Analysis of the spatiotemporal structure of rice sheath blight epidemics in a farmer's field

The spatiotemporal structures of two rice sheath blight epidemics were stud ied during two consecutive (rainy and dry) rice cropping seasons in a farme r's field in the Philippines. The amount of primary inoculum in the soil me asured at transplanting was higher during the dry than during the rainy sea son, and showed nonaggregated spatial patterns in both seasons. Disease inc idence (percentage tillers infected per hill) was monitored at weekly inter vals on a (14 x 17) grid of rice hills 1 m apart, at the centre of adjacent , nonoverlapping elementary quadrats of 5 x 5 hills. The rate of disease in crease at the onset of the epidemics was higher in the dry than in the rain y season, but was afterwards higher in the rainy than in the dry season. A C(alpha) test was used to test for overdispersion, and the beta-binomial pa rameter theta was used to assess disease aggregation in quadrats consisting of one, two, four, eight and 16 elementary quadrats. Aggregation was detec table at all quadrat sizes throughout the epidemic, according to the C(alph a) test. theta was highest at epidemic onset and at the smallest (elementar y) quadrat size, indicating a very strong aggregation of disease at the ind ividual hill scale. theta increased until disease incidence reached 40%, th en declined over time. theta values were generally smaller in the dry seaso n. Semivariograms for both epidemics showed strong initial discontinuities, that is, large variance of disease incidence at the elementary quadrat sca le. Some aggregation in the spatial distribution of disease incidence was i ndicated at higher scales in the rainy season, coinciding with the highest rate of disease increase. Semivariograms did not suggest any structure in t he dry season, except for one date in the early epidemic stage. Spatiotempo ral autocorrelation of disease incidence at 45 days after transplanting yie lded very few variables that contributed significantly to describing the di sease incidence and its increase in a given hill. In the rainy season these variables were disease incidence (or increase) in the nearest sampled neig hbours, and disease incidence (or increase) observed 1 week earlier in the same hill. In the dry season only disease incidence (or increase) observed 1 week earlier in the same hill was significantly correlated with the curre nt incidence (or its increase). The development of a spatial structure in d istribution of disease is hypothesized to coincide with the spread of the p athogen from one hill to another, and with the shift from monocyclic to pre dominantly polycyclic processes in sheath blight epidemics. These results c oncur with previous reports in suggesting that management of rice sheath bl ight should first be directed against the leafborne, polycyclic phase of ep idemics, rather than their initial phase.