Page 83 - 《高原气象》2026年第2期
P. 83
2 期 黄嘉雯等:基于机器学习的青海湖水位变化模拟研究 383
赖性较高。特别是 RF, 在关键变量缺失时精度显 126086.
著降低, 这也使其特别适合用于变量相对重要性的 Hochreiter S, Schmidhuber J, 1997. Long short-term memory[J].
Neural Computation, 9: 1735-1780.
排序分析, 但预测精度远不及 LSTM 稳定。MLR
Hou P, Weidman R P, Liu Q, et al, 2023. Recent water-level fluctua‐
体现出极强的变量敏感性和较低的鲁棒性。因此,
tions, future trends and their eco-environmental impacts on Lake
实际应用中可优先选用 LSTM 模型, 并结合上述关 Qinghai[J]. Journal of Environmental Management, 333: 117461.
键特征变量以实现较高精度预测, 同时有效降低计 Jain S K, Gupta A K, 2023. Investigation of multilayer perceptron re‐
算复杂度。 gression-based models to forecast reference Evapotranspiration
(ETo)[J]. Research Reports on Computer Science, 62(1):
(3) 预测结果表明, 2017 -2030 年青海湖水位
104-116.
将会上升 2. 55 m, 水位的上升在短期内有助于青海
Kadam A K, Wagh V M, Muley A A, et al, 2019. Prediction of water
湖的保护, 但如果持续快速扩张, 可能引发一系列 quality index using artificial neural network and multiple linear re‐
生态环境问题。建议相关部门加强对生态因素的 gression modelling approach in Shivganga River basin, India[J].
监测, 持续推进生态综合治理工程, 确保青海湖流 Modeling Earth Systems and Environment, 5: 951-962.
域生态环境的可持续发展。 Kim D K, Lee D I, 2015. Atmospheric thickness and vertical structure
properties in wintertime precipitation events from microwave radi‐
参考文献(References): ometer, radiosonde and wind profiler observations[J]. Meteoro‐
logical Applications, 22: 599-609.
Arnold J G, Srinivasan R, Muttiah R S, et al, 1998. Large area hy‐ Lenters J D, Kratz T K, Bowser C J, 2005. Effects of climate variabil‐
drologic modeling and assessment part I: model development 1 ity on lake evaporation: Results from a long-term energy budget
[J]. Journal of the American Water Resources Association, 34 study of Sparkling Lake, northern Wisconsin (USA)[J]. Journal
(1): 73-89. of Hydrology, 308: 168-195.
Blenckner T, Järvinen M, Weyhenmeyer G A, 2004. Atmospheric cir‐ Liu W, Fedorov A V, 2019. Global impacts of Arctic sea ice loss me‐
culation and its impact on ice phenology in Scandinavia[J]. Bore‐ diated by the Atlantic meridional overturning circulation[J]. Geo‐
al Environment Research, 9: 371-380. physical Research Letters, 46: 944-952.
Box G E P, Jenkins G M, Reinsel G C, et al, 1978. Time series anal‐ Ma S, Liao J, Jing R, et al, 2024. A dataset of lake level changes in
ysis: forecasting and control[J]. The Statistician, 27: 265-265. China between 2002 and 2023 using multi-altimeter data[J]. Big
Buechel M, Knoll A, 2018. Deep reinforcement learning for predic‐ Earth Data, 8: 166-188.
tive longitudinal control of automated vehicles[C]. 2018 21st In‐ Ma Y, Li X Y, Liu L, et al, 2018. Measurements and modeling of
ternational Conference on Intelligent Transportation Systems the water budget in semiarid high‐altitude Qinghai Lake Basin,
(ITSC): 2391-2397. northeast Qinghai‐Tibet Plateau[J]. Journal of Geophysical Re‐
Cannon A J, Whitfield P H, 2002. Downscaling recent streamflow search: Atmospheres, 123: 10857-10871.
conditions in British Columbia, Canada using ensemble neural Mosavi A, Ozturk P, Chau K W, 2018. Flood prediction using ma‐
network models[J]. Journal of Hydrology, 259: 136-151. chine learning models: literature review[J]. Water, 10: 1536.
Chen T, Guestrin C, 2016. XGBoost: a scalable tree boosting system Nash J E, Sutcliffe J V, 1970. River flow forecasting through concep‐
[C]. Proceedings of the 22nd ACM SIGKDD International Con‐ tual models part I-A discussion of principles[J]. Journal of Hy‐
ference on Knowledge Discovery and Data Mining. drology, 10: 282-290.
Chen Y, Deng H, Li B, et al, 2014. Abrupt change of temperature Pedregosa F, Varoquaux G, Gramfort A, et al, 2012. Scikit-learn:
and precipitation extremes in the arid region of Northwest China machine learning in python[J]. Journal of Machine Learning Re‐
[J]. Quaternary International, 336: 35-43. search, 12.
Cortes C, Vapnik V N, 1995. Support-vector networks[J]. Machine Pitman A J, Perkins S E, 2009. Global and regional comparison of dai‐
Learning, 20: 273-297. ly 2-m and 1000-hPa maximum and minimum temperatures in
Ghorbani M A, Deo R C, Karimi V, et al, 2018. Design and imple‐ three global reanalyses[J]. Journal of Climate, 22: 4667-4681.
mentation of a hybrid MLP-GSA model with multi-layer percep‐ Rasouli K, Hsieh W W, Cannon A J, 2012. Daily streamflow forecast‐
tron-gravitational search algorithm for monthly lake water level ing by machine learning methods with weather and climate inputs
forecasting[J]. Stochastic Environmental Research and Risk As‐ [J]. Journal of Hydrology, 414: 284-293.
sessment, 33: 125-147. Singh A P, Imtiyaz M, Isaac R K, et al, 2012. Comparison of soil
Hagen J S, Leblois É, Lawrence D, et al, 2021. Identifying major and water assessment tool (SWAT) and multilayer perceptron
drivers of daily streamflow from large-scale atmospheric circula‐ (MLP) artificial neural network for predicting sediment yield in
tion with machine learning[J]. Journal of Hydrology, 596: the Nagwa agricultural watershed in Jharkhand, India[J]. Agri‐

