Validation of a photothermal phenology model for predicting dates of flowering and maturity in legume cover crops using field observations

Lack of information about the specific environmental adaptation of cover cr op species remains a serious constraint in the efficient design of agronomi c experiments examining options for more suitable and more sustainable mana gement of hillside farming systems in the tropics. Predictive models of cro p phenology for cover crop species exist but how robust these are for legum e cover crop species is largely unproven. We determined the predictive abil ity of phenological models, derived from glasshouse studies and driven by t emperature and photoperiod, across a diverse range of cover crop genotypes and tropical hillside environments. The models were designed to predict the duration from germination to first flowering, and from first flowering to first pod maturity. Seeds of eleven legume species of cover and/or green ma nure crops collected from different hillside locations world-wide were sown in two groups of nurseries (tropical short-day plants in early summer and sub-tropical long-day plants in early winter) at Kabale and Namulonge in Ug anda, Godavari and Lumle in Nepal, Cochabamba in Bolivia, Zamorano in Hondu ras, and Valenca in Brazil. Dates of sowing, first flowering and first pod maturity were taken and daily temperature data were recorded at each site. Similar observations for the same genotypes were available from independent experiments conducted at Islamabad, Pakistan, Hattiban, Nepal and at three locations in Cyprus. Model predictions wen compared with field observation s. The proportion of variation accounted for in the period from sowing to f irst pod maturity was 88% and 89% for the short-day and the long-day groups of genotypes, respectively. Likewise, the average difference from sowing t o pod maturity between the model predictions and the field observations was 6.3% and 7.9% for the combined short-day species and the combined long-day species, respectively. It is clear that the model predictions, for this da taset at least, an sufficiently robust to serve as a filter for determining the environmental suitability of germplasm.