Improvement of Deep Learning Method for Daily Precipitation Downscaling
-
摘要: 为了提高深度学习方法对全球气候模式（GCMs）日降水的降尺度效果，以长江流域为研究对象，基于20种GCMs输出的日降水数据，构建了4种深度学习降尺度模型，并与日偏差校正方法（DBC）结合，提出一种混合降尺度方法（DL-DBC）.4种深度学习方法对GCMs日降水的降尺度表现接近；与日偏差校正方法相比，其降尺度后的多年平均日降水的平均绝对相对误差（MARE）更低，但多年平均月降水和多年平均年降水的MARE略高，与深度学习方法相比，DL-DBC得到的多年平均年降水的MARE降低了6.7%~11.3%，多年平均月降水的MARE降低了6.3%~7.6%，且在降水量频率分析等方面同样表现更好.混合方法DL-DBC能提高深度学习模型对GCMs日降水数据的降尺度效果，进一步减小GCMs日尺度降水数据的偏差.Abstract: To improve the downscaling effectiveness of deep learning methods for daily precipitation from global climate models (GCMs). Targeting at the Yangtze River basin, we constructed four deep learning downscaling models based on historical daily precipitation outputs from 20 GCMs. A hybrid method (DL-DBC) was proposed by integrating these models with the daily bias correction method. The four deep learning models exhibited comparable performance in daily precipitation downscaling. Compared to DBC, they achieved lower mean absolute relative error (MARE) for multi-year average daily precipitation but slightly higher MARE for multi-year average monthly and annual precipitation. The DL-DBC method outperformed standalone deep learning models, reducing MARE for multi-year average annual precipitation by 6.7%-11.3% and monthly precipitation by 6.3%-7.6%, while also demonstrating superior performance in precipitation frequency analysis. The DL-DBC method enhances the downscaling effectiveness of deep learning models and further reduces biases in daily precipitation data from GCMs.
-
Key words:
- downscaling /
- deep learning /
- precipitation /
- global climate models /
- Yangtze River basin /
- climate change
-
图 1 长江流域1979-2014年间多年平均年降水量
数据来源于CN05.1数据集
Fig. 1. The multi-year average annual precipitation in the Yangtze River basin from 1961 to 2014
图 2 深度学习模型架构
图a为4种模型主干架构；图b为SPC上采样法的架构；图c介绍了深度学习架构中所用到的具体结构
Fig. 2. Deep learning model architecture
图 3 不同降尺度方法基于不同CGMs的降尺度降水与观测降水在测试期的MARE
图a为多年平均降水的计算结果；图b为多年平均月降水计算结果；图c为多年平均日降水计算结果
Fig. 3. MARE of downscaled precipitation based on different GCMs and observed precipitation during the testing period for various downscaling methods
图 4 任意一种GCMs降尺度结果的季节性平均降水量偏差（mm/d）的空间分布图以及偏差箱线图
季节以12~2月（DJF），3~5月（MAM），6~8月（JJA）和9~11月（SON）划分
Fig. 4. Seasonal average precipitation bias (mm/d) spatial distribution and boxplot of bias for the downscaled results of an arbitrary GCMs
图 5 模拟降水的IOA空间分布(a)以及20种GCMs降尺度结果的流域IOA均值(b)
Fig. 5. Spatial distribution of IOA for simulated precipitation (a) and the mean IOA of the watershed from the downscaling results of 20 GCMs (b)
图 6 任意1种GCMs模拟降水与观测降水在高降水区域(a)和低降水区域(b)的流域日平均降水Q-Q图
Fig. 6. Quantile-quantile (Q-Q) plots of basin daily average precipitation between simulated and observed precipitation for one selected GCMs in (a) high-precipitation areas and (b) low-precipitation areas
图 7 逐格点计算的模拟降水与观测降水的年湿日总数(a)、年最长湿期(b)、年高降水日总数(c)和年最高日降水量(d)的MAE均值
Fig. 7. Mean absolute error (MAE) for grid cell-wise calculations of annual wet days count (a), annual longest wet spell (b), annual high precipitation days count (c), and annual maximum daily precipitation (d) between simulated and observed precipitation
图 8 所有GCMs模拟降水与观测降水的年均月降水序列的RMSE(a)和MAE(b)的统计结果
Fig. 8. Statistical results of RMSE (a) and MAE (b) for the annual mean monthly precipitation sequences of observed and simulated precipitation from all GCMs
