Research On Some Problems In Modeling Of Crop Physiological Parameters

Crop model is an important component of precision agriculture,digital agriculture and intelligent agriculture.Simulating the processes of crop growth and production,serving agricultural research and production practice in various forms combined with other information technologies,crop model has greatly promoted the development of agricultural information and agricultural modernization.In the process of crop modeling,it is difficult and therefore of great significance to choose the appropriate monitoring sampling interval,obtain the key influencing factor combination and build the multi output model.In this paper,we study the three key issues by analyzing photosynthetic physiology and environment data of various crops.The main contents and results include:（1）Effect of different monitoring and sampling intervals on crop model performance.The relationship between CO₂ exchange rate and transpiration rate of tomato leaves and main environmental factors was taken as the experimental object.Different sampling intervals were used in three classical model construction methods,namely regression,BP neural network and BP neural network optimized by GA.The results of modeling,prediction,and comparative studies show that the optimal sampling interval for simulating CO₂ exchange rate is 30 minutes,and for transpiration rate is 15 minutes.（2）Multi-factor analysis based on GEP function mining.The relationship between CO₂ exchange rate,transpiration rate and various environmental factors of tomato and cucumber was established via GEP function mining.The method was compared with correlation analysis and path analysis.The results showed that,the most suitable influencing factors of cucumber CO₂ exchange rate were air temperature,light intensity,water vapor pressure deficit and dew point,the most suitable influencing factors of cucumber transpiration rate were light intensity,air temperature,water vapor pressure and air flow rate,and the optimal factor combination for tomato CO₂ exchange rate and transpiration rate were light intensity,air temperature,relative humidity,water vapor pressure deficit and dew point.The results are consistent with the classical methods.（3）Multi output RBF network based on GEP optimization for crop physiological parameters modeling.The structure of GEP-optimized multi-output RBF network algorithm was designed.The CO₂ exchange rate and transpiration rate of the cucumber and tomato leaves were predicted.It was compared with RBF multi-output network and GA-RBF multi-output network.The results show that GEP-optimized multi-output RBF networks owns better prediction accuracy,and it has advantage over RBF and GA-RBF models on the balance of multi-output results.