2:15 PM - 2:30 PM
[MIS17-02] Reconstruction of Precipitation Rainfall in Modern China from Qing Dynasty Diaries: Using the Modified Green Ampt Model
Keywords:old diaries, soil moisture
Reconstructing past climates is important for understanding future climate variables, and many efforts have been made to estimate paleoclimate using diaries and official documents. In this study, we attempt to reconstruct rainfall in Gansu Province, northwest China, by using Yu-Fen-Cun (soil infiltration depth data after rainfall) officially recorded during the Qing Dynasty (1747-1860) at 92 sites.
Several previous studies have attempted to reconstruct rainfall using Yu-Fen-Cun. However, all of them are based on field observations and infiltration tests conducted in the present day, and the relationship between rainfall, initial soil moisture content, and infiltration depth is determined by regression analysis and applied to the infiltration depth data. However, conducting observations and infiltration tests costs, and it is not realistic to conduct field experiments at all observation points. In addition, the number of samples for regression analysis is not abundant, and this regression model varies. Therefore, a robust physical model across observation points and over time to reproduce rainfall is required.
In this study, the Green and Ampt model (GA model), the first theoretical analysis of infiltration, is modified to reproduce rainfall in the arid zone, the target area of this study. The GA model tends to overestimate the infiltration depth, and there have been various corrections. In this study, the correction is based on the method of Gill (1978), which assumes that the permeability is a vertical average of the transmit zone and the unsaturated zone. Furthermore, based on the observed fact that the transmit zone is not saturated, the moisture content of the transmit zone was also taken as a parameter.
The validation of this model is based on an artificial rainfall experiment conducted in northern China in 2008, in which rainfall of 4.06~14.93 mm fell for 10 minutes, and the correlation coefficient was found to be 0.86.
Using this improved physical model, the rainfall for the period 1801~1850, when infiltration depth data existed for more than 180 days per year, at 92 sites where infiltration depth data existed.
First, sensitivity experiments were conducted. The model parameters, initial water content, and moisture content in the transmit zone were the experimental values in northwestern China described above and varied from the minimum to the maximum values of the same experiment. The sensitivity of the model to the difference between the moisture content of the transmit zone and the initial soil moisture content was low, and all simulation results were within a range of ±10% of the average calculated value. The sensitivity was similarly low for initial moisture content, except when the value was less than 0.05. This indicates that the model is robust to the input parameters except for extreme dryness.
Next, rainfall was reproduced using the average values obtained from the experiment. The average annual rainfall at each site ranged from 0.1 to 121 mm, and the average rainfall for all sites was 46 mm. The average monthly rainfall at all stations was 0 mm in December ~ February, with a peak in July ~ September, consistent with the current climate trend in the arid zone of northwestern China. In terms of time-series variation, rainfall exceeding 500 mm was reproduced at Koran in 1813. Ge et al. (2004) also reported that the heaviest rainfall of more than 1000 m was reproduced at Shijiazhuang, also in the 50 years from 1801 to 1850. The spatial distribution of rainfall averaged 19.72 mm west of 100° longitude and 88 mm east of 100° longitude. This is consistent with the distribution in modern China, where precipitation increases from northwest to southeast.
Several previous studies have attempted to reconstruct rainfall using Yu-Fen-Cun. However, all of them are based on field observations and infiltration tests conducted in the present day, and the relationship between rainfall, initial soil moisture content, and infiltration depth is determined by regression analysis and applied to the infiltration depth data. However, conducting observations and infiltration tests costs, and it is not realistic to conduct field experiments at all observation points. In addition, the number of samples for regression analysis is not abundant, and this regression model varies. Therefore, a robust physical model across observation points and over time to reproduce rainfall is required.
In this study, the Green and Ampt model (GA model), the first theoretical analysis of infiltration, is modified to reproduce rainfall in the arid zone, the target area of this study. The GA model tends to overestimate the infiltration depth, and there have been various corrections. In this study, the correction is based on the method of Gill (1978), which assumes that the permeability is a vertical average of the transmit zone and the unsaturated zone. Furthermore, based on the observed fact that the transmit zone is not saturated, the moisture content of the transmit zone was also taken as a parameter.
The validation of this model is based on an artificial rainfall experiment conducted in northern China in 2008, in which rainfall of 4.06~14.93 mm fell for 10 minutes, and the correlation coefficient was found to be 0.86.
Using this improved physical model, the rainfall for the period 1801~1850, when infiltration depth data existed for more than 180 days per year, at 92 sites where infiltration depth data existed.
First, sensitivity experiments were conducted. The model parameters, initial water content, and moisture content in the transmit zone were the experimental values in northwestern China described above and varied from the minimum to the maximum values of the same experiment. The sensitivity of the model to the difference between the moisture content of the transmit zone and the initial soil moisture content was low, and all simulation results were within a range of ±10% of the average calculated value. The sensitivity was similarly low for initial moisture content, except when the value was less than 0.05. This indicates that the model is robust to the input parameters except for extreme dryness.
Next, rainfall was reproduced using the average values obtained from the experiment. The average annual rainfall at each site ranged from 0.1 to 121 mm, and the average rainfall for all sites was 46 mm. The average monthly rainfall at all stations was 0 mm in December ~ February, with a peak in July ~ September, consistent with the current climate trend in the arid zone of northwestern China. In terms of time-series variation, rainfall exceeding 500 mm was reproduced at Koran in 1813. Ge et al. (2004) also reported that the heaviest rainfall of more than 1000 m was reproduced at Shijiazhuang, also in the 50 years from 1801 to 1850. The spatial distribution of rainfall averaged 19.72 mm west of 100° longitude and 88 mm east of 100° longitude. This is consistent with the distribution in modern China, where precipitation increases from northwest to southeast.