Climate Hazards Center Early Estimates and the East Africa March-to-May 2019 Drought
This post was written by Laura Harrison, Juliet Way-Henthorne, and Chris Funk with the Climate Hazards Center.
The Climate Hazards Center (CHC) has been using the CHC Early Estimates for agroclimatological monitoring and forecast-plus-observation rainfall scenarios since late 2018. The CHC Early Estimates are regional maps that show rainfall conditions for multiple accumulation periods, which are useful for agroclimatic monitoring. They combine the latest-available CHIRPS precipitation estimates and unbiased 15-day rainfall forecasts based on NCEP Global Ensemble Forecast System (GEFS) weather forecasts, downscaled by the CHC. (See https://chc.ucsb.edu/monitoring and the end of this article to access CHC Early Estimates.)
In this article, we review several recent applications of CHC Early Estimates at FEWS NET and in the GEOGLAM Crop Monitor for Early Warning that supported early warning of the March-to-May 2019 drought in the eastern Horn of Africa.
The March-to-May 2019 drought was the latest in a series of East Africa droughts, and it followed a previously poor October-to-December 2018 Short Rains/Deyr season. A recent blog article from the CHC explains the historical context of this drought and contributing climate factors. A previous Agrilinks article describes some of the drought impacts seen during May 2019.
As discussed below, there was no advance pre-season warning of this drought from dynamic climate model forecasts. Early warning for this drought was made possible through close monitoring of two main information sources: rainfall estimates, which showed increasing season-to-date rainfall deficits starting in late February in some areas, and relatively short lead-time weather and climate model forecasts (~ one month to less than a month).
The strategy of combining observed final and preliminary data and short-to-medium term rainfall forecasts can yield valuable and actionable information during major cropping seasons. In 2019, this type of information helped to successfully identify and communicate early concern for a poor Long Rains/Belg/Gu spring season rainfall in the eastern Horn. FEWS NET and the Food Security and Nutrition Analysis Unit-Somalia (FSNAU) released alerts and multiple analyses to the donor community in April and May, which detailed the severe drought in the eastern Horn, the poor start of Somalia’s April-June 2019 Gu rainfall season, and the anticipated deterioration in food security conditions across the region.
(Those reports can be accessed here: FEWS NET Seasonal Monitor for April and May 2019 and FSNAU April 2019).
March-to-May 2019 drought: Dynamic model forecasts did not provide advance warning (before March)
International seasonal forecast systems did not provide a clear indication of upcoming poor rainfall performance in advance of this season. However, at least one of the one-month-ahead forecasts, beginning in early March, provided actionable information to support drought early warning.
Figure 1 (top panel) shows the forecasts for the most likely March-to-May 2019 rainfall category, according to the ECMWF/Met Office/Meteo-France/CMCC/DWD C3S multi-system seasonal forecast, the North American Multi-Model Ensemble (NMME) forecast, and the Greater Horn of Africa Regional Consensus Climate Outlook (GHACOF 51), all released in February. These collectively indicated a below-normal MAM rainfall only in limited areas of the eastern Horn of Africa, and the absence of a below-normal rainfall signal in key agricultural and pastoral areas of Kenya, southern Ethiopia, and southwestern Somalia. Figure 1 (bottom panel) shows two- and one-month ahead NMME ensemble mean rainfall anomaly for April 2019. The March 2019 initial conditions (forecast released in early March) indicated below-average April rainfall, whereas the February 2019 initial conditions did not.Figure 1. A retrospective look at forecasts for March to May and April 2019. The top panel shows February 2019 probabilistic seasonal forecasts for March-to-May rainfall from (left to right) the C3S multi-system season forecast, North American Multi-Model Ensemble (NMME), and the 51st Greater Horn of Africa Climate Outlook Forum. The bottom panel shows the February and March 2019 NMME forecasts for April rainfall anomaly. Of these forecasts, only the 1-month ahead NMME forecast for April (lower right) clearly indicated a cause for concern. Source: Laura Harrison and Gideon Galu (UCSB CHC/FEWS NET) and Andrew Hoell (NOAA ESRL/FEWS NET).
March-to-May 2019 drought: Early warning based on CHC Early Estimates and March NMME assumptions
In mid-March 2019, prompted by the rainfall deficits observed in the region since late February and the NMME forecast for below-average April rainfall, the CHC used the Early Estimates (a combination of CHIRPS Preliminary and CHIRPS-GEFS) and the North American Multi-Model Ensemble (NMME) rainfall forecast for April and May as inputs to a MAM 2019 rainfall scenario for the eastern Horn. This visual (Figure 2) proved to be valuable for communicating the current situation. In the end, it also turned out to be a generally accurate portrayal of March-to-May 2019 rainfall outcomes. Eastern Kenya, southeastern Ethiopia, and parts of southern Somalia ended up with substantially below-normal rainfall, with amounts even lower than shown in the scenario.
