基于LSTM神经网络的青藏高原月降水量预测

doi:10.12082/dqxxkx.2020.190378

摘要/Abstract

摘要：

青藏高原的降水量预测不仅为该地区水资源合理规划利用提供依据,同时对中国及周边国家气候变化研究有着重要的意义。论文利用1990—2016年青藏高原降水量数据,采用长短期记忆神经网络（LSTM）对青藏高原月降水量进行预测,主要包括：① 使用青藏高原86个测站1990—2013年的月降水资料,预测各个测站2014—2016年的月降水量,并与传统的RNN、NAR、SSA和ARIMA预测模型相比,平均决定系数R²分别提高了0.07、0.15、0.13和0.36,均方根误差（RMSE）和平均绝对误差（MAE）表现更低;② 分析了降水量预测精度的空间分布特征,将各模型的R²在青藏高原地区内插值,分析R²的空间分布特征,发现所有模型降雨稀少的干旱地区和降雨多的湿润地区R²较低,在气候稳定、降水规律性明显的地区R²较高,且LSTM模型R²≥0.6的空间范围远大于传统模型;③ 分析了不同预测长度对各模型预测精度的影响,发现所有模型会随着预测长度增加而预测精度降低,但在不同的预测长度下LSTM预测的RMSE值都低于其他模型。

关键词: 长短期记忆网络, 降水量, 预测, 循环神经网络, 青藏高原, 时间序列, 机器学习

Abstract:

Precipitation prediction on the Qinghai-Tibet Plateau not only provides a basis for rational planning and utilization of water resources, but also has significance for climate change research in China and neighboring countries. In this paper, the Long Short Term Memory neural network （LSTM） was used to predict the monthly precipitation over the Qinghai-Tibet Plateau using data from 1990 to 2016. Firstly, the monthly precipitation data of 86 stations in the Qinghai-Tibet Plateau from 1990 to 2013 were used to predict the monthly precipitation of each station from 2014 to 2016. Comparing with the traditional RNN, NAR, SSA, and ARIMA prediction models, LSTM increased the average coefficient of determination （R²） by 0.07, 0.15, 0.13, and 0.36, respectively. Simultaneously, LSTM had lower Root Mean Squared Error （RMSE） and Mean Absolute Error （MAE）. Among them, the observation of station 56106 showed that the LSTM model predicted the period more accurately with less displacement deviation, and that the prediction of the valley between July and September was more accurate with R² reaching 0.87. Secondly, the spatial distribution characteristics of precipitation prediction accuracy were analyzed. The R²of each model was interpolated in the Qinghai-Tibet Plateau, and the spatial distribution characteristics of R²were analyzed. All the drought areas with rare rainfall and the wet areas with heavy rainfall were of lower R², while the areas with stable climate and obvious precipitation were of higher R². Areas of R² over 0.6 were much larger when using the LSTM model than the traditional model. Finally, influence of different prediction lengths on the prediction accuracy was analyzed for each model. All models showed decreased prediction accuracy as the prediction length increased, yet the RMSE values predicted by LSTM were lower than by other models with the varying prediction lengths.

Key words: LSTM neural network, precipitation, prediction, RNN neural network, Qinghai-Tibet Plateau, time series, machine learning

刘新, 赵宁, 郭金运, 郭斌. 基于LSTM神经网络的青藏高原月降水量预测[J]. 地球信息科学学报, 2020, 22(8): 1617-1629.DOI:10.12082/dqxxkx.2020.190378

LIU Xin, ZHAO Ning, GUO Jinyun, GUO Bin. Prediction of Monthly Precipitation over the Tibetan Plateau based on LSTM Neural Network[J]. Journal of Geo-information Science, 2020, 22(8): 1617-1629.DOI:10.12082/dqxxkx.2020.190378

