Previous View
APSnet Home
Phytopathology Home


Ecology and Epidemiology

Spatio-Temporal Analysis of Epidemic Development of Leather Rot of Strawberry. K. M. Reynolds, Postdoctoral research associate, The Ohio State University (OSU) and Ohio Agricultural Research and Development Center (OARDC), Wooster 44691; L. V. Madden, and M. A. Ellis. Associate professors, The Ohio State University (OSU) and Ohio Agricultural Research and Development Center (OARDC), Wooster 44691. Phytopathology 78:246-252. Accepted for publication 13 July 1987. Copyright 1988 The American Phytopathological Society. DOI: 10.1094/Phyto-78-246.

The temporal and spatial patterns of strawberry leather rot, caused by Phytophthora cactorum, were monitored in three field plots near Wooster, OH, from 15 May to 15 June 1986. Each plot was 2 m in length, three rows wide, and partitioned into 60 quadrats (10 20-cm-long quadrats on each side of a crop row). Straw mulch was removed from the two interior aisles, but left on the two exterior sides. Plots were infested on 15 May with strawberry fruit on which P. cactorum was sporulating. Assessments of disease incidence in each quadrat were made by counting the number of cymes bearing at least one infected fruit. By 8 June, disease incidence was > 60% in all but two of the interior sides but < 10% in the exterior sides. Data were analyzed using the spatio-temporal (ST) autocorrelation analysis program, STAUTO, to identify the appropriate temporal and spatial lag orders for specification of model parameters in ST autoregressive moving average (STARMA) models. When spatial autocorrelations and partial spatial autocorrelations of disease incidence between quadrats were calculated for each plot over the 10 rain events, first-, second-, and third-order spatial autocorrelations exhibited clear positive trends in each plot, indicating that disease incidence within neighboring quadrats at a given time was becoming progressively similar as the epidemic developed. No such trends were apparent in the partial autocorrelations. The first-order spatial partial autocorrelation was significant in all three plots for the last six rain events, whereas higher-order partial autocorrelations never attained significance in any plot, indicating a lack of spatial dependence between disease incidence in quadrats beyond the first spatial lag at any given time. Analyses of separate across- and within row-side ST autocorrelations of disease incidence demonstrated that a strong barrier effect was operating across crop rows. The highest levels of ST autocorrelation for all three plots were observed using the rook’s definition of spatial proximity and a crop row barrier specification that was obtained by trial and error. Initial model identification was based on the use of temporally differenced data, since analysis of the raw data and ST autocorrelograms indicated that the data were nonstationary over time. Interpretation of ST autocorrelograms and partial ST autocorrelograms suggested that the ST transfer functions that generated the observed patterns of fruit infection in plots 3 and 6 were very similar, and that change in the logit of disease incidence within a quadrat could be predicted in terms of the mean change and a single disturbance (error) term. However, the ST transfer function in plot 2 was best represented by either a STIMA(1, I) or a STARIMA(1, 1, 1, 1) model. Differences in model forms required to represent the epidemic processes in the three plots appeared to be the result of differences in topographic and edaphic factors that affected splash dispersal of the pathogen.