Short Range Forecast Verification Program

This handy reference introduces the subject of forecastverification and provides a review of the basic concepts,discussing different types of data that may be forecast. Each chapter covers a different type of predicted quantity(predictand), then looks at some of the relationships betweeneconomic value and skill scores, before moving on to review the keyconcepts and summarise aspects of forecast verification thatreceive the most attention in other disciplines. The book concludes with a discussion on the most importanttopics in the field that are the subject of current research orthat would benefit from future research. An easy to read guide of current techniques with real life casestudies An up-to-date and practical introduction to the differenttechniques and an examination of their strengths andweaknesses Practical advice given by some of the world?s leadingforecasting experts Case studies and illustrations of actual verification and itsinterpretation Comprehensive glossary and consistent statistical andmathematical definition of commonly used terms

This report describes the development and testing of an objective technique to forecast cloudiness and precipitation through extrapolation of satellite imagery. By utilizing on objectively determined cloud-motion vector, the technique makes local forecasts of satellite parameters (brightness and IR temperature), with high temporal resolution, using simple linear extrapolation. Algorithms are then used to convert the satellite parameters to total cloud cover, probability of 1-hour precipitation, and presence of low, middle, and high clouds. The test program computed motion vectors and made forecasts out to 7 hours, in half-hour steps, at 30 locations. The program was tested on 12 spring and fall cases, using half-hourly GOES imagery. For periods beyond 2 hours, forecasts of cloud cover and precipitation were markedly better than persistence, which deficiencies in specification hindered short-period performance. Forecasts of cloud layers were worse than persistence due to inadequate specification algorithms. The net results were quite encouraging, and further refinements and developments are planned.