| Crop model has become an important tool for the research of water saving in modern agriculture.However,the internal structure of crop model is complex and the parameters inside are numerous,which cause the model shows some non-linear characteristics and potential uncertainties.These all hinder the further development of the crop model.The low efficiency of agricultural water use is still an important factor restricting the sustainable development of agriculture in the grain production area of China.The current research on saving irrigation water has gradually shifted from the field scale to the regional scale.Therefore,the research on the optimizing irrigation schedule at a regional scale and considering the spatial distribution of various environmental factors are significant for improving irrigation efficiency and water resources allocation on a large scale.In response to the above problems,this study used the AquaCrop model as a research tool,wheat and maize as the research crops and the field experiment data in the valley plain as basics,we first conduct a systematic sensitivity analysis and uncertainty analysis for the model;then on the basis of these analysis,the elite genetic algorithm was used to calibrate the crop parameters of wheat and maize in the model,and the calibration efficiency setting three different objective functions respectively in the algorithm were compared;then,the simulation performance of crop growth,water use and soil moisture were validated under different irrigation treatments;finally,we combined with the AquaCrop model,GIS,multi-objective optimization algorithm and efficiency coefficient methods to construct an optimization framework for winter wheat irrigation schedule in the valley plain at the regional scale.The following results were achieved:(1)The changing characteristic of sensitivity and uncertainty of the AquaCrop model under different water and fertilizer conditions were discoveredThe sensitivity analysis of parameters in the AquaCrop model showed that the sensitive parameters and sensitivity rankings of maize and wheat are significantly different,but the parameter sensitivity variation of the two crops show similar laws and characteristics:first,parameters sensitive to model time series output such as canopy cover variation and biomass variation have a significant impact on this type of output during each crop growth period;second,some parameters that are not sensitive or generally sensitive to the model's non-time series output such as yield and maximum biomass would be very sensitive to the time series output,for example,CGC and MCC parameters(for maize)and STBIO and CCS parameters(for wheat)are generally sensitive to the maximum biomass and become very sensitive to the biomass production of maize and wheat;third,the sensitivity results under different water and fertilizer conditions are different,and the differences are small;compared to the impact of a single crop parameter on the model output,the calibration should pay more attention to the impact of the interaction between the parameters on the model output.The uncertainty analysis of model output showed that the uncertainty of the model output increase with the increase of water and fertility stress level,and fertility stress has a greater effect on the uncertainty increment.The fertilizer factor in the AquaCrop model has greater impact on parameter sensitivity and model output.According to the final results we recommend to select the treatment with no or low environmental stress as the basic data for model calibration.The above results provide reference and guidance for future calibration and simplification of crop models.(2)The AquaCrop model was automatically calibrated by elite genetic algorithmBased on the analysis of sensitivity and uncertainty,we used elite genetic algorithm to calibrate the maize and wheat crop parameters of AquaCrop model and achieved satisfactory results.The results showed that the calibration speed of the genetic algorithm is significantly improved compared to the traditional manual trial and error method,and the calibration accuracy is slightly better than the manual trial and error method;and we also set three different objective functions in the algorithm(i.e.different error sources and weighting factors),the results showed that the calibration speed and accuracy is different when the same objective function calibrating the parameters of two different crops of maize and wheat,and different objective functions also have some differences when calibrating the parameters of same crop.The three objective functions have respective advantages in calibration speed and accuracy.Considering that accuracy is more important in model calibration,so the Obj2objective function(different weight factors)is recommended to use in the algorithm.The results also showed that the size of the calibration parameter range(search space)has a greater impact on the calibration speed than the setting of the objective function.(3)The adaptability of AquaCrop model was evaluatedThe crop parameters of wheat and maize which calibrated based on the elite genetic algorithm are validated.The results show that the calibrated parameters can predict the crop canopy cover development(R~2 and d are all greater than 0.92),above ground biomass production(R~2 and d are all greater than 0.93),yield and water use efficiency(RE are less than10%)under different water stress conditions.Compared with the above indicators,the model's simulation accuracy of soil moisture was declined,but it is still within the acceptable range;there are some differences in the simulation accuracy between the three different objective functions and between wheat and maize.This is mainly because these differences have been expressed in the model calibration results,and the calibration accuracy of the model affect the error performance of the model validation directly;similar to previous results,this study also showed that the simulation accuracy of canopy cover and biomass under different water conditions is different.The simulation accuracy will decrease with the increase of water stress level,and in the later growth stage of crop,the model slightly overestimates the soil moisture.These issues should be paid attention to in future model improvements and applications.(4)A multi-objective irrigation schedule optimization model for the Fenwei Plain was constructed at a regional scaleA framework composed of the AquaCrop model,GIS,multi-objective optimization algorithm and the exponential efficiency coefficients method was developed to optimize the irrigation schedule of winter wheat and summer maize.The optimization was analyzed under different typical hydrological years,which the division of different hydrological years was based on grid rainfall data.The results showed that the AquaCrop model was calibrated based on field experimental data can predict crop yield and water use under different water conditions in each subregion of the study region,and the meteorological data generated by interpolation can more accurately describe the spatial difference of meteorological elements in the study area.The final optimization results showed that the optimized irrigation schedule has achieved satisfactory results in improving crop yield,water use efficiency and irrigation economic benefits of the overall study area under different typical hydrological years.For example,compared with the current local irrigation schedule,the average crop yield increased by 1.6-5.1%,the average water use efficiency and irrigation water use efficiency increased by1.1-10.2%,and the total amount of irrigation water also has a reduction between 4.2-6.5%.The results also showed that the spatial distribution of soil types and crop parameters have a significant impact on the optimization results of the irrigation schedule.The optimization framework re-optimizes the distribution of irrigation water resources among the sub-regions by increasing or reducing the irrigation water consumption in each research sub-region,thereby improving the overall crop yield,water use efficiency and economic benefit of the study area.This study clarified the variation laws of crop model simulation,developed a new method of model calibration and applied crop model in the optimization of irrigation schedule,and provided a theoretical basis for efficient water use in regional agriculture production.Large uncertainty is a difficult issue in the application of crop models.This research had involved some uncertainty analysis,but it is not in-depth and systematic.This issue still needs further in-depth research in the future. |
PDF Full Text Request