Downscaling of Air Temperature in the Yarlung Zangbo River Basin Based on the Random Forest Model

  • REN Meifang ,
  • PANG Bo ,
  • XU Zongxue ,
  • ZHAO Yanjun
Expand
  • Key Laboratory for Water and Sediment Science, Ministry of Education, College of Water Sciences, Beijing Normal University/Beijing Key Laboratory of Urban Hydrological Cycle and Sponge City Technology, Beijing 100875, China

Received date: 2017-11-11

  Online published: 2018-10-28

Abstract

Random forest (RF) model was used to downscale the daily air temperature at 20 meteorological stations in and around the Yarlung Zangbo River basin.For the purpose to explore the better downscaling method of air temperature in the study area, three methods were used to compare, namely, the Multiple Linear Regression (MLR), Artificial Neural Network (ANN), and Support Vector Machine (SVM).Principal Component Analysis (PCA) and Partial Correlation Analysis (PAR) were used to select the characteristic variables.The model performance was assessed using four criteria, namely, the NASH coefficient of efficiency (NASH), the root mean squared error (RMSE), the mean absolute error (MAE), and the coefficient of correlation (r).The results showed that the performance of RF model was obviously better than other models; the results obtained by PAR to select characteristic variables were not only better than those used by PCA method, but also more stable.In addition, the NASH efficiency coefficients of various models were all above 0.86 and the correlation coefficients were all above 0.93 in the validation periods.Therefore, all of models used in this study can well simulate the average temperature in the Yarlung Zangbo River basin.The experimental data of two typical extreme concentration paths(Representative Concentration Pathway, RCP)emission scenarios RCP2.6 and RCP8.5 of MPI-ESM-LR model in the future (2016-2050)were chosen to study the future trend of temperature in the Yarlung Zangbo River basin.The results showed increasing trend of daily air temperature both under RCP2.6 and RCP8.5 scenarios in the future years of 2016-2050 in the Yarlung Zangbo River basin.The daily air temperature will increase by 0.14℃ under RCP2.6 scenario, and 0.30℃ under RCP8.5 scenario from 2016 to 2050.

Cite this article

REN Meifang , PANG Bo , XU Zongxue , ZHAO Yanjun . Downscaling of Air Temperature in the Yarlung Zangbo River Basin Based on the Random Forest Model[J]. Plateau Meteorology, 2018 , 37(5) : 1241 -1253 . DOI: 10.7522/j.issn.1000-0534.2018.00026

