Page 144 - 《高原气象》2025年第3期
P. 144
高 原 气 象 44 卷
702
表2 三个模型的三种评估结果 (1): 55. DOI: 10. 3390/su11010055.
Tabel 2 The evaluation results for the three models Feng L, Li T, Yu W D, 2014. Cause of severe droughts in Southwest
are as follows China during 1951-2010[J]. Climate Dynamics, 43: 2033-
2042. DOI: 10. 1007/s00382-013-2026-z.
模型 时间 σ Bias ME
Gao J M, Sang Y H, 2017. Identification and estimation of landslide-
SST-ConvLSTM 2019 1. 333 1. 061 -0. 058
debris flow disaster risk in primary and middle school campuses
2020 1. 329 1. 074 -0. 075 in a mountainous area of Southwest China[J]. International Jour‐
2021 1. 327 0. 972 -0. 154
nal of Disaster Risk Reduction, 25: 60-71. DOI: 10. 1016/j. ij‐
2022 0. 863 0. 707 -0. 049
drr. 2017. 07. 012.
ConvLSTM 2019 1. 660 1. 244 0. 791 Ho H C, Knudby A, Sirovyak P, et al, 2014. Mapping maximum ur‐
2020 1. 744 1. 231 0. 918 ban air temperature on hot summer days[J]. Remote Sensing of
2021 1. 855 1. 328 1. 028 Environment, 154: 38-45. DOI: 10. 1016/j. rse. 2014. 08. 012.
Jiang X W, Shu J C, Wang X, et al, 2017. The roles of convection
2022 1. 120 0. 850 -0. 501
over the western Maritime Continent and the Philippine Sea in in‐
NCC 2019 2. 150 1. 850 0. 692
terannual variability of summer rainfall over southwest China[J].
2020 2. 788 2. 297 1. 137 Journal of Hydrometeorology, 18(7): 2043-2056. DOI: 10.
2021 2. 952 2. 363 0. 915 1175/JHM-D-16-0292. 1.
2022 1. 574 1. 334 -0. 675 Kim S, Hong S, Joh M, et al, 2017. Deeprain: Convlstm network for
precipitation prediction using multichannel radar data[J]. arXiv
黑体突出表示效果最好(the best performance highlighted)
preprint arXiv: 1711. 02316. DOI: 10. 48550/arXiv. 1711. 02316.
结构能够增强模型的非线性映射能力, 从而更精准 Koster R D, Dirmeyer P A, Guo Z C, et al, 2004. Regions of strong
地捕捉降水变化中的复杂特征。同时, 对超参数的 coupling between soil moisture and precipitation[J]. Science, 305
调整, 如学习率和批量大小, 能够平衡模型的训练 (5687): 1138-1140. DOI: 10. 1126/science. 1100217.
Kusiak A, Zheng H Y, Song Z, 2009. Wind farm power prediction: a
速度和稳定性, 确保模型在迭代过程中不断优化。
data‐mining approach[J]. Wind Energy: An International Journal
在数据输入方面, 本文建议可以尝试使用不同分辨
for Progress and Applications in Wind Power Conversion Technol‐
率的海表温度和降水数据, 以探索不同尺度信息对 ogy, 12(3): 275-293. DOI: 10. 1002/we. 295.
预测精度的影响。此外, 通过调整海表温度与降水 Li J Z, Shi Y, Zhang T, et al, 2023. Radar precipitation nowcasting
数据在模型中的权重比例, 可以进一步挖掘两种数 based on ConvLSTM model in a small watershed in north China
据之间的协同作用, 提升模型对降水变化的敏感度 [J]. Natural Hazards, 1-23. DOI: 10. 1007/s11069-023-
和准确性。此外, 考虑到降水变化的局地性特征, 06193-6.
Lim Y, Jo S, Lee J, et al, 2012. Prediction of East Asian summer pre‐
使用与预测区域更为相关的局部海表温度数据也
cipitation via independent component analysis[J]. Asia-Pacific
是一个值得探索的方向。
Journal of Atmospheric Sciences, 48: 125-134. DOI: 10. 1007/
SST-ConvLSTM模型作为创新性的降水预测工 s13143-012-0012-8.
具, 在西南地区的降水预测方面已经展现显著的优 Lin W, Wen C, Wen Z, et al, 2015. Drought in Southwest China: a
势。其在捕捉降水时空变化特征、 预测极端降水事 review[J]. Atmospheric and Oceanic Science Letters, 8(6):
件等方面的能力, 使得该模型在降水研究领域具有 339-344. DOI: 10. 3878/AOSL20150043.
巨大的应用潜力。因此, 未来的研究可深入探索 Liu W, Wang Y Q, Zhong D, et al, 2022. ConvLSTM network-based
rainfall nowcasting method with combined reflectance and radar-
SST-ConvLSTM 模型的性能边界和适用范围, 进一
retrieved wind field as inputs[J]. Atmosphere, 13(3): 411.
步提升其在降水预测中的准确性和可靠性。
DOI: 10. 3390/atmos13030411.
参考文献(References): Ma Z F, Liu J, Zhang S Q, et al, 2013. Observed climate changes in
southwest China during 1961-2010[J]. Advances in Climate
Behrang M A, Assareh E, Ghanbarzadeh A, et al, 2010. The poten‐ Change Research, 4(1): 30-40. DOI: 10. 3724/SP. J. 1248.
tial of different artificial neural network (ANN) techniques in dai‐ 2013. 030.
ly global solar radiation modeling based on meteorological data Mellit A, Pavan A M, Benghanem M, 2013. Least squares support
[J]. Solar Energy, 84(8): 1468-1480. DOI: 10. 1016/j. solen‐ vector machine for short-term prediction of meteorological time
er. 2010. 05. 009. series[J]. Theoretical and Applied Climatology, 111: 297-307.
Chou J M, Xian T, Dong W J, et al, 2018. Regional temporal and DOI: 10. 1007/s00704-012-0661-7.
spatial trends in drought and flood disasters in China and assess‐ Michaelides S, Levizzani V, Anagnostou E, et al, 2009. Precipita‐
ment of economic losses in recent years[J]. Sustainability, 11 tion: Measurement, remote sensing, climatology and modeling