| 试验 | 特征 | 辅助变量 | 嵌入层 | 时间滞后变量 |
| 1 | 有 | 无 | 无 | 无 |
| 2 | 有 | 有 | 无 | 无 |
| 3 | 有 | 有 | 有 | 无 |
| 4 | 有 | 有 | 有 | 有 |
| Citation: | Zhao Linna, Lu Shu, Qi Dan, et al. Daily maximum air temperature forecast based on fully connected neural network. J Appl Meteor Sci, 2022, 33(3): 257-269. DOI: 10.11898/1001-7313.20220301.
|
Objective forecast of maximum temperature is an important part in numerical weather prediction(NWP). The forecast uncertainty of near-surface meteorological elements is greater than that of upper atmospheric elements due to the impact of uncertainty in numerical forecasting models for sub-grid and boundary layer schemes.In recent years, meteorological observations expand rapidly, making traditional error correct method difficult to deal with the massive data. As a result, artificial intelligence has an increasingly obvious advantage in processing big data. Based on the fully connected neural network, four sensitivity experiments are designed in order to investigate the importance of auxiliary variable, time-lagged variable and the effectiveness of embedding layer in the neural network. The output products of high resolution(HRES) model of European Centre for Medium-Range Weather Forecasts(ECMWF) and the observations of basic meteorological elements of totally 2238 basic weather stations from 15 January 2015 to 31 December 2020 are employed. The training period is from 15 January 2015 to 31 December 2019, and the rest part is test period.The results show that the forecast error of daily maximum air temperature from the HRES in test period is reduced greatly by the sensitivity experiments, which add auxiliary variables, daily maximum air temperature with 1-2 lag days and embedding layer structures and their combination. The root mean square error is reduced by 29.72%-47.82% and the accuracy of temperature forecast are increased by 16.67%-38.89%, and the effects for Qinghai-Tibet Plateau is especially remarkable where the forecast error of HRES model is very high. It is preliminarily proved that the fully connected neural network with embedding layer has better overall performance than the raw fully connected neural network, and the features also affect the forecast errors and forecast skills of the model. Besides, the prediction error of neural network model with embedding layer is more stable when auxiliary variables and lag time variables are added. Positive forecasting techniques are available for almost all stations in the study, and it is possible to reduce the mean absolute error to less than 1℃ at many stations.
Fig. 1 Topography(the shaded) of the target area and distribution of stations(black dots)
Fig. 2 Flow chart of multi-input fully connected neural network
Fig. 3 Scatter plot and kernel density of daily maximum air temperature between observation and forecasting
(the red solid line denotes the diagonal, the black dashed line denotes the fitting line)
Fig. 4 Prediction skill score for forecasted daily maximum temperature(unit: %)
Fig. 5 Box plot of root mean square error of HRES and each test in test dataset during Jan-Dec
Table 1 Experiments of features and embedding layers on the structure of multi-input neural network
| 试验 | 特征 | 辅助变量 | 嵌入层 | 时间滞后变量 |
| 1 | 有 | 无 | 无 | 无 |
| 2 | 有 | 有 | 无 | 无 |
| 3 | 有 | 有 | 有 | 无 |
| 4 | 有 | 有 | 有 | 有 |
DownLoad: Download CSV
Table 2 Ratio of positive skills in different regions(unit: %)