图 9 20种GCMs在测试期的模拟与观测年均日降水的空间半变函数图的RMSE(a)与皮尔逊相关系数r(c)，(b)、(d)为对20种GCMs测试的指标均值
Fig. 9. RMSE (a) and Pearson correlation coefficients r (c) of spatial semivariance for simulated and observed annual mean daily precipitation from 20 GCMs during the test period and mean values of these metrics across the 20 GCMs are presented in (b) and (d)
表 1 GCMs降水数据的初始分辨率及来源
Table 1. The initial resolution and sources of precipitation data from GCMs
GCMs 分辨率(纬度×经度) 国家/机构 GFDL-ESM4 1.00°×1.25° 美国国家海洋和大气管理局 ACCESS-CM2 1°×1° 澳大利亚 MPI-ESM1 1°×1° 德国 INM-CM4 1.5°×2.0° 俄罗斯 CSM2-MR 1°×1° 美国国家大气研究中心 CMCC-CM2-SR5 1.125°×1.112° 意大利 CNRM-CM6 1.41°×1.41° 法国 FGOALS-f3 1.00°×1.25° 中国 IPSL-CM6A 1.26°×2.50° 法国 MIROC6 1.41°×1.41° 日本 CESM2-WACCM 0.94°×1.25° 美国国家大气研究中心 TaiESM1 1.25°×0.94° 中国台湾区域预报中心 NorESM2-MM 0.94°×1.25° 挪威 INM-CM5 1.5°×2.0° 俄罗斯 CESM2 0.94°×1.25° 美国国家大气研究中心 EC-Earth3 0.70°×0.70° 地球系统模型联盟 UKESM1-0-LL 1.25°×1.88° 英国 CAMS-CSM1-0 1.11°×1.13° 中国 BCC-CSM2-MR 1.11°×1.13° 中国 AWI-ESM-1-1-LR 1.85°×1.88° 德国 
下载: 导出CSV
表 2 偏差校正与降尺度方法的名称
Table 2. Titles of various bias correction and downscaling methods
编号 校正方法 A1 LS校正(linear scaling bias correction) A2 DBC校正(daily bias correction method) B1 卷积神经网络CNN (convolutional neural network) B2 超分辨率深度残差网络SRDRN(super resolution deep residual network) B3 ConvNext网络 B4 密集连接卷积网络DenseNet (densely connected convolutional network) B1A2 基于CNN与DBC校正的混合方法 B1A2 基于SRDRN与DBC校正的混合方法 B1A2 基于ConvNext与DBC校正的混合方法 B1A2 基于DenseNet与DBC校正的混合方法
下载: 导出CSV
表 3 不同季节的空间半变函数图RMSE与皮尔逊相关系数r
Table 3. RMSE and Pearson correlation coefficient r of spatial semivariance for different seasons
方法 12~2月 3~5月 6~8月 9~11月 r RMSE r RMSE r RMSE r RMSE BL 0.836 0.842 0.877 5.238 0.852 8.133 0.889 3.191 A1 0.900 1.009 0.879 3.007 0.852 6.754 0.897 3.157 A2 0.918 0.669 0.912 2.233 0.863 4.022 0.905 1.709 B1 0.904 0.655 0.897 2.211 0.848 4.100 0.900 1.735 B2 0.908 0.661 0.903 2.253 0.847 4.092 0.904 1.735 B3 0.905 0.661 0.902 2.252 0.848 4.119 0.903 1.745 B4 0.905 0.650 0.903 2.178 0.848 4.322 0.900 1.823 B1A2 0.929 0.676 0.924 2.233 0.867 3.765 0.913 1.659 B2A2 0.930 0.674 0.925 2.214 0.867 3.785 0.916 1.664 B3A2 0.930 0.667 0.924 2.145 0.866 3.771 0.916 1.678 B4A2 0.930 0.670 0.924 2.237 0.866 3.789 0.917 1.676
下载: 导出CSV
-
Baño-Medina, J., Manzanas, R., Gutiérrez, J. M., 2020. Configuration and Intercomparison of Deep Learning Neural Models for Statistical Downscaling. Geoscientific Model Development, 13(4): 2109-2124. https://doi.org/10.5194/gmd-13-2109-2020 Bennett, J. C, Grose, M. R., Post, D. A., et al., 2011. Performance of Quantile-Quantile Bias-Correction for Use in Hydroclimatological Projections. In: Chan, F., Marinova, D., Anderssen, R. S., eds., 19th International Congress on Modelling and Simulation. Perth. https://doi.org/10.36334/modsim.2011.f5.bennett Cannon, A. J., 2018. Multivariate Quantile Mapping Bias Correction: An N-Dimensional Probability Density Function Transform for Climate Model Simulations of Multiple Variables. Climate Dynamics, 50(1): 31-49. https://doi.org/10.1007/s00382-017-3580-6 Chen, J., Brissette, F. P., Chaumont, D., et al., 2013. Performance and Uncertainty Evaluation of Empirical Downscaling Methods in Quantifying the Climate Change Impacts on Hydrology over Two North American River Basins. Journal of Hydrology, 479: 