References

[1]Baudron P, Alonso-Sarría F, García-Aróstegui J L, et al, 2013.Identifying the origin of groundwater samples in a multi-layer aquifer system with Random Forest classification[J].J Hydrology, 499:303-315.
[2]Breiman L, 2001.Random Forest[J].Machine Learning, 45(1):5-32.
[3]Duhan D, Pandey A, 2015.Statistical downscaling of temperature using three techniques in the Tons River basin in Central India[J].Theor Appl Climatol, 121(3-4):605-622.
[4]Hewitson B C, Crane R G, 1996.Climate downscaling:Techniques and application[J].Climate Res, 7(2):85-95.
[5]IPCC, 2007.Contribution of working groups Ⅰ, Ⅱ, and Ⅲ to the fourth assessment report of the intergovernment panel on climat change[M].Cambridge and New York:Cambridge University Press.
[6]IPCC, 2013.Summary for policymaker of climate change 2013:The physical science basis.Contribution of working group 1 to the fifth assessment report of the intergovernment panel on climate change[M].Cambridge University Press:Cambridge, UK.
[7]Jing W L, Yang Y P, Yue X F, et al, 2016.A comparison of different regression algorithms for downscaling monthly satellite-based precipitation over north China[J].Remote Sens, 8(10):1-17.
[8]Kharin V V, Zwiers F W, Zhang X, et al, 2013.Changes in temperature and precipitation extremes in the CMIP5 ensemble[J].Climate Change, 119:345-357.
[9]Kostopoulou E, Giannakopoulos C, Anagnostopoul C, et al, 2007.Simulation maximum and minimum temperature over Greece:A comparison of three downscaling techniques[J].Theor Appl Climatol, 90(1/2):65-82.
[10]Li F P, Zhang Y Q, Xu Z X, et al, 2014.Runoff predictions in ungagged catchments in southeast Tibetan Plateau[J].J Hydrology, 511(511):28-38.
[11]Liu T, Chen B D, 2000.Climate warming in the Tibetan Plateau during recent decades[J].Climate, 20:1729-1742.
[12]Malekipirbazari M, Aksakalli V, 2015.Risk assessment in social lending via random forest[J].Expert Systems with Applications, 42(10):4621-4631.
[13]Schoof J T, Pryor S C, 2001.Downscaling temperature and precipitation:a comparison of regression-based methods and artificial neural networks[J].Int J Climatol, 21970:773-790.
[14]Screen J A, 2014.Arctic amplification decreases temperature variance in Northern mid-to high-latitudes[J].Nature Climate Change, 4(7):577-582.
[15]Sharma A, 2000.Seasonal to interannual rainfall probabilistic forecasts for improved water supply management:Part1-A strategy for system predictor identification[J].J Hydrology, 239(1-4):232-239.
[16]Su F G, Duan X L, Chen D L, et al, 2013.Evaluation of the global climate models in the CMIP5 over the Tibetan Plateau[J].J Climate, 26(10):3187-3208.
[17]Tang Q C, Xiong Y, 1998.River hydrology in China[M].Beijing:Science Press.
[18]Tatsumi K, Oizumi T, Yamashiki Y, 2015.Effects of climate change on daily minimum and maximum temperatures and cloudiness in the Shikoku region:a statistical downscaling model approach[J].Theor Appl Climatol, 120:87-98.
[19]Vapnik V N, 1997.The nature of statistical learning theory[M].New York:Springer-Verlag.
[20]Wang Z, Lai C, Chen X, et al, 2015.Flood hazard risk assessment model based on random forest[J].J Hydrology, 527:1130-1141.
[21]Xu J W, Gao Y H, Chen D L, et al, 2017.Evaluation of global climate models for downscaling applications centred over the Tibetan Plateau[J].Int J Climatol, 37(2):657-671.
[22]Yang C L, Wang N L, Wang S J, et al, 2016.Performance comparison of three predictor selection methods for statistical downscaling of daily precipitation[J].Theor Appl Climatol, 1-12.DOI:10.1007/s00704-016-1956-x.
[23]You Q L, Kang S C, Wu Y H, et al, 2007.Climate change over the Yarlung Zangbo River Basin during 1961-2005[J].J Geograp Sci, 17(4):409-420.
[24]Zhu X H, Wang W Q, Fraedrich K, 2013.Future climate in the Tibetan plateau from a statistical regional climate model[J].J Climate, 26(4):10125-10138.
[25]Chang Y, Lü S H, Luo S Q, et al, 2016.Evaluation and projections of permafrost on the Qinghai-Xizang Plateau by CMIP5 coupled climate models[J].Plateau Meteor, 35(5):1157-1168.DOI:10.7522/j.issn.1000-0534.2015.00090.<br/>常燕, 吕世华, 罗斯琼, 等, 2016.CMIP5耦合模式对青藏高原动土变化的模拟和预估[J].高原气象, 35(5):1157-1168.
[26]Gao X, Mao X Z, 2017.Study on non-point source load in river network based on the BP neural network method[J].Water &amp; Wastewater Engineering, 43(1):156-160.<br/>高翔, 毛献忠, 2017.基于BP神经网络方法计算河网区面源负荷[J].给水排水, 43(1):156-160.
