspiration (PET) used in the new composite drought indicators is based on the FAO-56 Penman-Monteith equation (alfalfa reference) provided by the University of California, Merced.
- To account for climate non-stationarity, the NDMC developed the short-term indices based on a 40-year period of record (POR) with a 30-year reference mean.
- To account for climate non-stationarity, the NDMC developed the mid-term indices based on a 50-year period of record (POR) with a 30-year reference mean.
- To account for climate non-stationarity, the NDMC developed the long-term indices based on a 60-yr period of record POR and 30-year mean reference.
- The grid size for the CDI products is 1/25° x 1/25° (4km x 4km).
Sources
- Precipitation and temperature data are from the Applied Climate Information System (ACIS) PRISM model. Precipitation and temperature data from Oregon State University’s PRISM model
- PET/SWE values are from the University of California, Merced
- Soil moisture is from the North American Land Data Assimilation System (NLDAS) NOAH model.
- NDVI/EVI are from EROS Visible Infrared Imaging Radiometer Suite (EVIIRS).