Analysis of the association between the incidence of two spatially aggregated foliar diseases of strawberry

Association of the incidence of leaf blight (caused by Phomopsis obscurans) and leaf spot of strawberry (caused by Mycosphaerella fragariae) was asses sed at multiple scales in perennial plantings at several commercial farms o ver 3 years (1996 to 1998). For each field, the presence or absence of each disease was recorded from n = 15 leaflets in each of N approximate to 70 e venly spaced sampling units, and the proportion of leaflets with blight, sp ot, and total disease (blight or spot) was determined. Individual diseases and total disease incidence were all well described by the beta-binomial di stribution but not by the binomial distribution, indicating overdispersion of disease. The Jaccard similarity index was used to measure disease co-occ urrence at the leaflet, sampling-unit, and field scales. Standard errors of this index for the lower two scales were obtained using the jackknife (res ampling) procedure, and data randomizations were used to determine the expe cted Jaccard index for an independent arrangement of the two diseases, cond itioned on the incidence and spatial heterogeneity of the observed disease data. Results based on these statistics showed that only 4 of 52 data sets at the leaflet level and no data sets at the sampling-unit level had Jaccar d index values significantly different from that expected under an independ ent rearrangement of the two diseases. Rank correlation and cross-correlati on statistics were calculated to determine the degree of covariation in inc idence between the two diseases. Additionally, covariation between diseases was tested using a new procedure in the Spatial Analysis by Distance Indic Es (SADIE) class of tests. Covariation was detected in 21% of the data sets using rank correlation methods and in 15% of the data sets using the SADIE -based approach. The discrepancy between these two methods may be due to th e rank correlation procedure not taking into account the effects of spatial pattern of disease incidence. There was no relationship between mean disea se incidence per field of spot and blight or between degree of heterogeneit y of the two diseases las measured by a of the beta-binomial distribution), demonstrating lack of covariation at the field scale. Incidence of leaflet s with either disease (total disease incidence) could be well predicted usi ng a linear combination of the estimated probabilities of leaf blight and l eaf spot incidence based on independence of the two diseases. Heterogeneity of total disease incidence, measured with the estimated theta parameter of the beta-binomial distribution, could also be well predicted using a linea r combination of the weighted a values for leaf blight and leaf spot, with weights proportional to incidence of the individual diseases.