
EEMD-ARIMA在干旱预测中的应用
许德合, 丁严, 张棋, 黄会平
EEMD-ARIMA在干旱预测中的应用
Application of the EEMD-ARIMA Combined Model in Drought Prediction: A Case Study in Xinjiang Uygur Autonomous Region
近年来,国内干旱灾害频发,影响了正常的农业生产和经济发展,因此精确预测干旱发生具有重要意义。基于1960-2019年新疆维吾尔自治区气象站点的逐日降水量数据,计算了1、3、6、9、12及24个月时间尺度的标准化降水指数(SPI),利用差分自回归移动平均模型(ARIMA)和集合经验模态分解(EEMD)-ARIMA组合模型,分别对多尺度的SPI进行预测,并通过均方根误差(RMSE)、平均绝对误差(MAE)和决定系数(R 2)对预测结果进行评价。结果表明:EEMD-ARIMA组合模型的预测结果与新疆年鉴记录情况较为一致,能够用于对干旱进行预测;组合模型能够有效减少序列的非平稳性,相较单一模型能更好地预测SPI序列;EEMD-ARIMA组合模型在干旱预测中具有明显优势,在各时间尺度,组合模型预测精度均高于单一模型,能更准确地进行预测。
In the context of global warming, drought becomes more and more frequent, causing negative impacts on agricultural and social activities. Based on the daily precipitation data of meteorological stations from 1960 to 2019 in Xinjiang Uygur Autonomous Region, this paper calculates the Standard Precipitation Index (SPI) in a timeframe of 1, 3, 6, 9, 12, 24 months, then time series SPI at different temporal scales are predicted by ARIMA model and EEMD-ARIMA combined model. And the effectiveness of model is judged by the evaluation standard of RMSE, MAE, and R2 . The main conclusions are as follows: the forecast results of the EEMD-ARIMA combined model in Xinjiang are consistent with Xinjiang yearbook. Therefore, the combined model can be used in the prediction of drought. Compared with ARIMA model, EEMD-ARIMA combined model can effectively reduce the non-stationary of series and match the SPI series better. The prediction accuracy of EEMD-ARIMA combined model is higher than that of ARIMA model at each time scale. The combined model has significant advantages in drought prediction.
干旱预测 / ARIMA / EEMD-ARIMA组合模型 / SPI {{custom_keyword}} /
drought prediction / ARIMA model / EEMD-ARIMA combined model / SPI {{custom_keyword}} /
表1 示例气象站点信息Tab.1 Information of sample meteorological stations |
区站号 | 站名 | 经度/(°E) | 纬度/(°N) | 海拔/m |
---|---|---|---|---|
51068 | 福海 | 87.28 | 47.07 | 500.9 |
51542 | 巴音布鲁克 | 84.09 | 43.02 | 2458 |
51811 | 莎车 | 77.16 | 38.26 | 1 231.2 |
表2 标准化降水指数干旱分级Tab.2 Drought classification based on SPI |
等级 | 类型 | SPI范围 |
---|---|---|
1 | 无旱 | SPI>-0.5 |
2 | 轻旱 | -1.0<SPI≤-0.5 |
3 | 中旱 | -1.5<SPI≤-1.0 |
4 | 重旱 | -2.0<SPI≤-1.5 |
5 | 特旱 | SPI≤-2.0 |
表3 六尺度SPI序列的ARIMA模型定阶Tab.3 Order the ARIMA model based on six scales SPI values |
站点 | SPI序列 | p | d | q | AIC | BIC |
---|---|---|---|---|---|---|
福海 | SPI 1 | 1 | 0 | 1 | 2 015.450 | 2 033.761 |
SPI 3 | 1 | 0 | 3 | 1 395.748 | 1 423.198 | |
SPI 6 | 5 | 0 | 2 | 950.322 | 991.460 | |
SPI 9 | 8 | 0 | 3 | 615.929 | 675.296 | |
SPI 12 | 6 | 0 | 3 | 314.084 | 364.271 | |
SPI 24 | 1 | 0 | 1 | -162.364 | -144.182 | |
巴音布鲁克 | SPI 1 | 1 | 0 | 0 | 2 024.907 | 2 038.641 |
SPI 3 | 0 | 0 | 2 | 1 570.411 | 1 588.711 | |
SPI 6 | 3 | 0 | 2 | 1 382.345 | 1 414.341 | |
SPI 9 | 1 | 0 | 0 | 1 127.860 | 1 141.560 | |
SPI 12 | 1 | 0 | 0 | 697.394 | 711.081 | |
SPI 24 | 1 | 0 | 0 | 132.625 | 146.261 | |
莎车 | SPI 1 | 2 | 0 | 1 | 1 646.146 | 1 669.036 |
SPI 3 | 4 | 0 | 1 | 1 470.492 | 1 502.518 | |
SPI 6 | 5 | 0 | 0 | 1 226.602 | 1 258.599 | |
SPI 9 | 2 | 0 | 1 | 936.908 | 959.741 | |
SPI 12 | 8 | 0 | 1 | 442.679 | 492.866 | |
SPI 24 | 2 | 0 | 1 | -78.242 | -55.515 |