| 试验 | 东北 | 新疆 | 西北地区东部 | 华北 | 青藏高原中南部 | 西南地区东部 | 长江中下游 | 华南 |
| 1 | 96.35 | 87.50 | 79.38 | 91.45 | 89.87 | 85.64 | 96.71 | 97.46 |
| 2 | 78.54 | 66.67 | 73.75 | 76.92 | 83.54 | 78.10 | 84.87 | 87.31 |
| 3 | 99.09 | 94.79 | 96.88 | 98.72 | 98.73 | 98.78 | 99.18 | 97.97 |
| 4 | 99.54 | 98.96 | 98.13 | 98.93 | 100.00 | 99.76 | 99.34 | 100.00 |
DownLoad: Download CSV
Table 3 Positive skill ratio of mean absolute error no more than 2℃ in different regions(unit: %)
| 试验 | 东北 | 新疆 | 西北地区东部 | 华北 | 青藏高原中南部 | 西南地区东部 | 长江中下游 | 华南 |
| 1 | 98.58 | 89.29 | 96.85 | 99.07 | 57.75 | 78.69 | 97.96 | 97.92 |
| 2 | 97.09 | 82.81 | 95.76 | 98.89 | 51.52 | 78.82 | 97.48 | 97.67 |
| 3 | 99.54 | 97.80 | 100.00 | 99.57 | 79.49 | 92.86 | 99.67 | 99.48 |
| 4 | 100.00 | 100.00 | 100.00 | 100.00 | 98.73 | 99.02 | 99.83 | 100.00 |
DownLoad: Download CSV
Table 4 Positive skill ratio of mean absolute error no more than 1℃ in different regions(unit: %)
| 试验 | 东北 | 新疆 | 西北地区东部 | 华北 | 青藏高原中南部 | 西南地区东部 | 长江中下游 | 华南 |
| 1 | 17.54 | 17.86 | 14.17 | 39.72 | 5.63 | 0.00 | 29.25 | 6.25 |
| 2 | 0.58 | 4.69 | 5.93 | 8.89 | 4.55 | 0.00 | 6.59 | 4.07 |
| 3 | 24.42 | 19.78 | 20.65 | 39.18 | 8.97 | 1.97 | 23.38 | 5.70 |
| 4 | 43.12 | 48.42 | 48.41 | 70.41 | 37.97 | 11.95 | 43.71 | 22.34 |
DownLoad: Download CSV
Table 5 Average positive skill scores in different regions(unit: %)
| 试验 | 东北 | 新疆 | 西北地区东部 | 华北 | 青藏高原中南部 | 西南地区东部 | 长江中下游 | 华南 |
| 1 | 19.19 | 23.38 | 35.86 | 23.46 | 57.81 | 33.38 | 25.31 | 28.44 |
| 2 | 10.84 | 20.05 | 34.72 | 17.60 | 57.16 | 32.47 | 16.14 | 21.17 |
| 3 | 20.75 | 27.78 | 36.85 | 23.84 | 64.97 | 39.74 | 26.88 | 30.56 |
| 4 | 27.20 | 37.40 | 43.47 | 30.72 | 71.04 | 46.46 | 33.18 | 37.53 |
DownLoad: Download CSV
| [1] |
Lin A L, Gu D J, Peng D D, et al. Climatic characteristics of regional persistent heart event in the eastern China during recent 60 years. J Appl Meteor Sci, 2021, 32(3): 302-314. doi: 10.11898/1001-7313.20210304
|
| [2] |
Shi L, Liang N, Xu X, et al. SA-JSTN: Self-attention joint spatiotemporal network for temperature forecasting. IEEE J Sel Top Appl Earth Obs Remote Sens, 2021, 14: 9475-9485. doi: 10.1109/JSTARS.2021.3112131
|
| [3] |
Tran T T K, Lee T, Shin J Y, et al. Deep learning-based maximum temperature forecasting assisted with meta-learning for hyperparameter optimization. Atmosphere, 2020, 11(5): 487. doi: 10.3390/atmos11050487
|
| [4] |
Yin S, Li Y, Ma J, et al. Ensemble learning for bias correction of station temperature forecast based on ECMWF products. Meteor Mon, 2020, 46(3): 412-419. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX202003012.htm
|
| [5] |
Wu Q S, Han M, Guo H, et al. The optimal training period scheme of MOS temperature forecast. J Appl Meteor Sci, 2016, 27(4): 426-434. doi: 10.11898/1001-7313.20160405
|
| [6] |
Wang D, Wang J P, Bai Q M, et al. Comparative correction of air temperature forecast from ECMWF model by the decaying averaging and the simple linear regression methods. Meteor Mon, 2019, 45(9): 1310-1321. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201909011.htm
|
| [7] |
Zhao L N, Liu Y, Bao H J, et al. The probabilistic flood prediction based on implementation of the Schaake shuffle method over the Huaihe Basin. J Appl Meteor Sci, 2017, 28(5): 544-554. doi: 10.11898/1001-7313.20170503
|
| [8] |
