暴雨洪涝淹没灾情时空变化预测方法研究

Study on the Flood Disaster Spatio-temporal Variation Prediction Method

暴雨洪涝灾害一直是全球发生最频繁,影响范围最广的自然灾害之一,科学而全面地对暴雨洪涝灾害进行评价管理,具有十分重要的意义。针对暴雨洪涝淹没灾情科学预测的需求,以湖北省气象局提供的时间序列的暴雨洪涝淹没数据为基础,利用ARIMA模型(差分自回归滑动平均模型)对襄阳市2014-07-04 19:00和20:00的暴雨洪涝淹没灾情进行预测。诊断结果表明所建立的ARIMA(1,1,1)模型对襄阳市暴雨洪涝淹没数据可信度较好,对未来1 h和2 h暴雨洪涝淹没情况预测平均绝对误差分别是0.013 856和0.051 5,与实际监测值的相关性R的平方分别为0.769 9、0.458 9。该ARIMA(1,1,1)预测模型可以满足实际的预测需求。未来1 h的预测结果优于2 h预测,且未来1 h预测结果相关性较强,表明所建立的模型能够较好地应用于襄阳市未来1 h内的暴雨洪涝淹没灾情预测。

Prediction of flood disaster has very important guiding significance for disaster prevention and relief. Based on the time series rainstorm flood data provided by Hubei Provincial Meteorological Bureau, we applied the ARIMA model to predict the flood and flood disaster in Xiangyang at 19 and 20 July 4, 2014.The diagnostic results show that the ARIMA(1,1,1) model has a good reliability to the flood inundation data in Xiangyang. For the next hour and two hours, the average absolute errors of rainstorm flood are 0.013 856 and 0.051 5, and the correlations between R and the actual monitoring value are 0.769 9 and 0.458 9. The forecast results for the next hour are better than the two hour forecast and the forecast results are stronger in the next hour, which shows that the model can be better applied to the forecast of flood in Xiangyang in the next hour.