Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Chine → 2021 → Evaluation of the Applicability of Machine Learning Models for Reference Crop Evapotranspiration

Kunming University of Science and Technology (昆明理工大学) 2021

Evaluation of the Applicability of Machine Learning Models for Reference Crop Evapotranspiration

董建华

Titre : Evaluation of the Applicability of Machine Learning Models for Reference Crop Evapotranspiration

Auteur : 董建华

Grade : Master 2021

Université : Kunming University of Science and Technology (昆明理工大学)

Résumé partiel
Reference crop evapotranspiration(ET0)is an important indicator to characterize the evapotranspiration capacity of the atmosphere and is an important component of crop water demand research.Accurate forecasting and estimation of ET0 allows for real-time monitoring of ET0,reducing the risk of crop exposure to disasters and making timely regulation.Clarifying the trend of ET0 under different climatic environments and meteorological data deficiencies can provide an important theoretical basis for accurately predicting the future trend of ET0,which is of guiding significance for crop planting structure adjustment,zoning and layout,thus realizing the purpose of sustainable utilization of regional agricultural water resources.In this paper,the results of the Penman-Monteith(P-M)formula calculation recommended by the Food and Agriculture Organization of the United Nations(FAO)are used as the ET0 standard values to assess the accuracy of forecasting and estimating ET0 from various perspectives.1)The applicability of estimating monthly ET0 under two input patterns(i.e.,local data only as inputs or local data in combination with cross-station data as inputs)is evaluated using data such as ET0 from multiple cross-stations combined with local meteorological data to form different parameter combinations in the input extreme gradient boosting(XGBoost)and support vector machine(SVM)models.2)This paper uses different parameter combinations to input four hybrid models(including the kernel-based nonlinear extension of Arps decline(KNEA)with grasshopper optimization algorithm(GOA-KNEA),KNEA with grey wolf optimizer algorithm(GWO-KNEA),KNEA with particle swarm optimization algorithm(PSO-KNEA),and KNEA with salp swarm algorithm(SSA-KNEA))to evaluate the performance of the hybrid model and the impact of different climatic conditions on the model degree.3)The meteorological data of five stations in northwest China were selected,and the five meteorological factors and ET0 were forecasted one by one in the forecasting period of 1~16 d using the Equi Distant-CDF-matching method(EDCDFm),XGBoost and Light GBM models,and the ET0 values calculated by the P-M model were compared to analyze the effects of different meteorological factors and different The effects of different meteorological factors and different seasons on the forecast ET0 were analyzed throughout the forecasting period

Mots clés : reference crop evapotranspiration ;estimation;forecasting ;FAO Penman-Monteith formula ;climate zones ;machine learning models ;

Présentation (CNKI)

Page publiée le 22 février 2022