Sequential sampling for panicle caterpillars (Lepidoptera: Noctuidae) in sorghum

Abstract

Panicle caterpillars comprise an economically important insect pest complex of sorghum throughout the Great Plains of the United States, particularly in Kansas, Oklahoma, and Texas. The sorghum panicle caterpillar complex consists of larvae of two polyphagous lepidopteran species: the corn earworm, Helicoverpa zea (Boddie), and fall armyworm, Spodoptera frugiperda (J.E. Smith) (Lepidoptera: Noctuidae). Sampling for panicle caterpillars in sorghum fields is usually accomplished by the beat bucket sampling technique with a fixed sample size of 30 beat bucket samples of one sorghum panicle each per 16.2 ha of field. We used Wald's sequential probability ratio test for a negative binomial distribution to develop a sequential sampling plan for panicle caterpillars. In total, 115 sorghum fields were sampled in Kansas, Oklahoma, and Texas from June to August 2010. Panicle caterpillars had an aggregated distribution of counts confirmed by Pearson's chi-square statistic for lack of fit to the negative binomial distribution for each sampled field. A sequential sampling plan was developed using a high threshold (an economic threshold) of 0.5 caterpillars per sorghum panicle, a low threshold (a safe level) of 0.20 caterpillars per panicle, and fixed error rates (α = 0.10 and β = 0.05). At caterpillar densities >0.45 and <0.12 per panicle, the average number of panicles inspected to make a decision was less than the current recommendation of 30. In a 2013 validation test of 25 fields, the expected number of samples taken from average sample number curve was in close agreement with the number of samples required using the sequential plan (r 2 = 0.93), and all fields were correctly classified when compared with a fixed sample size result. The plan improved upon current sampling recommendations for panicle caterpillars in sorghum because at known acceptable fixed error rates fewer samples were required when caterpillars are scarce or abundant, whereas more samples were required to make decisions with the same acceptable error rates when densities were near the economic thresholds.