图4 基于ARIMA模型与EEMD-ARIMA组合模型对福海站多时间尺度SPI值的预测结果与观测值计算结果对比(2008-2019)Fig.4 Comparison of predicted and observed value of multi-time scale SPI of ARIMA model and EEMD-ARIMA combined model in Fuhai Station(2008-2019) |
图5 基于ARIMA模型与EEMD-ARIMA组合模型对巴音布鲁克站多时间尺度SPI值的预测结果与观测值计算结果对比(2008-2019)Fig.5 Comparison of predicted and observed value of multi-time scale SPI of ARIMA model and EEMD-ARIMA combined model in Bayinbuluke Station(2008-2019) |
表4 ARIMA模型和 EEMD-ARIMA组合模型的MAE、RMSE、R 2值Tab.4 MAE、RMSE and R 2 values for ARIMA and EEMD-ARIMA models |
站点 | 时间尺度 | 模型 | MAE | RMSE | R2 |
---|---|---|---|---|---|
福海站 | 1 | ARIMA | 0.808 0 | 1.030 7 | -0.054 4 |
EEMD-ARIMA | 0.513 4 | 0.606 7 | 0.634 7 | ||
3 | ARIMA | 0.478 8 | 0.601 5 | 0.540 8 | |
EEMD-ARIMA | 0.262 4 | 0.320 0 | 0.870 0 | ||
6 | ARIMA | 0.373 1 | 0.465 4 | 0.662 7 | |
EEMD-ARIMA | 0.139 6 | 0.170 8 | 0.954 6 | ||
9 | ARIMA | 0.268 3 | 0.359 2 | 0.742 0 | |
EEMD-ARIMA | 0.099 2 | 0.125 7 | 0.968 4 | ||
12 | ARIMA | 0.212 6 | 0.285 6 | 0.816 3 | |
EEMD-ARIMA | 0.079 0 | 0.094 3 | 0.980 0 | ||
24 | ARIMA | 0.150 6 | 0.200 6 | 0.848 4 | |
EEMD-ARIMA | 0.048 7 | 0.060 7 | 0.986 1 | ||
巴音布鲁克站 | 1 | ARIMA | 0.847 0 | 1.074 5 | -0.086 6 |
EEMD-ARIMA | 0.538 9 | 0.636 8 | 0.618 3 | ||
3 | ARIMA | 0.600 3 | 0.775 0 | 0.375 1 | |
EEMD-ARIMA | 0.323 2 | 0.392 4 | 0.839 8 | ||
6 | ARIMA | 0.504 5 | 0.653 6 | 0.514 1 | |
EEMD-ARIMA | 0.213 5 | 0.266 4 | 0.919 3 | ||
9 | ARIMA | 0.387 6 | 0.552 1 | 0.621 4 | |
EEMD-ARIMA | 0.160 3 | 0.223 3 | 0.938 1 | ||
12 | ARIMA | 0.272 5 | 0.413 4 | 0.766 7 | |
EEMD-ARIMA | 0.117 5 | 0.165 3 | 0.962 7 | ||
24 | ARIMA | 0.165 1 | 0.241 4 | 0.918 4 | |
EEMD-ARIMA | 0.062 2 | 0.078 6 | 0.991 3 | ||
莎车站 | 1 | ARIMA | 0.550 5 | 0.742 7 | 0.000 4 |
EEMD-ARIMA | 0.372 0 | 0.460 0 | 0.616 5 | ||
3 | ARIMA | 0.522 7 | 0.699 5 | 0.445 3 | |
EEMD-ARIMA | 0.254 4 | 0.341 0 | 0.868 1 | ||
6 | ARIMA | 0.392 4 | 0.584 0 | 0.655 8 | |
EEMD-ARIMA | 0.253 0 | 0.295 1 | 0.912 1 | ||
9 | ARIMA | 0.315 3 | 0.497 0 | 0.706 1 | |
EEMD-ARIMA | 0.139 1 | 0.192 1 | 0.956 1 | ||
12 | ARIMA | 0.201 0 | 0.326 0 | 0.845 2 | |
EEMD-ARIMA | 0.089 0 | 0.130 1 | 0.975 3 | ||
24 | ARIMA | 0.136 4 | 0.214 3 | 0.935 4 | |
EEMD-ARIMA | 0.049 9 | 0.064 1 | 0.994 2 |
1 |
莫兴国,胡 实,卢洪健,等. GCM预测情景下中国21世纪干旱演变趋势分析[J]. 自然资源学报,2018,33(7) :1 244-1 256.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
2 |
黄荣辉,杜振彩. 全球变暖背景下中国旱涝气候灾害的演变特征及趋势[J]. 自然杂志,2010,32(4):187-195,184.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
3 |
郭嘉豪,王会肖,赵茹欣,等. 基于多维Copula函数的松嫩平原干旱特征分析[J]. 节水灌溉,2020(12):57-63.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
4 |
高珊,陈杰,许朗. 黄淮海地区干旱分区研究[J]. 节水灌溉,2020(10):101-106.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
5 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
6 |
袁梦,畅建霞,黎云云. 基于综合干旱指数的渭河流域干旱时空分析[J]. 武汉大学学报(工学版),2018,51(5):401-408.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
7 |
杜瑞麒,张智韬,巨娟丽,等. 基于波文比和降水的综合干旱指数的构建及应用[J]. 节水灌溉,2020(8):63-71.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
8 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
9 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
10 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
11 |
岳 元,申双和,王 琪. 基于标准化降水指数的近45年吉林省干旱特征分析[J]. 中国农村水利水电,2017(7):166-172
{{custom_citation.content}}
{{custom_citation.annotation}}