Wei G F, Liu H J, Wu Q S, et al. Multi-model consensus forecasting technology with optimal weight for precipitation intensity levels. J Appl Meteor Sci, 2020, 31(6): 668-680. doi: 10.11898/1001-7313.20200603
|
| [9] |
Ouyang X, Chen D, Lei Y. A generalized evaluation scheme for comparing temperature products from satellite observations, numerical weather model, and ground measurements over the Tibetan Plateau. IEEE Trans Geosci Remote Sens, 2018, 56(7): 3876-3894. doi: 10.1109/TGRS.2018.2815272
|
| [10] |
Sun J, Cao Z, Li H, et al. Application of artificial intelligence technology to numerical weather prediction. J Appl Meteor Sci, 2021, 32(1): 1-11. doi: 10.11898/1001-7313.20210101
|
| [11] |
Han F, Yang L, Zhou C X, et al. An experimental study of the short-time heavy rainfall event forecast based on ensemble learning and sounding data. J Appl Meteor Sci, 2021, 32(2): 188-199. doi: 10.11898/1001-7313.20210205
|
| [12] |
Liu N, Xiong A Y, Zhang Q, et al. Development of basic dataset of severe convective weather for artificial intelligence training. J Appl Meteor Sci, 2021, 32(5): 530-541. doi: 10.11898/1001-7313.20210502
|
| [13] |
Sun Q D, Jiao R L, Xia J J, et al. Adjusting wind speed prediction of numerical weather forecast model based on machine learning methods. Meteor Mon, 2019, 45(3): 426-436. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201903012.htm
|
| [14] |
Jin Z Q, Wang X M, Bao Y S, et al. Squall line identification method based on convolution neural network. J Appl Meteor Sci, 2021, 32(5): 580-591. doi: 10.11898/1001-7313.20210506
|
| [15] |
Wei J, Li Z, Cribb M, et al. Improved 1 km resolution PM2.5 estimates across China using enhanced space-time extremely randomized trees. Atmos Chem Phys, 2020, 20(6): 3273-3289. doi: 10.5194/acp-20-3273-2020
|
| [16] |
Li W, Gao X, Hao Z, et al. Using deep learning for precipitation forecasting based on spatio-temporal information: A case study. Climate Dyn, 2022, 58(1): 443-457.
|
| [17] |
Cai J Q, Tan G R, Niu R Y. Circulation pattern classification of persistent heavy rainfall in Jianghuai Region based on the transfer learning CNN model. J Appl Meteor Sci, 2021, 32(2): 233-244. doi: 10.11898/1001-7313.20210208
|
| [18] |
Wang B, Lu J, Yan Z, et al. Deep Uncertainty Quantification: A Machine Learning Approach For Weather Forecasting//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019: 2087-2095.
|
| [19] |
Jeong S, Park I, Kim H S, et al. Temperature prediction based on bidirectional long short-term memory and convolutional neural network combining observed and numerical forecast data. Sensors, 2021, 21(3): 941. doi: 10.3390/s21030941
|
| [20] |
Rasp S, Lerch S. Neural networks for postprocessing ensemble weather forecasts. Mon Wea Rev, 2018, 146(11): 3885-3900. doi: 10.1175/MWR-D-18-0187.1
|
| [21] |
Chen Y W, Huang X M, Li Y, et al. Ensemble learning for bias correction of station temperature forecast based on ECMWF products. J Appl Meteor Sci, 2020, 31(4): 494-503. doi: 10.11898/1001-7313.20200411
|
| [22] |
Tan J H, Chen W L, Wang S S. Using a machine learning method for temperature forecast in Hubei Province. Adv Meteor Sci Tech, 2018, 8(5): 46-50. doi: 10.3969/j.issn.2095-1973.2018.05.006
|
| [23] |
Cho D, Yoo C, Im J, et al. Comparative assessment of various machine learning-based bias correction methods for numerical weather prediction model forecasts of extreme air temperatures in urban areas. Earth Space Sci, 2020, 7(4): e2019EA000740.