图/表 16

图1

图2

图3

图4

图5

表1

图6

图7

图8

图9

表2

图10

表3

图11

图12

表4

参考文献 29

[1]	Immerzeel W W, Beek L P H V, Bierkens M F P. Climate change will affect the asian water towers[J]. Science, 2010,328(5984):1382-1385. doi: 10.1126/science.1183188 pmid: 20538947
[2]	Song C, Sheng Y. Contrasting evolution patterns between glacier-fed and non-glacier-fed lakes in the Tanggula Mountains and climate cause analysis[J]. Climatic Change, 2016,135(3-4):1-15.
[3]	冯松, 汤懋苍, 王冬梅. 青藏高原是我国气候变化启动区的新证据[J]. 科学通报, 1998,43(6):633-636.
	[ Feng S, Tang M C, Wang D M. The Qinghai-Tibet Plateau is a new evidence for the climate change start-up zone in China[J]. Chinese Science Bulletin, 1998,43(6):633-636. ]
[4]	潘保田, 李吉均. 青藏高原:全球气候变化的驱动机与放大器-Ⅲ.青藏高原隆起对气候变化的影响[J]. 兰州大学学报, 1996,32(1):108-115.
	[ Pan B J, Li J J. Qinghai-Tibetan Plateau: A driver and amplifier of the global climatic change[J]. Journal of LanZhou University, 1996,32(1):108-115. ]
[5]	Yao T, Thompson L, Yang W, et al. Different glacier status with atmospheric circulations in Tibetan Plateau and surroundings[J]. Nature Climate Change, 2012,2(9):663-667.
[6]	Ke L, Ding X, Li W, et al. Remote sensing of glacier change in the central Qinghai-Tibet Plateau and the relationship with changing climate[J]. Remote Sensing, 2017,9(2):114-130.
[7]	Nayagam L R, Janardanan R, Mohan H S R. An empirical model for the seasonal prediction of southwest monsoon rainfall over Kerala, a meteorological subdivision of India[J]. International Journal of Climatology, 2010,28(6):823-831.
[8]	赵莹. 葠窝水库降雨量灰色预测模型应用分析[J]. 中国水能及电气化, 2016,141(12):68-70.
	[ Zhao Y. Analysis on application of rainfall grey prediction model for Shenwo reservoir[J]. China Water Power & Electrification, 2016,141(12):68-70. ]
[9]	潘刚, 芦冰, 邹兵, 等. 马尔可夫链在水库主汛期降雨状态预测中的应用[J]. 水利科技与经济, 2011,17(6):33-36.
	[ Pan G, Lu B, Zou B, et al. Markov chain state in the reservoir the main flood season rainfall forecast[J]. Water Conservancy Science & Technology & Economy, 2011,17(6):33-36. ]
[10]	Feng C Y. Application of singular spectrum analysis to summer precipitation prediction[J]. Meteorological Monthly, 2002,28(11):22-25.
[11]	Burlando P, Rosso R, Cadavid L G, et al. Forecasting of short-term rainfall using ARMA models[J]. Journal of Hydrology, 1993,144(1-4):193-211.
[12]	Liu C, Tian Y, Wang X H. Study of rainfall prediction model based on GM (1, 1) - Markov chain[C]. Xi'an: 2011 International Symposium on Water Resource and Environmental Protection, 2011,18(1):744-747.
[13]	宋帆, 杨晓华, 武翡翡, 等. 基于聚类分析的模糊马尔科夫链在降雨量预测中的应用[J]. 节水灌溉, 2018,278(10):38-41,46.
	[ Song F, Yang X H, Wu F F, et al. Rainfall prediction using clustering-fuzzy-markov chain model[J]. Water Saving Irrigation, 2018,278(10):38-41,46. ]
[14]	Luk K C, Ball J E, Sharma A. An application of artificial neural networks for rainfall forecasting[J]. Mathematical & Computer Modelling, 2001,33(6):683-693.
[15]	Nanda T, Sahoo B, Beria H, et al. A wavelet-based non-linear autoregressive with exogenous inputs (WNARX) dynamic neural network model for real-time flood forecasting using satellite-based rainfall products[J]. Journal of Hydrology, 2016,539(87):57-73.
[16]	Chattopadhyay S, Chattopadhyay G. Comparative study among different neural net learning algorithms applied to rainfall time series[J]. Meteorological Applications, 2010,15(2):273-280.
[17]	Nasseri M, Asghari K, Abedini M J. Optimized scenario for rainfall forecasting using genetic algorithm coupled with artificial neural network[J]. Expert Systems with Applications, 2008,35(3):1415-1421.
[18]	Baratta D, Masulli F, Cicioni G, et al. Application of an ensemble technique based on singular spectrum analysis to daily rainfall forecasting[J]. Neural Networks, 2003,16(3):375-387.
[19]	Wu C L, Chau K W. Prediction of rainfall time series using modular soft computingmethods[J]. Engineering Applications of Artificial Intelligence, 2013,26(3):997-1007.
[20]	Yaseen Z M, Ghareb M I, Ebtehaj I, et al. Rainfall pattern forecasting using novel hybrid intelligent model based ANFIS-FFA[J]. Water Resources Management, 2017,8(32):105-122.
[21]	Chau K W, Wu C L. A hybrid model coupled with singular spectrum analysis for daily rainfall prediction[J]. Journal of Hydroinformatics, 2010,12(4):458-473.
[22]	Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997,9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735 pmid: 9377276
[23]	Graves A . Supervised sequence labelling with recurrent neural networks[J]. Studies in Computational Intelligence, 2012,2(385):42-45.
[24]	Graves A, Schmidhuber J. Framewise phoneme classification with bidirectional LSTM and other neural network architectures[J]. Neural Networks, 2005,18(5-6):602-610. doi: 10.1016/j.neunet.2005.06.042 pmid: 16112549
[25]	Duchi J, Hazan E, Singer Y. Adaptive subgradient methods for online learning and stochastic optimization[J]. Journal of Machine Learning Research, 2011,12(7):257-269.
[26]	Kingma D P, Ba J. Adam: A method for stochastic optimization[C]. 3rd International Conference for Learning Representations, 2015,78(1):116-130.
[27]	Vautard R. Singular-spectrum analysis: A toolkit for short, noisy chaotic signals[J]. Physica D: Nonlinear Phenom. 1992,9(58):95-126.
[28]	Golyandina N, Korobeynikov A. Basic singular spectrum analysis and forecasting with R[J]. Computational Statistics & Data Analysis, 2014,71(12):934-954.
[29]	Box G E, Jenkins G M. Time series analysis: Forecasting and control rev.ed.[J]. Journal of Time, 1976,31(4):238-242.

自相关系数	偏相关系数	模型定阶
拖尾	p阶截尾	AR（p）模型
q阶截尾	拖尾	MA（q）模型
拖尾	拖尾	ARMA（p, q）模型

评价指标	LSTM	RNN	NAR	SSA	ARIMA
RMSE	19.69	29.82	32.15	33.77	54.46
MAE	14.06	18.57	24.17	22.17	40.67
R²	0.87	0.70	0.65	0.61	0.28

评价指标	LSTM	RNN	NAR	SSA	ARIMA
RMSE	26.83	28.87	31.64	30.17	38.01
MAE	17.91	20.65	22.77	21.04	26.47
R²	0.61	0.55	0.46	0.48	0.25

预测长度/个月	LSTM	RNN	NAR	SSA	ARIMA
36	26.83	28.87	31.64	30.17	38.01
24	26.23	28.16	30.45	29.37	35.82
12	25.21	27.52	29.95	28.78	36.24
6	19.89	23.85	26.58	25.10	27.30