|
12 |
刘卫林,朱圣男,刘丽娜,等. 基于SPEI的1958-2018年鄱阳湖流域干旱时空特征及其与ENSO的关系[J]. 中国农村水利水电,2020(4):116-123,128.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
13 |
王盈盈,王志良,张泽中,等. 基于SPEI的贵州省分区干旱时空演变特征[J]. 灌溉排水学报,2019,38(6):119-128.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
14 |
于家瑞,艾 萍,袁定波, 等. 基于SPI的黑龙江省干旱时空特征分析 [J] . 干旱区地理,2019,42(5):1 059-1 068.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
15 |
郭 伟,李 莹,杜莉丽. 基于SPI的山西省1972-2012年春夏干旱特征及对玉米产量的影响分析[J]. 干旱地区农业研究,2018,36(1):230-236,265.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
16 |
王理萍,王树仿,王新华,等. 五种干旱指数在云南省的适用性分析 [J] . 灌溉排水学报,2017,36(7):117-124.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
17 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
18 |
韩 萍,王鹏新,王彦集,等. 多尺度标准化降水指数的ARIMA模型干旱预测研究[J].干旱地区农业研究,2008,26(2):212-218.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
19 |
李佳佳,贺新光,卢希安. 长江流域月降水的EEMD多时间尺度遥相关分析[J]. 长江流域资源与环境,2019,28(8):1 898-1 908.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
20 |
王 佳,王 旭,王 浩,等. 基于EEMD与ANN混合方法的水库月径流预测[J]. 人民黄河,2019,41(5):43-46.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
21 |
李 辉, 焦 毛, 杨晓萍, 等. 基于EEMD和SOM神经网络的水电机组故障诊断[J]. 水力发电学报,2017,36(7):83-91.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
22 |
李勃旭, 南西康, 郑向东, 等. 基于EMD-ARIMA模型的地铁门传动系统早期故障预测[J]. 计算机系统应用,2019,28(9):110-117.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
23 |
白致威, 张 雷, 王 杰, 等. 基于ARIMA模型的云南气象干旱预测研究[J]. 人民长江,2015,46(15) :6-9.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
24 |
吾米提·居马太, 瓦力江·瓦黑提, 黄娟. 新疆气象灾害对主要作物产量影响的灰色关联分析[J]. 沙漠与绿洲气象,2020,14(1):132-137.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
25 |
轩俊伟,郑江华,刘志辉.基于SPEI的新疆干旱时空变化特征[J]. 干旱区研究,2016,33(2):338-344.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
26 |
冯冬蕾,程志刚,赵 雷,等. 4种干旱判别指数在东北地区适用性分析[J]. 干旱区地理,2020,43(2):371-379.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
27 |
郭嘉豪,王会肖,赵茹欣,等. 基于最优拟合函数的SPI指数的松嫩平原干旱特征分析[J] . 北京师范大学学报(自然科学版),2020,56(2):240-249.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
28 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
29 |
杨 蕊,王 龙,高 瑞,等. 基于标准化降水指数的云南冬春干旱特征分析[J]. 中国农村水利水电,2017(4):36-40,44.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
30 |
杨慧荣,张玉虎,崔恒建,等. ARIMA和ANN模型的干旱预测适用性研究[J]. 干旱区地理,2018,41(5):945-953.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
31 |
左秀霞. 带高次趋势项的ADF单位根检验 [J] . 数量经济技术经济研究,2019,36(1):152-169.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
32 |
汉密尔顿. 时间序列分析[M]. 北京:中国社会科学出版社,1999:531-542.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
33 |
张 莹,谭艳春,彭发定,等. 基于EEMD和ARIMA的海温预测模型研究[J]. 海洋学研究,2019,37(1):9-14.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
34 |
李 宁,刘 瑜,王大刚. 基于ARIMA和EEMD的东江流域季节降水预报研究[J]. 人民珠江,2019,40(3):52-58,70.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
35 |
张建海,张 棋,许德合,等. ARIMA-LSTM组合模型在基于SPI干旱预测中的应用:以青海省为例[J]. 干旱区地理,2020,43(4):1 004-1 013.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
36 |
温克刚, 史玉光. 中国气象灾害大典(新疆卷)[M]. 北京:气象出版社,2006.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
37 |
孙 鹏,张 强,刘剑宇,等. 新疆近半个世纪以来季节性干旱变化特征及其影响研究[J]. 地理科学,2014,34(11):1 377-1 384.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_ref.label}} |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
/
〈 |
|
〉 |