Johnson Dennis A, Cummings Thomas F, Fox Alan D
Department of Plant Pathology, Washington State University, Pullman 99164-6430.
Fox Weather LLC, Fortuna, CA 95540.
Plant Dis. 2015 May;99(5):683-690. doi: 10.1094/PDIS-06-14-0602-RE.
Accuracy of prediction was analyzed for 17- and 30-day rain forecasts at two locations in the Columbia Basin to determine whether forecasts were sufficiently accurate to be included as a model component to schedule fungicide applications for potato late blight. Accuracy was partitioned into specificity (percentage of forecasted nonrainfall events classified correctly) and sensitivity (percentage of forecasted daily rainfall events classified correctly). An adjusted sensitivity, which included the forecasted rain day plus the next 2 days, was also used to give a wider target than only 1 day for evaluating accuracy of forecasted rain events. For 17-day forecasts, specificity during the seasonal test period was ≥70% from mid-June through September and specificity over the days of the forecast was >70% for the first 8 days at both locations both years. Adjusted sensitivity over days of the forecast was initially >80% and then decreased as forecasts increased from 7 to 17 days for 17-day forecasts at both locations and years. Sensitivity and adjusted sensitivity during the seasonal test period were both positively correlated with the number of rainy days while specificity was negatively correlated. Adjusted sensitivity was considerably higher for May (month with highest incidence of rain) than July (month with lowest incidence of rain) at both locations. For 30-day forecasts, specificity during the test period was >75% in July and August and adjusted sensitivity ranged from 60 to 100% for time periods occurring in May and June during both sample seasons. Specificity was generally above 80% as days of the forecast increased and adjusted sensitivity varied greatly over days of the forecasts, with extremes between 0 and 100% at both locations and years for the 30-day forecasts. Specificity of 17- and 30-day rain forecasts and adjusted sensitivity of 17-day rain forecasts have utility in scheduling late blight fungicides in the Columbia Basin.
分析了哥伦比亚盆地两个地点17天和30天降雨预报的预测准确性，以确定这些预报是否足够准确，可作为安排马铃薯晚疫病杀菌剂施用的模型组成部分。准确性被划分为特异性（正确分类的预测非降雨事件的百分比）和敏感性（正确分类的预测每日降雨事件的百分比）。还使用了一种调整后的敏感性，它包括预测降雨日加上接下来的2天，以给出比仅1天更宽的目标来评估预测降雨事件的准确性。对于17天的预报，在季节性测试期间，从6月中旬到9月，特异性≥70%，并且在这两年中，两个地点前8天的预报期间特异性>70%。在两个地点和年份，对于17天的预报，随着预报天数从7天增加到17天，预报期间的调整后敏感性最初>80%，然后下降。季节性测试期间的敏感性和调整后敏感性均与降雨天数呈正相关，而特异性呈负相关。在两个地点，5月（降雨发生率最高的月份）的调整后敏感性比7月（降雨发生率最低的月份）高得多。对于30天的预报，在7月和8月的测试期间，特异性>75%，并且在两个采样季节的5月和6月期间，调整后敏感性在60%至100%之间。随着预报天数增加，特异性通常高于80%，并且调整后敏感性在预报天数上变化很大，在两个地点和年份的30天预报中，极端值在0%至100%之间。17天和30天降雨预报的特异性以及17天降雨预报的调整后敏感性在哥伦比亚盆地安排晚疫病杀菌剂方面具有实用性。