200-214. https://doi.org/10.1016/j.jhydrol.2012.11.062 Chen, J., Xu, C. Y., Guo, S. L., et al., 2016. Progress and Challenge in Statistically Downscaling Climate Model Outputs. Journal of Water Resources Research, 5(4): 299-313 (in Chinese with English abstract). Dickinson, R. E., Errico, R. M., Giorgi, F., et al., 1989. A Regional Climate Model for the Western United States. Climatic Change, 15(3): 383-422. https://doi.org/10.1007/BF00240465 Eden, J. M., Widmann, M., 2014. Downscaling of GCM-Simulated Precipitation Using Model Output Statistics. Journal of Climate, 27(1): 312-324. https://doi.org/10.1175/jcli-d-13-00063.1 Eum, H. I., Cannon, A. J., 2017. Intercomparison of Projected Changes in Climate Extremes for South Korea: Application of Trend Preserving Statistical Downscaling Methods to the CMIP5 Ensemble. International Journal of Climatology, 37(8): 3381-3397. https://doi.org/10.1002/joc.4924 Guo, Q., Chen, J., Zhang, X. C., et al., 2019. A New Two-Stage Multivariate Quantile Mapping Method for Bias Correcting Climate Model Outputs. Climate Dynamics, 53(5): 3603-3623. https://doi.org/10.1007/s00382-019-04729-w Harris, L., McRae, A. T. T., Chantry, M., et al., 2022. A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecasts. Journal of Advances in Modeling Earth Systems, 14(10): e2022MS003120. https://doi.org/10.1029/2022MS003120 He, X. G., Chaney, N. W., Schleiss, M., et al., 2016. Spatial Downscaling of Precipitation Using Adaptable Random Forests. Water Resources Research, 52(10): 8217-8237. https://doi.org/10.1002/2016WR019034 Huang, G., Liu, Z., Van Der Maaten, L., et al., 2017. Densely Connected Convolutional Networks. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu. https://doi.org/10.1109/CVPR.2017.243 Kingma, D., Ba, J., 2014. Adam: A method for Stochastic Optimization. The 3rd International Conference for Learning Representations, San Diego. Legasa, M. N., Manzanas, R., Calviño, A., et al., 2022. A Posteriori Random Forests for Stochastic Downscaling of Precipitation by Predicting Probability Distributions. Water Resources Research, 58(4): e2021WR030272. https://doi.org/10.1029/2021WR030272 Li, W. T., Pan, B. X., Xia, J. J., et al., 2022. Convolutional Neural Network-Based Statistical Post-Processing of Ensemble Precipitation Forecasts. Journal of Hydrology, 605: 127301. https://doi.org/10.1016/j.jhydrol.2021.127301 Lim, B., Son, S., Kim, H., et al., 2017. Enhanced Deep Residual Networks for Single Image Super-Resolution. 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Honolulu. https://doi.org/10.1109/CVPRW.2017.151 Liu, J., Chen, Q., Xu, Y., et al., 2024. Simulation of Phosphorus Inflow and Outflow Fluxes and Water Quality Prediction in Dongting Lake Area of the Yangtze River Basin: A Coupled Approach of Machine Learning and Traditional Hydrological Modeling. Earth Science, 49(11): 3995-4007 (in Chinese with English abstract). Liu, Y. H., Guo, W. D., Feng, J. M., et al., 2011. A Summary of Methods for Statistical Downscaling of Meteorological Data. Advances in Earth Science, 26(8): 837-847 (in Chinese with English abstract). Liu, Y. M., Ganguly, A. R., Dy, J., 2020. Climate Downscaling Using YNet: A Deep Convolutional Network with Skip Connections and Fusion. The 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. San Diego. https://doi.org/10.1145/3394486.3403366 