基于GM(1,1)-AR模型和AR模型的黑土区坡耕地土壤水分预测

Forecasting of Soil Water Content in Sloping Farmland of Black Soil Region Based on GM(1,1)-AR Model and AR Model

选用AR模型和基于小波分解的GM(1,1)-AR模型,模拟和预测北安市红星农场径流小区内鼠洞、鼠洞+垄向区田、鼠洞+暗管、鼠洞+暗管+垄向区田和无措施处理的坡耕地耕层土壤水分。结果表明,该2种模型预测精度均较高,其中基于小波分解的GM(1,1)-AR模型预测精度高于AR模型,基于小波分解的GM(1,1)-AR模型和AR模型的预测平均相对误差分别为3.18%和7.3%。

The AR model and GM（1,1）-AR model based on wavelet decomposing were applied to forecast the soil moisture of 5 runoff plots in Bei＇an Hongxing farm, including mole drain, the combinations of mole drain and subsurface pipe, mole drain and furrow damming, mole drain and subsurface pipe and furrow damming, conventional treatment. The results showed that： prediction accuracy of the two models were both higher, the accuracy of GM（1,1）-AR based on wavelet decomposition was higher compared to AR model, and the relative errors of GM（1,1）-AR model and AR model in average were 3.18% and 7.3%, respectively.