Variation of measured banana yields in a Costa Rican plantation as explained by soil survey and thematic mapper data

Variation in yield was studied on a 370 ha banana plantation in Costa Rica. Yield for 42, 6–12 ha large sections, were expressed in terms of: number of bunches/ha, weight/bunch, gross production/ha. Yield variation was explained by readily accessible soil and thematic mapper (TM) information, using three models. First, a TM image was used. A number of band combinations (vegetation indices) and the Near Infrared Band (TM4) were tested to characterize each section. The TM4 band and the “Greenness” band combination appeared to be the best estimators with 46% of the variation explained. Second, yield variation was explained on the basis of a detailed soil map that was analysed for each section. Of the variance in gross production, 67% could be explained by production per soil unit. Gross production of the units varied from 504 to 896 kg/ha/week. In the third model, the TM and soil data were combined. The explained part of the variation in a combined model remained 67% or was even lower.

The study indicates that the soil map of this area was a good estimator for banana yields in terms of gross production. Since the production figures were only available per section, i.e., for relatively large, rectangular areas, the averaged TM image for a section did not provide extra information. Site specific, stratified sampling per soil map unit is proposed to improve results of the type of models used in the study.