This post is written by Greg Husak and Juliet Way-Henthorne, Climate Hazards Center at the University of California, Santa Barbara.

The Climate Hazards Center (CHC) at the University of California, Santa Barbara works with USAID’s Family Early Warning Systems Network (FEWS NET) to monitor food insecurity across several countries in the developing world. The CHC contribution to FEWS NET includes components related to forecasting and monitoring the agroclimatic conditions in its primary regions of interest. Foremost among these efforts is the characterization of rainfall conditions.

To this end, the CHC created the Climate Hazards InfraRed Precipitation with Stations (CHIRPS) gridded rainfall product. CHIRPS estimates the rainfall every 5 days, with a spatial resolution of 0.05 degrees. While CHIRPS data are freely available, it requires some expertise to take the digital data and put them in a graphical form that allows a user to visualize the spatial characteristics, as well as some of the derived products that help characterize the agroclimatic conditions. To help users, the CHC has developed the CHC Early Estimates (EE), a set of products and accompanying graphics that assess conditions over a variety of time intervals and spatial domains, assisting in the interpretation of rainfall conditions.

The EE products blend the best-available data to produce up-to-date products. The foundational data input is the CHIRPS rainfall estimate. CHIRPS blends a long-term climatology, satellite-derived estimates, and the best-available rainfall station data from a variety of sources to produce a rainfall estimate that has been shown to be reasonably accurate across many spatial domains. The temporal resolution of CHIRPS is the pentad — a ~5-day interval that aligns with the months of the year. 

One of the shortcomings of CHIRPS is that it is available in the middle of the month following the period of observation. So, for example, CHIRPS estimates for January are available in the middle of February. Its availability is obviously problematic for monitoring near-real-time conditions; to counteract this limitation, the CHC developed CHIRPS-Prelim. This product uses a subset of the stations that go into CHIRPS, including only the Global Telecommunication System (GTS) stations, which are available in a timely and automated way, facilitating their inclusion in a quick product. CHIRPS-Prelim is used to bridge the temporal gap between the latest-available CHIRPS and the current pentad. These two products are used to get up-to-date estimates of rainfall conditions two days after the end of each pentad.

Finally, for those interested in a slightly forward-looking product, CHC combines CHIRPS and CHIRPS-Prelim with the CHIRPS-GEFS 3-pentad (~15-day) forecast. This product takes the Global Ensemble Forecast System (GEFS) precipitation forecast and performs quantile matching to convert the rainfall anomalies produced by that system into CHIRPS-compatible rainfall amounts. Blending the CHIRPS, CHIRPS-Prelim, and CHIRPS-GEFS forecast can give an estimate of what the conditions over a specified temporal window will look like 15 days in the future. 

The combination of CHIRPS and CHIRPS-prelim provides a best estimate of the conditions to-date — adding CHIRPS-GEFS (and the associated uncertainty of forecast products) to that can give a forward-looking estimate of the conditions. These products are blended to produce the EE, which have two flavors of products, the Recent Rainfall Monitor and the Seasonal Rainfall Monitor. 

The Recent Rainfall Monitor (RRM) analyzes conditions for the previous 1/2/6/12/18 pentads (roughly equivalent to 5/10/30/60/90 day accumulations) for nine different spatial windows. The longest four time periods are also combined with a 3-pentad forecast component. The Seasonal Rainfall Monitor (SRM) analyzes rainfall for specific agricultural growing seasons for nine unique spatial-temporal windows. The monitoring spans from the start of the temporal window to current conditions in order to analyze the progress of a particular agricultural growing season, rather than just a fixed number of days. This implies that certain monitoring windows will be out of season, and will not be monitored for parts of the year. The SRM also includes a forecast component up to 3 pentads into the future or the end of the temporal window — whichever is shorter.

Figure 1. Seasonal Rainfall Monitor anomalies covering 1-December 2019 through 29-February 2020. 
This graphic highlights the abnormally wet conditions over East Africa and the predominantly dry conditions across much of Zambia, Zimbabwe, and Mozambique. Impacts of the dry conditions at the heart of the southern Africa growing season will be realized during the current harvest in the region. 

Rainfall for the RRM and SRM are depicted in varying ways to convey the values and their significance to the user. The first variable is simply the total rainfall accumulation over the temporal interval. The second variable is the anomaly, or the current season’s difference from the average of the CHIRPS historical period (Figure 1). The third variable is the Percent of Average, which divides the current total by the historical average and multiplies that by 100. The fourth variable is the Standardized Precipitation Index (SPI) which presents the current value as a normalized score based on the historical distribution of points. Finally, the rank is shown if the current accumulation is in the driest/wettest three events of the historical record. 

Additionally, for the SRM product, the Seasonal Performance Probability (SPP) is available. The SPP provides the to-date accumulations with historical totals for the remainder of the season to identify the likelihood of rainfall being in the lower/middle/upper tercile of seasonal accumulations. This product can help determine when there is a lot of uncertainty remaining in the seasonal outcome or, conversely, when it is fairly certain that the seasonal rainfall total will be in a given tercile. For instance, rainfall that is 60 percent of average early in the season may just indicate a late start, with plenty of time to recover and have normal rainfall performance. However, that same 60 percent of average late in the season may indicate a failed rainy season, and associated reduced agricultural yields. In the former example, the 60 percent of average may only mean a 50 percent chance of “below-normal” seasonal performance (i.e., still half a chance of normal rains or better), whereas the latter may indicate an 80 percent chance of below-normal rains (i.e., little chance of recovery).

The CHC Early Estimates products have been available for nearly a year, and are proving to be very useful to the monitoring community. Because they present conditions in a variety of terms, it allows for a fuller characterization of the rainfall. While none of these products can individually tell the whole story of rainfall conditions, in combination, they can better color the tapestry of agroclimatic conditions, assisting in identifying both the spatial extent and severity of water shortages.

These products regularly appear in FEWS NET monitoring and reporting streams, strengthening FEWS NET’s ability to highlight regions facing water deficits by combining recent rainfall totals with a robust historical record. The graphics have also been highlighted in GEOGLAM Crop Monitor products, as well as other outlets focused on early warning and food security analysis. This critical information provides advance warning and can help mobilize and target humanitarian relief. The CHC hopes that the increased exposure of such products and the data they produce will assist all partners and agencies in assessing conditions and providing the best-possible characterization of rainfall.

To learn more about CHC Early Estimates, please visit our website, or contact Will Turner and Laura Harrison with comments or inquiries.

Acknowledgements: Support to create and maintain the CHC EE came from the United States Geological Survey (USGS) cooperative agreement #G14AC00042 and United States Agency for International Development (USAID) cooperative agreement #72DFFP19CA00001.