Recalculation of precipitation trends in the long cycle region of Ethiopia: Sep 2003

Situation Report
Originally published

A recent report released by USGS/FEWS NET 'Estimating Meher Crop Production Using Rainfall in the 'Long Cycle' Region of Ethiopia' noted some significant trends in the 'Long Cycle' crop-growing region of Ethiopia (regions A and B in Fig 1). A recent visitor to EROS Data Center, Dr. Abdulkarim Hussein Seid of Addis Ababa University, presented a comparison of gauge data and the Collaborative Historical African Rainfall Model (CHARM) data with and without its orographic component (i.e., local effect due to mountains). From his work, it was apparent that the orographic enhancement component exaggerated the negative precipitation trends reported in that study. This brief document presents a recalculation of these trends using a dataset without these effects. The Climatologically Aided Interpolation (CAI) rainfall time series2 is based on the interpolation of carefully quality controlled station data3. The results shown here are for 1961-1996. We are currently working with Dr. Seid to extend our results to the present.

Fig 1. "Long cycle" crop growing region

One way of examining the strength of trends is to examine maps of the correlation of rainfall and time. In places where this correlation value is close to zero, trends are negligible. The observed correlation patterns, calculated over the 1961-1996 time period, are weak throughout most of the Belg crop growing regions for the months of March through May (Fig 2, left). The correlations can vary from -- 1 to 1, and indicate the relative strength of the rainfall change over time. Correlation values between -0.3 and 0.3 represent trends that explain less than 10% of the rainfall variability. Most of the March -May correlations fall within this range. For the months of June through September (Fig 2, right), however, portions of the western highlands exhibit fairly strong negative correlations (<-0.4) over the 1961-1996 period. This pattern suggests that the southwestern highlands have been experiencing a drying trend during the height of their rainy season. Spatial averages of the April-September rainfall for the 'long cycle' region likewise show that the western region (area A, Fig 1) appears to have a downward trend (Fig 3), while the eastern region (area B, Fig 1) does not.

Fig 2. Correlation [r] between seasonal CAI precipitation totals and time, calculated over the 1961-1996 period. Correlations are calculated at select stations with near continuous periods of record. Note that interpolated fields are weighted averages of multiple stations, and that more stations than those plotted here were used to derive the interpolated fields.

Fig 3. Time series of spatially averaged western and eastern April-Sept rainfall

Western and Eastern Apr-Sep Long Cycle CAI Presip

Combining the eastern and western time series and calculating the trend over the 1961-1996 period yields a value of -3.1 mm/year. This value is substantially smaller than that reported in our previous report (-23 mm/year). Using this new trend value to estimate a potential decline in yields produces a proportionately smaller value (-0.007 Quintals per hectare per year, versus -- 0.06 as originally reported). Assuming 8 million hectares are under production, the corresponding reduction in production per decade would be 57,000 Quintals, a value on the order of 10% of average annual food aid quantities, rather than a full year's food aid. Thus, while the trend is clearly unhelpful, especially in southwestern Ethiopia, it is not by itself sufficient to create large food deficits on a national scale. We regret any sense of alarm that the first trend calculation might have conveyed.


1 This document was prepared by Chris Funk, Gabriel Senay, Jim Verdin and Jim Rowland of the USGS/FEWS NET.

2 Willmott, C. J. and S. M. Robeson, 1995: Climatologically Aided Interpolation (CAI) of Terrestrial Air Temperature. International Journal of Climatology, 15, 221-229.

3 Vose and co-authors, 1992: The Global Historical Climatology Network: Long-term monthly temperature, precipitation, sea-level pressure, and station pressure data. Oak Ridge National Laboratory, ORNL/CDIAG-53, 26 pp.