Validation of a site yield potential model for oil palms in Malaysia

The genetic yield potential of oil palms is the largest yield obtainable if all the environmental conditions, agronomic and management decisions were perfect. The actual yield obtained on a site however is usually lower and dependent on site, agronomic and management deficiencies which interact to limit the full expression of the genetic potential. If all constraints relating to incorrect agronomic and management inputs are removed, then the yield obtained would be the site yield potential (SYP). The ability to reliably predict SYP would be very useful for setting realistic yield targets and to focus attention on yield limiting factors that need to be alleviated. AAR has developed an empirical model for prediction of SYP of oil palm based on palm, soil, site and climatic characteristics. Preliminary evaluations of the model (ASYP 1) for mature palms on a range of soils had indicated satisfactory agreement. A revised model (ASYP 2.6) has been developed to predict SP of palms from planting to maturity. Two trials were set up to test the revised model with young oil palms on Munchong *Typic Hapludox) and with mature oil palms on Kawang (Typic Kanhapludult) soils. The model projected SYP of 19, 26, 32 and 36 t/ha/y for the first 4 years of harvesting on Munchong series. The corresponding best treatment yields obtained were 11.5, 25, 33 and 34 t/ha/y. Except for the year of harvesting, the differences between the best treatment yields and SYP were within 5 percent of SYP from year 4-6. SYP on Kawang series ranged from 30 to 3l t/ha/y for palms from 9-14 years old. Trial mean yield were however more variable, ranging from 26 to 3l t/ha/y or 83 to 102 percent of SYP. Differences between SYP and total yields over the duration of the trials were less than 8 percent of the SYP for the 2 sites. Although results indicated highly satisfactory agreement between predicted SYP and actual yields achieved for the soils, further validation work is needed with specifically designed trials for model improvement and over a wider range of environments