Estimating solar radiation for water-use and yield simulations under present and projected future climate using Cropsyst.
Abraha, Michael Ghebrekristos.
MetadataShow full item record
Agricultural scientists are faced with the challenge of producing enough food for the increasing world population. Hence the need to develop tools for managing soil and plant systems to increase food production in order to meet the world food demand in the future. Crop simulation models have become promising tools in predicting yield and related components fi'om a set of weather, soil, plant and management data inputs. This study describes the estimation of solar radiant density, a crucial input in crop simulation models; calibration and validation of a soil-plant growth simulator, CropSyst, for management purposes; and generation of weather data for assessment of crop production under possible climate changes in the future. Daily solar radiant density, an input required by most crop simulation models, is infiequently observed in many stations. This may prevent application of crop simulation models for specific locations. Long-term data records of daily minimum and maximum air temperatures, precipitation, sunshine hours and/or solar radiant density were obtained for Cedara, Durban, Seven Oaks and Ukulinga in KwaZulu-Natal, South Africa. Solar radiant density was estimated fi'om sunshine hours using the Angstrom equation and ten other models that involved daily minimum and maximum air temperatures and/or precipitation along with extratelTestrial radiant density. Coefficients for the Angstrom equation and one of the other ten models were specifically developed for South African conditions; the remaining models required fitting coefficients using the available data for all locations. The models were evaluated using (i) conventional statistics that involved, root mean square elTor (RMSE) along with its systematic and unsystematic components, slope, intercept, index of agreement (d), and coefficient of determination (R\ and (ii) a fuzzy expert system that involved a single modular indicator (Ira d) aggregated from the modules of accuracy (aggregation of the indices relative RMSE, model efficiency and I-student probability), con'elation (Pearson's correlation coefficient) and pattem (aggregation of pattem index vs day of year and pattem index vs minimum air temperature). For each index, two functions describing membership to the fuzzy subsets Favourable (F) and Unfavourable (V) were defined. The expelt system calculates the modules according to both the degree of membership and a set of decision rules. Solar radiant density estimated from sunshine hours for the Durban station resulted in R2 , RMSE (MJ m,2) and d index of 0.90, 2.32 and 0.97 respectively. In the absence of observed solar radiant density data, estimations from sunshine hours were used for derivation of coefficients as well as evaluation of the models. For Durban, the performance of the models was generally poor. For Cedara, Seven Oakes and Ukulinga two of the models resulted in a high d index and smallest systematic RMSE. The solar radiant density estimated from each model was also used as an input to simulate maize grain yields using the soil-plant growth simulator, CropSyst. The models were ranked according to their ability to simulate grain yields that match those obtained from using the observed solar radiant density. The rankings according to crop simulation, conventional statistics and expert system were compared. The CropSyst model was also evaluated for its ability to simulate crop water-use of fallow and cropped (oats, Italian ryegrass, rye and maize) plots at Cedara, KwaZulu-Natal, South Africa. Soil characteristics, initial soil water conditions, irrigation and weather data were inputted to CropSyst. Crop input parameters for oats, Italian ryegrass and rye were used, with little modifications, as determined from field experiments conducted at Kromdraai open cast mine, Mpumalanga province, South Africa. Crop input parameters for maIze were either determined fi'om field experiments or taken from CropSyst crop input parameters documentation and adjusted within a narrow specification range of values as dictated by CropSyst. The findings indicated that CropSyst was generally able to simulate reasonably well the water-use of fallow and cropped (oats, Italian ryegI°ass, rye and maize) plots; leaf area index and crop evapotranspiration of rye; and grain yield and developmental stages of maize. The validated CropSyst model was also used to simulate timing and amount of irrigation water, and investigate incipient water stress in oats, Italian ryegrass and rye. The CropSyst model was used to investigate potential effects of future climate changes on the productivity of maize grain yields at Cedara, KwaZulu-Natal, South Africa. The effect of planting date (local planting date, a fortnight earlier and a fortnight later) was also included in the study. A 30-year baseline weather data input series were generated by a stochastic weather generator, ClimGen, using 30 years of observed weather data (l971 to 2000). The generated weather data series was compared with the observed for its distributions of daily rainfall and wet and dry series, monthly total rainfall and its variances, daily and monthly mean and variance of precipitation, minimum and maximum air temperature, and solar radiant density. Four months of the year failed to reproduce distributions of wet and dry series, daily precipitation, and monthly variances of precipitation of the observed weather data series. In addition, Penman-Monteith reference evaporation (ETa) was calculated using the observed and generated data series. Cumulative probability function of ETa calculated using the generated weather data series followed the observed distribution well. Moreover, maize grain yields were simulated using the generated and observed weather data series with local, a fortnight earlier and a fortnight later planting dates. The mean simulated grain yields for the respective planting dates were not statistically different from each other; the grain yields simulated using the generated weather data had significantly smaller variance than the grain yields simulated using the observed weather data series. When the generated weather data series was used an input, the early planting date as compared to the locally practiced and late planting dates resulted in significantly greater simulated grain yields. The grain yields simulated using the observed weather data for the early and local planting dates were not statistically different from each other. The baseline period was modified by synthesized climate projections to create future climatic scenarios. The climate changes considered corresponded to doubling of [C02] from 350 to 700 ~t1 ,-I without air temperature and water regime changes, and doubling of [C02] accompanied by increases in mean air temperature and precipitation changes of 2 (lC and 10%, 2 (le and 20%>, 4 °c and 10%, and 4 (lC and 20% respectively. Solar radiant density was also estimated from daily air temperature range for all scenarios that involved change in mean air temperature. In addition, input crop parameters of radiation-use and biomass transpiration efficiencies were modified for maize, in CropSyst, to accommodate changes in elevated levels of [C02]. Equivalent doubling of [C02], without air temperature or water regime changes, resulted in increased simulated grain yields as compared to the baseline period. Adding 2 QC to the mean daily temperature and 10% to the daily precipitation of a [C02] elevated atmosphere reduced the grain yield but still kept it above the level of the baseline period grain yield. Adding 4 QC to the mean daily temperature and 10% to the daily precipitation fLllther decreased the yield. Increasing the daily precipitation by 20% instead of 10% did not change the simulated grain yield as compared to the 10% increments. Early planting date, for all scenarios, also resulted in higher yields, but the relative increment in grain yield was higher for the late planting dates with scenarios that involved increment in mean air temperature. In general, this study confi1l11ed that doubling of [C02] increases yield but the accompanied increase in mean air temperature reduces yield.