日尺度下的干旱区成龄枣树耗水量敏感因素建模分析
Modeling analysis on sensitive factors of water consumption for mature jujube tree under daily scale in arid area

以新疆阿克苏地区成龄枣树为研究对象，以日均气温(x1)、日均相对湿度(x2)、日均风速(x3)、日太阳辐射总量(x4)、日均大气压(x5)、0～100 cm土壤日均含水率(x6)及0～20 cm土壤日均温度(x7)为模型影响因子，采用偏最小二乘回归法建立了枣树耗水量预测模型，在此基础上运用缺省因子法分析了枣树耗水量对各因子的敏感性，并采用灰色关联分析法加以验证。结果表明：偏最小二乘回归模型(PLSR)具有较高的模拟精度（相关系数r=0.9789），不仅能够定量预测枣树耗水量（平均相对误差为6.40%），而且能够从机理上解释各因素对耗水量的影响；枣树耗水量对太阳辐射能量、土壤含水率和温度这3因素最为敏感（敏感性指数分别为3.24、2.18和2.09）；基于缺省因子法的枣树耗水敏感因素排序（x4＞x1＞x6＞x3＞x7＞x2＞x5）与灰色关联分析计算结果（x4＞x1＞x6＞x3＞x7＞x5＞x2）基本一致，尤其在主要影响因素的判别上是完全一致的。
Taking the mature jujube tree in Akesu, Xinjiang as the research object, the average daily temperature(x1), relative humidity(x2), wind speed(x3), solar radiation(x4), atmospheric pressure(x5), soil moisture content in 0～100 cm(x6) and soil temperature in 0～20 cm(x7) as the model influnce factors, has established the water consumption forecast model by using the PLSR method. On the basis of default factor method, has analyzed the sensitivity of water comsuption to each factor, furthermore using the gray relational analysis method to test and verify. The results showed that: The PLSR model had rather high accuracy(correlation coefficient r=0.9789), not only quantitative forecast the water consumption of jujube tree (average relative error was 6.4%), but also explain the impact of each factor on water consumption through mechanism. The three factors as solar radiation, soil moisture and temperature were the most sensitive factors on water consumption of jujube (sensitivity index were 3.24, 2.18 and 2.09, respectively). Based on the default factor method, has sequenced the sensitive factors as (x4＞x1＞x6＞x3＞x7＞x2＞x5) almost agreement with the calculated results by the grey relational analysis as(x4＞x1＞x6＞x3＞x7＞x5＞x2), especially the main influence factors were completely consistent.