|
| [24] |
Ren P, Chen M X, Cao W H, et al. Error analysis and correction of short-term numerical weather prediction under complex terrain based on machine learning. Acta Meteor Sinica, 2020, 78(6): 1002-1020. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXB202006009.htm
|
| [25] |
Zamani J M, Cao C, Ni X, et al. PM2.5 prediction based on random forest, XGBoost, and deep learning using multisource remote sensing data. Atmosphere, 2019, 10(7): 373. doi: 10.3390/atmos10070373
|
| [26] |
Jiang G Q, Xu J, Wei J. A deep learning algorithm of neural network for the parameterization of typhoon-ocean feedback in typhoon forecast models. Geophys Res Lett, 2018, 45(8): 3706-3716. doi: 10.1002/2018GL077004
|
| [27] |
Tran T T K, Lee T, Kim J S. Increasing neurons or deepening layers in forecasting maximum temperature time series. Atmosphere, 2020, 11(10): 1072. doi: 10.3390/atmos11101072
|
| [28] |
Veldkamp S, Whan K, Dirksen S, et al. Statistical postprocessing of wind speed forecasts using convolutional neural networks. Mon Wea Rev, 2021, 149(4): 1141-1152. doi: 10.1175/MWR-D-20-0219.1
|
| [29] |
Yu X, Shi S, Xu L, et al. A novel method for sea surface temperature prediction based on deep learning. Math Probl Eng, 2020: 1-9.
|
| [30] |
Wang J, Xu Z F, Fan G Z, et al. Study on bias correction for the 2 m temperature forecast of GRAPES_RAFS. Meteor Mon, 2015, 41(6): 719-726. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201506006.htm
|
| [31] |
Zhang C J, Zeng J, Wang H Y, et al. Correction model for rainfall forecasts using the LSTM with multiple meteorological factors. Meteor Appl, 2020, 27(1): e1852.
|
| [32] |
Yang Q L. High-dimensional LBP Feature Selection Method For Meteorological Cloud Image Classification. Taiyuan: Shanxi University, 2021.
|
| [33] |
Liu M Y, Wu L J, Liang H, et al. A kind of high-precision LSTM-FC atmospheric contaminant concentrations forecasting model. Comput Sci, 2021, 48(6A): 184-189. doi: 10.11896/jsjkx.200600090
|
| [34] |
Hou H, Chen X, Li M, et al. A space prediction method for power outage in a typhoon disaster based on a Stacking integrated structure. Electr Power Syst Res, 2022(3): 76-84. https://www.cnki.com.cn/Article/CJFDTOTAL-JDQW202203009.htm
|
| [35] |
Lai X F, Liang X W, Xie Z C, et al. Intrusion detection method based on entity embedding and long short-term memory networks. Journal of University of Chinese Academy of Sciences, 2020, 37(4): 553-561. https://www.cnki.com.cn/Article/CJFDTOTAL-ZKYB202004016.htm
|
| [36] |
Mikolov T, Chen K, Corrado G, et al. Efficient estimation of word representations in vector space. arXiv preprint arXiv: 1301.3781, 2013.
|
| [37] |
Cerqueira V, Torgo L, Mozeti I. Evaluating time series forecasting models: An empirical study on performance estimation methods. Mach Learn, 2020, 109(11): 1997-2028. doi: 10.1007/s10994-020-05910-7
|
| [38] |
Wang Y. Verification of NMC subjective and objective precipitation prediction during the main flood season in 2002. Meteor Mon, 2003, 29(5): 21-25. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX200305006.htm
|
| [39] |
Xiong M Q. Calibrating daily 2 m maximum and minimum air temperature forecasts in the ensemble prediction system. Acta Meteor Sinica, 2017, 75(2): 211-222. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXB201702002.htm
|
| [40] |
Hao C, Zhang Y X, Wang Z W, et al. Application of analog ensemble rectifying method in objective temperature prediction. Meteor Mon, 2019, 45(8): 1085-1092. https://www.cnki.com.cn/Article/CJFDTOTAL-QXXX201908005.htm
|
Figures(5) / Tables(5)
