ARIMA-SVR组合模型在基于标准化降水指数干旱预测中的应用

许德合; 张棋; 黄会平

许德合,张棋,黄会平.ARIMA-SVR组合模型在基于标准化降水指数干旱预测中的应用[J].干旱地区农业研究,2020,38(2):276~282

ARIMA-SVR组合模型在基于标准化降水指数干旱预测中的应用

Application of the combined ARIMA-SVR model in drought prediction based on the Standardized Precipitation Index

DOI：10.7606/j.issn.1000-7601.2020.02.38

中文关键词: 干旱预测标准化降水指数 ARIMA-SVR组合模型 ARIMA模型 SVR模型

英文关键词:drought prediction Standardized Precipitation Index (SPI) ARIMA-SVR combined model ARIMA SVR

基金项目:国家自然科学基金项目(51679089);河南省重点研发与推广专项（192102310257）

作者	单位
许德合	华北水利水电大学测绘与地理信息学院，河南郑州 450000
张棋	华北水利水电大学地球科学与工程学院，河南郑州 450000
黄会平	华北水利水电大学测绘与地理信息学院，河南郑州 450000

摘要点击次数: 829

全文下载次数: 676

中文摘要:

开展干旱预测是有效应对干旱风险的前提基础，本研究利用1951—2017年河南省郑州气象站点逐日降水量数据计算多尺度标准化降水指数（SPI），并建立了SPI序列自回归移动平均模型（ARIMA）和自回归移动平均与支持向量机回归组合模型（ARIMA-SVR），对模型参数进行率定和验证后，利用所建立的模型对河南省郑州气象站点多尺度SPI值进行预测。借助均方根误差（RMSE）、平均绝对百分比误差（MAPE）对回归预测模型的有效性进行判定。结果表明：ARIMA-SVR 组合模型在SPI1（1个月）和SPI12（12个月）的RMSE值分别为80.05和0.74，均低于ARIMA 模型的92.25和1.24，说明ARIMA-SVR组合模型与单一的ARIMA模型对SPI的预测精度都与该指数的时间尺度长短有关，都随时间尺度的增加而逐渐提高；SPI12的两种模型预测精度均高于SPI1、SPI3（3个月）和SPI6（6个月)的预测精度。用实测数据与模型的预测数据相比较说明ARIMA-SVR组合模型相比于单一ARIMA模型预测精度更高，且能够很好拟合不同时间尺度的标准化降水指数。

英文摘要:

Carrying out drought prediction is the premise basis for effectively coping with drought risk. The multi\|scale Standardized Precipitation Index (SPI) was calculated by using the daily precipitation data of Zhengzhou meteorological station in Henan Province from 1951 to 2017, and the SPI sequence autoregressive moving average model (ARIMA) and autoregressive moving average and support vector machine regression combined model (ARIMA-SVR) were established. After the model parameters were determined and verified, the multi\|scale SPI value of Zhengzhou meteorological station in Henan Province was predicted by using the established model. The validity of the regression prediction model was determined by means of the root mean square error（RMSE） and the mean absolute percentage error (MAPE). The results showed that RMSE values of ARIMA-SVR combined model in SPI1 (1 month) and SPI12 (12 months) were 80.05 and 0.74, respectively, which were lower than 92.25 and 1.24 of ARIMA model, indicating that both the prediction accuracies of SPI of the ARIMA-SVR combined model and the single ARIMA model were related to the time scale of the index, and they gradually increased with the increase of time scale. The prediction accuracy of the two models of SPI12 was higher than that of SPI1, SPI3 (3 months), and SPI6 (6 months). Comparing the measured data with the predicted data of the model showed that the ARIMA-SVR combined model had higher prediction accuracy than the single ARIMA model, and can well fit the standardized precipitation index at different time scales.

查看全文查看/发表评论下载PDF阅读器