SPATIAL AND TEMPORAL VARIABILITY OF TRANSPLANTED RICE AT THE FIELD-SCALE

Measuring rice (Oryza sativa L.) growth in farmers' fields requires pl ant sampling techniques that take account of spatial and temporal vari ability. Our objective was to specify sampling recommendations for tra nsplanted rice at the field scale. Rice was grown in two fields, ferti lized and not, at 20- by 20-cm hill spacings. Plant height, tiller num ber per hill, dry matter yield, and foliage area index were measured d uring the growth period at 8 locations and harvest samples were taken at 56 locations per field. Variation was low for plant height at all g rowth stages (CV = 6.1-9.5%); tiller number varied more (CV = 21-33%), especially in the tillering stage. Variation at harvest was intermedi ate (CV = 12-21%) for most crop variables. Grain yield ranged from 2.4 1 to 5.88 t ha(-1) for the unfertilized field and 4.54 to 8.82 t ha(-1 ) for the fertilized field. Transplanting errors introduced variation among hills. Variation between quadrats within a field was due to vary ing growth rates between tillering and panicle initiation (PD and was related to the spatial distribution of inherent soil nutrient availabi lity and fertilizer nutrient supply. Nonhomogeneous topdressed applica tion of N caused location-specific microvariation in crop stands. At a ll growth stages, two samples of three randomly selected hills were su fficient to measure plant height in a field. Increasing the number of samples had more effect on the precision of tiller number estimates th an did increasing the size of the sampling unit. To measure tiller num ber between tillering and PI of rice, sampling units of five hills sho uld be used and samples should be collected from 12 to 19 quadrats. At flowering, seven samples are enough. Harvest estimates of agronomic c haracters should be based on 8 to 12 quadrats when 10-hill sampling un its are used. A stratified random sampling design should be used, and samples should be taken >3 m apart.