Liu, Z., Mao, H., et al., 2022. A Convnet for the 2020s. The IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle. Mearns, L. O., Arritt, R., Biner, S., et al., 2012. The North American Regional Climate Change Assessment Program: Overview of Phase I Results. Bulletin of the American Meteorological Society, 93(9): 1337-1362. https://doi.org/10.1175/bams-d-11-00223.1 Meehl, G. A., Covey, C., Delworth, T., et al., 2007. THE WCRP CMIP3 Multimodel Dataset: A New Era in Climate Change Research. Bulletin of the American Meteorological Society, 88(9): 1383-1394. https://doi.org/10.1175/bams-88-9-1383 Mei, Y. W., Maggioni, V., Houser, P., et al., 2020. A Nonparametric Statistical Technique for Spatial Downscaling of Precipitation over High Mountain Asia. Water Resources Research, 56(11): e2020WR027472. https://doi.org/10.1029/2020WR027472 Monjo, R., Gaitán, E., Pórtoles, J., et al., 2016. Changes in Extreme Precipitation over Spain Using Statistical Downscaling of CMIP5 Projections. International Journal of Climatology, 36(2): 757-769. https://doi.org/10.1002/joc.4380 Pan, B. X., Anderson, G. J., Goncalves, A., et al., 2021. Learning to Correct Climate Projection Biases. Journal of Advances in Modeling Earth Systems, 13(10): e2021MS002509. https://doi.org/10.1029/2021MS002509 Pan, B. X., Hsu, K., AghaKouchak, A., et al., 2019. Improving Precipitation Estimation Using Convolutional Neural Network. Water Resources Research, 55(3): 2301-2321. https://doi.org/10.1029/2018WR024090 Pierce, D. W., Cayan, D. R., Maurer, E. P., et al., 2015. Improved Bias Correction Techniques for Hydrological Simulations of Climate Change. Journal of Hydrometeorology, 16(6): 2421-2442. https://doi.org/10.1175/jhm-d-14-0236.1 Pour, S. H., Shahid, S., Chung, E. S., 2016. A Hybrid Model for Statistical Downscaling of Daily Rainfall. Procedia Engineering, 154: 1424-1430. https://doi.org/10.1016/j.proeng.2016.07.514 Rodrigues, E. R., Oliveira, I., Cunha, R., et al., 2018. DeepDownscale: A Deep Learning Strategy for High-Resolution Weather Forecast. 2018 IEEE 14th International Conference on e-Science (e-Science)., Amsterdam. https://doi.org/10.1109/eScience.2018.00130 Sachindra, D. A., Kanae, S., 2019. Machine Learning for Downscaling: The Use of Parallel Multiple Populations in Genetic Programming. Stochastic Environmental Research and Risk Assessment, 33(8): 1497-1533. https://doi.org/10.1007/s00477-019-01721-y Schoof, J. T., Pryor, S. C., 2001. Downscaling Temperature and Precipitation: A Comparison of Regression-Based Methods and Artificial Neural Networks. International Journal of Climatology, 21(7): 773-790. https://doi.org/10.1002/joc.655 Sharma, S. C. M., Kumar, B., Mitra, A., et al., 2024. Deep Learning-Based Bias Correction of ISMR Simulated by GCM. Atmospheric Research, 309: 107589. https://doi.org/10.1016/j.atmosres.2024.107589 Sillmann, J., Kharin, V. V., Zhang, X., et al., 2013. Climate Extremes Indices in the CMIP5 Multimodel Ensemble: Part 1. Model Evaluation in the Present Climate. Journal of Geophysical Research: Atmospheres, 118(4): 1716-1733. https://doi.org/10.1002/jgrd.50203 Sun, L., Lan, Y. F., 2021. Statistical Downscaling of Daily Temperature and Precipitation over China Using Deep Learning Neural Models: Localization and Comparison with Other Methods. International