AN NWP PRECIPITATION PRODUCTS DOWNSCALING METHOD BASED ON DEEP LEARNING
-
摘要: 提出一种基于深度学习的数值模式降水产品降尺度方法。利用深度学习的非线性映射能力和对栅格数据的信息提取能力，建立深度超分辨率模型提取不同分辨率数值模式降水产品间相对应的有效信息，从而将低分辨率数值模式降水产品利用提取的信息重构为高分辨率产品，继而通过构建多时次组合降尺度深度模型提取时间关联性进一步提升了重构准确性。基于欧洲中期天气预报中心不同尺度数值模式降水产品的实验表明所提方法能够比常用的双三次插值方法更有效地将低分辨率降水产品转换为对应的高分辨率产品。Abstract: A downscaling method of the Numerical Weather Prediction (NWP) precipitation products based on deep learning is proposed. Based on nonlinear mapping and information extraction of the deep learning network, a deep super-resolution model is established. The model is used to extract the effective corresponding information between low-resolution products and high-resolution products. Then high-resolution products can be reconstructed based on corresponding low-resolution products. Furthermore, we propose a multi-time combination downscaling deep learning model to extract time correlations, and the accuracy is further improved. Several experiments based on different scales of the ECMWF (European Centre for Medium-Range Weather Forecasts) NWP precipitation products are conducted. The results show that the proposed methods can be more effective in refining the resolution of NWP precipitation products than the commonly used method based on bi-cubic interpolation.
-
图 10 2019年7月5日15时降水产品降尺度效果
a. 1 °×1 °低分辨率产品；b.双三次插值得到的高分辨率产品；c.基于深度学习的方法得到的高分辨率产品；d.实际0.25 °×0.25 °高分辨率产品。
图 11 2019年7月15日00点降水产品降尺度效果
a. 0.5 °×0.5 °低分辨率产品；b.双三次插值得到的高分辨率产品；c.基于深度学习的方法得到的高分辨率产品；d.实际0.25 °×0.25 °高分辨率产品。
表 1 0.5 °×0.5 °分辨率产品向0.25 °×0.25 °分辨率产品
采用方法 MAE/mm MSE/mm2 多时次组合模型 0.208 0.160 单时次模型 0.216 0.175 双三次插值 0.388 0.498 
下载: 导出CSV
表 2 1 °×1 °分辨率产品向0.25 °×0.25 °分辨率产品转换
采用方法 MAE/mm MSE/mm2 多时次组合模型 0.471 0.753 单时次模型 0.494 0.819 双三次插值 0.553 0.985
下载: 导出CSV
-
[1] 刘永和, 郭维栋, 冯锦明, 等.气象资料的统计降尺度方法综述[J].地球科学进展, 2011, 26(8):837-847. [2] OSHIMA N, KATO H, KADOKURA S. An application of statistical downscaling to estimate surface air temperature in Japan[J]. J Geophys Res Atmos, 2002, 107(D10):ACL-1-ACL 14-10. [3] MPELASOKA F S, MULLAN A B, HEERDEGEN R G. New Zealand climate change information derived by multivariate statistical and artificial neural networks approaches[J]. Int J Climatol, 2001, 21(11):1 415-1 433. [4] ANANDHI A, SRINIVAS V V, KUMAR D N, et al. Role of predictors in downscaling surface temperature to river basin in India for IPCC SRES scenarios using support vector machine[J]. Int J Climatol, 2009, 29(4):583-603. [5] LIAN S M, WANG K, CUI M C, et al. Exercise in downscaling on sea surface temperature along Chinese coast[J]. Chinese Journal of Oceanology & Limnology, 2000, 18(2):110-116. [6] CUI B, TOTH Z, ZHU Y, et al. Bias correction for global ensemble forecast[J]. Wea Forecasting, 2012, 27(2): 396-410. [7] 李新, 程国栋, 卢玲.空间内插方法的比较[J].地球科学进展, 2000, 25(13) : 261-265 [8] 彭彬, 周艳莲, 高苹, 等.气温插值中不同空间插值方法的适用性分析——以江苏为例[J].地球信息科学学报, 2011, 13(4) : 539-547. [9] 徐振亚, 任福民, 杨修群, 等.日最高温度统计降尺度方法的比较研究[J].气象科学, 2012, 32(4):395-402. [10] KEYS R G. Cubic convolution interpolation for digital image processing[J]. IEEE Trans.on Acoust.speech.Signal Processing, 1981, 29(6): 1 153-1 160. [11] DONG C, LOY C C, HE K, et al. Image super-resolution using deep convolutional networks[J]. Pattern Analysis & Machine Intelligence IEEE Transactions on, 2016, 38(2):295-307. [12] LEGID C, THEIS L, HUSZARF, et al. Photo-realistic single image super-resolution using a generative adversarial network[C]. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, 2017: 105-114. [13] KIM J, LEE J K, Lee K M. Accurate image super-resolution using very deep convolutional networks[C].IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, 2016: 1 646-1 654. [14] MAO X J, SHEN C, YANG Y B. Image restoration using very deep convolutional encoder-decoder networks with symmetric skip connections[C]. Advances in Neural Information Processing Systems, 2016: 2 802-2 810. [15] SHI W, CABALLERO J, HUSZAR F, et al. Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network[C]. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 1 874-1 883. [16] LIM B, SON S, KIM H, et al. Enhanced deep residual networks for single image super-resolution[C]. IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2017: 1 132-1 140. [17] 姜丽黎, 余晖.基于动力相似方法的台风极端降水概率预报研究[J].热带气象学报, 2019, 35(3):353-364. [18] 李欣韵, 余锦华, 梁信忠.基于CWRF模式夏季日降水量的订正技术对比[J].热带气象学报, 2019, 35(6):842-851. [19] LECUN Y, BOTTOU L. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2 278-2 324. [20] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]. IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2016: 770-778. [21] NAIR V, HINTON G E. Rectified linear units improve restricted boltzmann machines vinod nair[C]. Proceedings of the 27th International Conference on Machine Learning(ICML), Haifa, Israel, 2010: 807-814. [22] KINGMA D P, BA J. Adam: A method for stochastic optimization[C]. Proceedings of the 3rd International Conference for Learning Representations(ICLR), 2015: 1-15. [23] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]. IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2018: 7 132-7 141. [24] 张阿珍, 刘政林.基于双三次插值算法的图像缩放引擎设计[J].微电子学与计算机, 2007, 24(1): 49-51. -
点击查看大图
图(11) / 表(2)
计量
- 文章访问数: 448
- HTML全文浏览量: 37
- PDF下载量: 74
- 被引次数: 0
粤公网安备 4401069904700003号