Journal of Geographical Sciences >
Attribution of trends in meteorological drought during 19602016 over the Loess Plateau, China
Guo Mengyao (1994), PhD Candidate, specialized in atmosphereecosystem interactions. Email: mengyao_guo@whu.edu.cn 
Received date: 20200922
Accepted date: 20210510
Online published: 20211025
Supported by
National Natural Science Foundation of China(41877159)
The National Key Research and Development Program of China(2017YFA0603704)
The Scholarship from China Scholarship Council(CSC), No(201906270109)
This study uses two forms of the Palmer Drought Severity Index (PDSI), namely the PDSI_TH (potential evapotranspiration estimatedby the Thornthwaite equation) and the PDSI_PM (potential evapotranspiration estimated by the FAO PenmanMonteith equation), to characterize the meteorological drought trends during 19602016 in the Loess Plateau (LP) and its four subregions. By designing a series of numerical experiments, we mainly investigated various climatic factors' contributions to the drought trends at annual, summer, and autumn time scales. Overall, the drying trend in the PDSI_TH is much larger than that in the PDSI_PM. The former is more sensitive to air temperature than precipitation, while the latter is the most sensitive to precipitation among all meteorological factors. Increasing temperature results in a decreasing trend (drying) in the PDSI_TH, which is further aggravated by decreasing precipitation, jointly leading to a relatively severe drying trend. For the PDSI_PM that considers more comprehensive climatic factors, the drying trend is partly counteracted by the declining wind speed and solar radiation. Therefore, the PDSI_PM ultimately shows a much smaller drying trend in the past decades.
GUO Mengyao , SHE Dunxian , ZHANG Liping , LI Lingcheng , YANG ZongLiang , HONG Si . Attribution of trends in meteorological drought during 19602016 over the Loess Plateau, China[J]. Journal of Geographical Sciences, 2021 , 31(8) : 1123 1139 . DOI: 10.1007/s114420211888y
Figure 1 Location of the Loess Plateau and the 79 meteorological stations and the distribution of the four subregions based on Kmeans clustering 
Table 1 Framework for the numerical experiments 
Case  Denotation  Computation 

Original  PDSI_TH_Org  PDSI_TH calculated on all original climatic variables 
PDSI_PM_Org  PDSI_PM calculated on all original climatic variables  
Base  PDSI_TH_Base  PDSI_TH calculated on all detrended climatic variables 
PDSI_PM_Base  PDSI_PM calculated on all detrended climatic variables  
P  PDSI_TH_P  PDSI_TH calculated on original P and detrended T 
PDSI_PM_P  PDSI_PM calculated on original P and detrended T, WS, R_{s}, and RH  
T  PDSI_TH_T  PDSI_TH calculated on original T and detrended P 
PDSI_PM_T  PDSI_PM calculated on original T and detrended P, WS, R_{s}, and RH  
WS  PDSI_PM_WS  PDSI_PM calculated on original WS and detrended P, T, R_{s}, and RH 
R_{s}  PDSI_PM_R_{s}  PDSI_PM calculated on original R_{s} and detrended P, T, WS, and RH 
RH  PDSI_PM_RH  PDSI_PM calculated on original RH and detrended P, T, WS, and R_{s} 
Table 2 Station numbers and climatic condition in each subregion 
Stations  P (mm)  PET (mm)  AI (P/PET)  

SR1  26  337.8  554.4  0.61 
SR2  13  433.5  429.6  1.01 
SR3  13  392.2  515.2  0.76 
SR4  27  591.7  641.2  0.92 
Figure 2 Time series of P, T, WS, R_{s}, RH, and PET at annual, summer, and autumn time scales for the averages of the whole LP and each subregion (19602016). The numbers in the top corner are the changing trends fitted by the linear regression model, with the different colors corresponding to different regions according to the legend. Underlined numbers indicate the trends have passed the Student's ttest of p < 0.05. 
Table 3 Pearson's correlation coefficient (r) between the PDSI_TH and PDSI_PM time series during 19602016 (all values have passed the Student's ttest of p < 0.001) 
Annual  Summer  Autumn  

LP  0.870  0.871  0.901 
SR1  0.897  0.895  0.926 
SR2  0.769  0.775  0.859 
SR3  0.930  0.928  0.929 
SR4  0.887  0.888  0.903 
Figure 3 Time series of the PDSI_TH, PDSI_PM and PDSI_TH minus the PDSI_PM at annual, summer, and autumn time scales for the averages of the whole LP and each subregion (19602016). The numbers in the top right corner are the changing trends fitted by the linear regression model, with the different colors corresponding to different PDSI according to the legend. Underlined numbers indicate the trends have passed the Student's ttest of p < 0.05. 
Figure 4 Fitting results between Cal_∂PDSI/∂t (the calculated PDSI trend, the sum of individual contributions of each climatic factor) and Obs_∂PDSI/∂t (the observed PDSI trend, the result of the PDSI_Org trend minus the PDSI_Base trend) from the 79 meteorological stations. The r values in the top left corner are the Pearson's correlation coefficients between Cal_∂PDSI/∂t and Obs_∂PDSI/∂t. 
Figure 5 The PDFs of the PDSI trends in the base case and the climatic factors' cases from the 79 meteorological stations. The numbers in the top right corner are the mean and standard deviation values, with the different colors indicating different experimental cases corresponding to the legend. 
Figure 6 The individual contributions of climatic factors and the Obs_∂PDSI/∂t (the observed PDSI trend, the result of the PDSI_Org trend minus the PDSI_Base trend) over the LP as a whole and in each subregion 
[1] 

[2] 

[3] 

[4] 

[5] 

[6] 

[7] 

[8] 

[9] 

[10] 

[11] 

[12] 

[13] 

[14] 

[15] 

[16] 

[17] 

[18] 

[19] 

[20] 

[21] 

[22] 

[23] 

[24] 

[25] 

[26] 

[27] 

[28] 

[29] 

[30] 

[31] 

[32] 

[33] 

[34] 

[35] 

[36] 

[37] 

[38] 

[39] 

[40] 

[41] 

[42] 

[43] 
Van Der Schrier G,

[44] 

[45] 

[46] 

[47] 

[48] 

[49] 

[50] 

[51] 

[52] 

[53] 

[54] 

/
〈  〉 