[27]Jian J, Yang Z G, Zhuo G, 2012.Temporal-spatial distributions of severe precipitation event during flood period in Tibet in last 30 years[J].Plateau Meteor, 31(2):380-386.<br/>建军, 杨志刚, 卓嘎, 2012.近30年西藏汛期强降水事件的时空变化特征[J].高原气象, 31(2):380-386.
[28]Kang Y, Chen Y F, Gu S H, et al, 2014.Assessment of sustainable utilization of regional water resources based on random forest[J].Water Resource and Power, 32(3):34-38.<br/>康有, 陈元芳, 顾圣华, 等, 2014.基于随机森林的区域水资源可持续利用评价[J].水电能源科学, 32(3):34-38.
[29]Lai C G, Chen X H, Zhao S W, et al, 2015.A flood risk assessment model based on random forest and its application[J].Journal of Hydraulic Engineering, 46(1):58-66.<br/>赖成光, 陈晓宏, 赵仕威, 等, 2015.基于随机森林的洪灾风险评价模型及其应用[J].水利学报, 46(1):58-66.
[30]Li Y L, Zhang Q, Li M, et al, 2015.Using BP neural networks for water level simulation in Poyang lake[J].Resources and Environment in the Yangtze Basin, 24(2):233-240.<br/>李云良, 张奇, 李淼, 等, 2015.基于BP神经网络的鄱阳湖水位模拟[J].长江流域资源与环境, 24(2):233-240.
[31]Liu W F, Xu Z X, Li F P, et al, 2014.Climate change scenarios in the Yarlung Zangbo river basin based on ASD model[J].Plateau Meteor, 33(1):26-36.DOI:10.7522/j.issn.1000-0534.2012.00176<br/>刘文丰, 徐宗学, 李发鹏, 等, 2014.基于ASD统计降尺度的雅鲁藏布江流域未来气候变化情景[J].高原气象, 33(1):26-36.
[32]Lu Q, Wu S H, Zhao D S, 2017.Variations in alpine grassland cover and its correlation with climate variables on the Qinghai-Tibetan in 1982-2013[J].Sci Geograp Sinica, 37(2):292-300.<br/>陆晴, 吴绍洪, 赵东升, 2017.1982-2013年青藏高原高寒草地覆盖变化及与气候之间的关系[J].地理科学, 37(2):292-300.
[33]Ma Y, Jiang Q G, Meng Z G, et al, 2016.Classification of land use in farming area based on random forest algorithm[J].Transactions of the Chinese Society for Agricultural Machinery, 47(1):297-303.<br/>马玥, 姜琦刚, 孟治国, 等, 2016.基于随机森林算法的农耕区土地利用分类研究[J].农业机械学报, 47(1):297-303.
[34]Pan B T, Li J J, 1996.Qing-hai-Tibetan plateau:A driver and amplifier of the global climate change:Ⅲ.The effect of the uplift of Qinghai-Tibetan plateau on climatic changes[J].Journal of Lanzhou University (Natural Sciences), 32(1):108-115.<br/>潘保田, 李吉均, 1996.青藏高原:全球气候变化的驱动机与放大器:Ⅲ.青藏高原隆起对气候变化的影响[J].兰州大学学报:自然科学版, 32(1):108-115.
[35]Ren G Y, Xu M Z, Chu Z Y, et al, 2005.Changes of surface air temperature in China during 1951-2004[J].Climatic Environ Res, 10:717-727.<br/>任国玉, 徐铭志, 初子莺, 等, 2005.近54年中国地面气温变化[J].气候与环境研究, 10:717-727.
[36]Shang K Z, Zhou H, Chen L Y, et al, 2010.Trend and Prediction of climate change in southern Tibet valley-taking Gyangze meteorological station for example[J].Plateau Meteor, 29(5):1111-1118.<br/>尚可政, 周海, 陈录元, 等, 2010.西藏南部河谷气候变化趋势及预测-以江孜站为例[J].高原气象, 29(5):1111-1118.
[37]Wang Y, Wang J, Yao Y B, et al, 2014.Evaluation of drought vulnerability in southern china based on principle component analysis[J].Ecology Environ Sci, 23(12):1897-1904.<br/>王莺, 王静, 姚玉璧, 等, 2014.基于主成分分析的中国南方干旱脆弱性评价[J].生态环境学报, 23(12):1897-1904.
[38]Wu Q, Jiang X W, Xie J, 2017.Evaluation of surface air temperature in Southwestern China simulated by the CMIP5 models[J].Plateau Meteor, 36(2):358-370.DOI:10.7522/j.issn.1000-0534.2016.00046.<br/>伍清, 蒋兴文, 谢洁, 2017.CMIP5模式对西南地区气温的模拟能力评估[J].高原气象, 36(2):358-370.
[39]Xu Z X, Zhang N, 2006.Long term trend of precipitation in the Yellow River basin during the past 50 years[J].Geograp Res, 25(1):27-34.<br/>徐宗学, 张楠, 2006.黄河流域近50年降水变化趋势分析[J].地理研究, 25(1):27-34.
[40]Zhang B, Dai X G, 2017.Evaluation and correction for the deviation of the surface air temperature based on 24 CMIP5 models over China for 2006-2015[J].Plateau Meteor, 36(6):1619-1629.DOI:10.7522/j.issn.1000-0534.2016.00136.<br/>张蓓, 戴新刚, 2017.基于CMIP5的2006-2015年中国气温预估偏差分析及订正[J].高原气象, 36(6):1619-1629.
[41]Zhang L, Wang L L, Zhang X D, et al, 2014.The basic principle of random forest and its applications in ecology:A case study of Pinus yunnanensis[J].Acta Ecologica Sinica, 34(3):650-659.<br/>张雷, 王琳琳, 张旭东, 等, 2014.随机森林算法基本思想及其在生态学中的应用-以云南松分布模拟为例[J].生态学报, 34(3):650-659.
[42]Zhao N, Yue T X, Shi W J, et al, 2017.Downscaling simulation of Annual Average Temperature and Precipitation of CMIP5 outputs by using HASM-A case study in Heihe river basin[J].J Desert Res, 37(6):1-10.<br/>赵娜, 岳天祥, 史文娇, 等, 2017.基于HASM方法对气候模式气温降水的降尺度研究-以黑河流域为例[J].中国沙漠, 37(6):1-10.
Outlines

/