For scenario development and application, important caveats at the time were that it was still early in the season and that this scenario was mainly based upon forecast information, which is inherently uncertain. All components were selected from the best-available sources; for example CHIRPS Preliminary was used for March 1st-15th, and the CHIRPS-unbiased GEFS forecast (CHIRPS-GEFS) for March 16th-31st. These data sources are regularly used in CHC Early Estimates. The longer-term outlook was based on NMME. Specifically, for April rainfall, the lower limit of NMME’s April mean standardized anomaly for this area, -0.25, (NMME map showed -0.25 to -0.5) was used to select corresponding per-pixel historical CHIRPS values, and for May rainfall, corresponding to the NMME forecast for May, the CHIRPS per-pixel historical average was used. Overall, the scenario map using a combination of preliminary data and forecast sources effectively communicated early warning of the situation. It is important to recognize its inaccuracies as well. Spatial variability is muted: it underestimated rainfall in eastern Tanzania, and it underestimated the severity of deficits in some of the worst-affected areas. Such is the nature of early outlooks. The main source of underestimation was the assumption that April rains would be -0.25 standard deviations below average. April 2019 rainfall ended up being in the -0.5 to -1.5 standard deviation range for most of the eastern Horn.Figure 2. Rainfall scenario and outcome for March-to-May 2019. The panel on the left shows a most likely rainfall scenario for the MAM 2019 season in the eastern Horn that was used in discussions at the Famine Early Warning System Network (FEWS NET) in March 2019. This scenario was based on rainfall data and outlooks from CHIRPS preliminary, a next two-week CHIRPS-GEFS forecast, and the NMME next one-month and two-month forecasts. On the right, the actual outcome for MAM 2019 percent of average rainfall based on CHIRPS final. Source: Laura Harrison and Gideon Galu (UCSB CHC/FEWS NET) and Andrew Hoell (NOAA ESRL/FEWS NET).
March-to-May 2019 drought: GEOGLAM Crop Monitor for Early Warning
CHC Early Estimates are regularly used to identify and communicate information about areas of concern in GEOGLAM Crop Monitor for Early Warning (CM4EW) reports. Monthly CM4EW bulletins identify crop conditions in countries at risk for food insecurity. CM4EW is a valuable multi-source consensus resource based on collaborations between UN WFP, USAID FEWS NET, EC-JRC MARS, South Africa ARC, and Asia-RiCE. CM4EW is coordinated and published by the University of Maryland, with support from NASA Harvest.
Regional Climate Outlooks in the CM4EW summarize current short-to-medium range climate forecasts to inform decision-makers about potential transitions towards improving or worsening cropping conditions in the upcoming weeks to months. Figure 3 shows the CM4EW April 2019 Regional Outlook for East Africa. In this outlook, which was published in early April, the CHC Early Estimate for February 21st to March 31st was paired with the CFSv2 rainfall forecast for the month of April (forecast issued March 28th) to clearly show that recent conditions, plus the two-week GEFS forecast and the next-month CFSv2 forecast, were cause for concern.Figure 3. Regional Outlook from Crop Monitor for Early Warning April 2019 report. This outlook documented early concern for poor distribution and below-normal rainfall totals during East Africa’s important March to May 2019 rainfall season. Source: https://cropmonitor.org
Accessing CHC Early Estimates
Thanks to website improvements in June 2019, public users can easily access the latest CHC Early Estimate maps for a number of monitoring regions: East Africa, West Africa, Southern Africa, Central America and Caribbean, Central Asia, and Continental Africa. These maps show precipitation totals, anomalies, percent of average, standardized precipitation index (SPI), and rank (1st-3rd wettest/driest since 1981). These maps show CHIRPS Preliminary and/or Final 5-day, 10-day, 30-day, 60-day, and 180-day rainfall totals, as well as a corresponding outlook that is extended by 15 days using an unbiased GEFS forecast (CHIRPS-GEFS).
Maps showing previous time periods, and the data used to make them, can be downloaded from our ftp site. The archived data are an excellent resource for users who want to make custom maps with GeoTIFFs. See https://chc.ucsb.edu/monitoring for more information and the latest CHC Early Estimates.Figure 4. Snapshot of web interface for the CHC Early Estimates. Maps that show the latest estimates and forecast for rainfall accumulations, and the historical context of these, are easy to view. Archived maps and GeoTIFFs are available through ftp. Source: https://chc.ucsb.edu/monitoring/early-estimates