FACTORS CAUSING FIELD VARIATION OF DIRECT-SEEDED FLOODED RICE

In 1989 a study was conducted on the variation of soil and plant properties within a rice field of the Krasnodar region (Russia). Fairly stable soil properties (pH, soil texture, soil organic matter) had lower coefficients of variation (CV 10-16%) and were normally distributed. In contrast, soil properties that were strongly affected by management had CV's between 29 and 122%. The rice yield within the field ranged from 0.73 to 7.91 t DM ha-1. Factor analysis and multiple linear regression were used as principal techniques to explore the nature of variate interactions that may significantly affect crop-yield patterns and to classify and map the soil fertility status in the field. Five factors (soil fertility status, land preparation, nitrogen fertilizer application, seeding rate, and P-availability) were extracted. Regression models based on these factors explained 75% of the plant density and 56% of the rice yield variation. Soil pH, microrelief, weed growth, and the seeding rate were found to be the major yield factors for rice. A soil fertility map was produced using three fertility classes derived from the extracted soil fertility factor. Soil pH is proposed as a criterion for soil fertility mapping. Co-spatial soil and crop survey and application of multivariate statistics are suggested as suitable methods for exploration of complex soil-crop interactions.