基于贝叶斯模式平均与区间二型模糊集的分区灌溉方法  

Partition irrigation method based on BMA method and IT2FS theory

作　　者：邵东国[1] 邹亮峰 顾文权[1] 农翕智 王鹤 王柏 SHAO Dongguo;ZOU Liangfeng;GU Wenquan;NONG Xizhi;WANG He;WANG Bai(State Key Laboratory of Water Resources and Hydropower Engineering Science,Wuhan University,Wuhan 430072,China;Heilongjiang Province Hydraulic Research Institute,Harbin 150080,China)

机构地区：[1]武汉大学水资源与水电工程科学国家重点实验室,湖北武汉430072 [2]黑龙江省水利科学研究院,黑龙江哈尔滨150080

出　　处：《水利学报》2021年第9期1070-1081,共12页Journal of Hydraulic Engineering

基　　金：“十三五”国家重点研发计划项目(2016YFC0400101);国家自然科学基金重点项目(51439006)。

摘　　要：为解决松嫩平原水稻灌溉水需求不断增长与地下水资源可持续利用之间的矛盾,实现节水控采目标,本文以松嫩平原北部和平灌区水稻种植区为例,通过2017-2019年降雨-地下水-产量动态监测试验,揭示灌域尺度地下水时空变化规律;采用聚类分析与空间叠加分析提出了基于地下水埋深的分区方法,确定研究区的地下水埋深分区界限;根据地下水埋深分区结果,采用贝叶斯模式平均(Bayesian Model Averaging,BMA)方法,融合AquaCrop、Jensen模型,进行平、枯水年各分区灌溉情景的产量模拟,提高了地下水埋深分区条件下的产量模拟精度。引入风险偏好因子刻画决策者的风险态度,基于区间二型模糊集(IT2FS)理论对各地下水埋深区域的灌溉制度进行综合评价,利用有序加权平均算子(Ordered Weighted Averaging,OWA)集结综合评价值,提出了不同地下水埋深区域水稻种植的优化灌溉制度。结果表明:研究区地下水埋深分区界限为1.5 m;采用BMA方法有利于提高地下水埋深分区条件下的产量模拟精度,与实测产量相关性系数超过0.95;分区灌溉模式下各区产量-灌水量响应规律存在差异,但与现行灌溉制度相比,分区灌溉能减少稻田灌水量15%~37.6%,减少地下水开采量12%以上。The management of irrigation water in the Songnen Plain(SNP)is facing complex issues and challenges such as climate change,groundwater and wetland degradation.This study takes the Heping Irriga tion District in the SNP as an example.Through the rainfall-groundwater dynamic monitoring field experi ment from 2017 to 2019,the law of groundwater variations is revealed.Using the cluster analysis and spa tial overlay analysis methods,the boundary of zoning irrigation based on groundwater depth is proposed.The Bayesian Model Averaging(BMA)method is applied to integrate the production simulation series of dif ferent irrigation scenarios with the AquaCrop and Jensen models,which improves the accuracy of the yield simulation under the conditions of subregional groundwater depth.Based on the Interval Type-2 Fuzzy Set(IT2FS)theory,while the risk preference factors are introduced to describe the risk attitude of deci sion-makers,and the district irrigation model suitable for different decision-makers risk preferences is pro posed.The results show that the boundary of the groundwater depth in the study area is 1.5 m;the BMA method is beneficial to improve the accuracy of yield simulation,with the correlation coefficient between the simulated and measured yield exceeds 0.95.Additionally,compared with the current irrigation system,zoning irrigation can reduce the amount of irrigation by 15%-37.6%,and reduce the amount of groundwa ter extraction by more than 12%.This article provides a new basis and method for optimizing rice intensive production irrigation methods in shallow groundwater irrigation areas and protecting local river,lake and wet land ecosystems.

关 键 词：灌溉制度优化 地下水动态 分区灌溉 区间二型模糊集 贝叶斯模式平均 风险态度 

分 类 号：S274.4[农业科学—农业水土工程]