自建模型间外推对森林可燃物含水率预测精度的影响

Effects of Extrapolation between Self-built Models on Forecast Model Accuracy of Moisture Contents

通过对2010年春、秋季大兴安岭地区盘古林场樟子松林、兴安落叶松林、白桦林林下细小可燃物含水率的连续观测并构建外推模型。结果表明:春季模型的外推效果好于秋季,模型的平均绝对误差降低了6.6%,平均相对误差降低了62.46%,白桦林的外推精度最高,破坏性取样的外推效果最好(平均相对误差349.83%);秋季樟子松林的外推精度最高,非破坏林荫下的外推精度最好(平均绝对误差较非破坏林空和破坏分别降低了13.7%和47.44%,平均相对误差分别降低了34.86%和83.30%)。模型外推虽不能减少误差,但有助于提高利用少量或仅有的几套含水率模型进行更大地区模型预测精度的工作加强关于模型参数和方程类型等的研究,以提高外推预测可燃物含水率的准确性。

The continuous observation was conducted on moisture contents of fine fuel under Pinus sylvestris, Xing＂ an Larch and birch forest of Pangu Forest Farm in Daxing＇ an Mountain during the spring and autumn of 2010, and the extrapolation model was constructed. The extrapolation effect of spring model is better than that of autumn, the average absolute error of model MAE （Mean, Absolute, Error） is decreased by 6.6%, and the average relative error MRE （Mean, Relative, Error） is decreased by 62.5%. The extrapolation accuracy of birch forest is the highest, and the extrapolation of destructive sampling is the best. The extrapolation accuracy of camphor forest in autumn is the highest, the extrapolation accuracy un- der non-destructive tree was the best, and MAE was 13.7% and 47.4% lower than that of non-destructive forest, respectively, and MRE was decreased by 34.9% and 83.3%, respectively. Model extrapolation does not reduce error, but it can help to improve the prediction accuracy of a larger area model using a few or only a few moisture models. In the future, the research on model parameters and equation types should be strengthened to improve the accuracy of extrapolation prediction .of fuel moisture content.