Verification of forecasts given to ciare project intervention areas

Downscaled forecasts or outlooks at timescale of 10 daily, monthly and seasonal have been prepared and provided to the CAIRE project areas since the third dekad of October 2016. In forecasting weather and climate, testing the value and quality of the forecasts must be taken as part of the whole forecasting process. Unless we test our forecast quality and measure where we stand we cannot see our progress or failures in forecast generation. So, verifying the issued forecasts and outlooks taken as one of the activities of the CIARE project. This study is done to test the performance of issued forecasts and outlooks from October 2016 third dekad to August 2017 first dekad for CIARE project areas. Methods for dichotomous (yes/no) forecasts and methods for forecasts of continuous variables used to test the performance of forecasts and outlooks given to the project areas. As we noticed, there are no meteorological stations in some of the CIARE interventions Woredas. For such Woredas, we cannot find normal meteorological data even for a single location in the Woreda. This means that our climatological knowledge of the dekal and monthly normal values can only be deduced from blended data or purely interpolation. Presenting the forecast for such Woredas in terms of normal rain could be very difficult for the assigned forecaster or meteorologist. In addition, if the season is not a rainfall season and dominantly sunny and dry, it does not matter if the forecast is presented in the form of rain, no rain or light to moderates rain. But for other Woredas where we have station and can get normal values, the dekadal and monthly forecast can be presented in relation to normal values. We can issue a forecast that clearly states whether the expected rainfall is above, near or below normal. The writer of this report finds very few forecasts of this kind. If possible, it would have value for the user as well as for one who verifies the forecast if the forecasts are presented in relation to normal values as it is done in the regular forecasts issued for the whole country in coarse resolution.  The NWP disseminated forecasts are from the outputs of NOAA. We have more localized weather and climate information than NOAA has. It would be much better if we were able to run the WRF model we have at home and issue ten daily forecasts. The period we took for verification is too short to arrive at concrete conclusion, but with the sample we used we have found out that, in relative terms, the accuracy of the dekadal worded forecast is much better than the worded monthly forecasts. Other statistics can be inferred from the summary tables given in the document. When we come to the NWP outputs, in all aspects, we do not get promising results.