Journal of Climatology, 41(2): 1128-1147. https://doi.org/10.1002/joc.6769 Taylor, K. E., Stouffer, R. J., Meehl, G. A., 2012. An Overview of CMIP5 and the Experiment Design. Bulletin of the American Meteorological Society, 93(4): 485-498. https://doi.org/10.1175/bams-d-11-00094.1 Vandal, T., Kodra, E., Ganguly, A. R., 2019. Intercomparison of Machine Learning Methods for Statistical Downscaling: The Case of Daily and Extreme Precipitation. Theoretical and Applied Climatology, 137(1): 557-570. https://doi.org/10.1007/s00704-018-2613-3 Wang, F., Tian, D., 2022. On Deep Learning-Based Bias Correction and Downscaling of Multiple Climate Models Simulations. Climate Dynamics, 59(11): 3451-3468. https://doi.org/10.1007/s00382-022-06277-2 Wang, F., Tian, D., Lowe, L., et al., 2021. Deep Learning for Daily Precipitation and Temperature Downscaling. Water Resources Research, 57(4): e2020WR029308. https://doi.org/10.1029/2020WR029308 Wang, L., Chen, W., 2014. A CMIP5 Multimodel Projection of Future Temperature, Precipitation, and Climatological Drought in China. International Journal of Climatology, 34(6): 2059-2078. https://doi.org/10.1002/joc.3822 Wilby, R. L., Wigley, T. M. L., Conway, D., et al., 1998. Statistical Downscaling of General Circulation Model Output: A Comparison of Methods. Water Resources Research, 34(11): 2995-3008. https://doi.org/10.1029/98WR02577 Wood, A. W., Leung, L. R., Sridhar, V., et al., 2004. Hydrologic Implications of Dynamical and Statistical Approaches to Downscaling Climate Model Outputs. Climatic Change, 62(1): 189-216. https://doi.org/10.1023/B: CLIM.0000013685.99609.9e doi: 10.1023/B:CLIM.0000013685.99609.9e Wu, J., Gao, X. J., 2013. A Gridded Daily Observation Dataset over China Region and Comparison with the Other Datasets. Chinese Journal of Geophysics, 56(4): 1102-1111 (in Chinese with English abstract). Yang, C. L., Wang, N. L., Wang, S. J., et al., 2018. Performance Comparison of Three Predictor Selection Methods for Statistical Downscaling of Daily Precipitation. Theoretical and Applied Climatology, 131(1): 43-54. https://doi.org/10.1007/s00704-016-1956-x Yang, W., Andréasson, J., Phil Graham, L., et al., 2010. Distribution-Based Scaling to Improve Usability of Regional Climate Model Projections for Hydrological Climate Change Impacts Studies. Hydrology Research, 41(3-4): 211-229. https://doi.org/10.2166/nh.2010.004 Zhang, Q., Wu, G. X., 2001. The Large Area Flood and Drought over Yangtze River Valley and Its Relation to the South Asia High. Acta Meteorologica Sinica, 59(5): 569-577 (in Chinese with English abstract). Zhang, Z. S., Li, S. L., Wang, H. J., et al., 2022. Introduction of Crossing Disciplines between Geology and Atmospheric Science. Earth Science, 47(10): 3569-3579 (in Chinese with English abstract). 陈杰, 许崇育, 郭生练, 等, 2016. 统计降尺度方法的研究进展与挑战. 水资源研究, 5(4): 299-313. 刘杰, 陈前, 许妍, 等, 2024. 长江流域洞庭湖区出入湖磷通量模拟及水质预测: 机器学习与传统水文模型耦合方法. 地球科学, 49(11): 3995-4007. doi: 10.3799/dqkx.2024.061 刘永和, 郭维栋, 冯锦明, 等, 2011. 气象资料的统计降尺度方法综述. 地球科学进展, 26(8): 837-847. 吴佳, 高学杰, 2013. 一套格点化的中国区域逐日观测资料及与其它资料的对比. 地球物理学报, 56(4): 1102-1111. 张琼, 吴国雄, 2001. 长江流域大范围旱涝与南亚高压的关系. 气象学报, 59(5): 569-577. 张仲石, 李双林, 王会军, 等, 2022. 浅谈大气科学与地质学的学科交叉. 地球科学, 47(10): 3569-3579. doi: 10.3799/dqkx.2022.350 -
点击查看大图
计量
- 文章访问数: 175
- HTML全文浏览量: 8
- PDF下载量: 17
- 被引次数: 0
