| The dynamic changes of soil temperature affect the soil physical and chemical process as well as the global climatic change.Soil temperatures,which were observed at meteorological stations,are weak at spatial continuity and limited at distribution quantity in the complex terrain,so they are difficult to represent the soil temperature dynamic changes of whole region.Consequently,simulating soil temperature at regional scale accurately and accessing its spatio-temporal dynamic changes precisely have become global issues,acquired growing attention and entered a period of increasing maturity.Based on the observed data gathered during 1981-2012 from local meteorological stations,exploration was done of soil temperature regimes,using classical statistics of climatic diagnosis,over the past 32 years in Liangshan Yi Autonomous Prefecture and Panzhihua lying on the south of Sichuan Province,China.Moreover,combined with remote sensing technology,SPSS and ArcGIS software,the simulation models of soil temperature were constructed.By means of multiple stepwise regression method with embedding environmental factors and the exponential function.Finally,according to the analog values,characteristics of soil temperature spatio-temporal dynamic changes were analyzed.The results suggestas follows:(1)Statistical characteristics of soil temperature in the study areaOn the mensal scale,soil surface temperature,which obtained from 8 selected weather stations,ranged from 5.2 to 30.8℃.Following cosine curve,soil surface temperature had the highest and the lowest value.The highest temperature appeared during the period May to August,and the lowest temperature occurred in December and January.On the annual scale,soil surface temperature had a significant increasingtrend(P<0.05).In 8 sites,meteorological stations that located at Tibetan Autonomous County of Muli and Leibo County had the highest increasing rates(1.54 and 1.03℃/10a)which are 2-7 times of other areas.However,meteorological station that located at Yuexi County had the lowest increasing rate(0.21℃/10a).In addition,the abrupt changes of mean annual soil surface temperature occurred during 1993-2004,after which the mean value of soil temperature raised up significantly.On vertical space,soil surface temperature(0 cm)and deep soil temperature(5-80 cm)had an extremely significant correlation(P<0.01)and explicit regular pattern.Vertical soil temperature in January,October,November and December entered the radiation cooling period.Vertical soil temperature during the period March to August restored solar radiation.In February and September,soil temperature decreased at 0-40 cm and increased at 40-80 cm,belonging to transition duration.(2)Construction of simulation model for soil temperature predictionThe simulation model of soil temperature which taken observed values obtained from weather stations as true values,combing the thermal factor,vegetation factor and terrain factor as auxiliary variables,using multiple stepwise regression method estimate the soil surface temperature,and using the exponential function as modeling method estimate the depth soil temperature in study area accurately.Among three influencing factors,LST(land surface temperature)contributed most to modeling,and was able to explain more than 60%of soil temperature spatial variability.In addition,the NDVI(normalized difference vegetation index)and S(slope)were used to improve the fitting degree of the model by 5.3%-20.9%in corresponding months and year.The accuracy assessment of model showed that,MAE(mean absolute error)and RMSE(root mean squared error)of soil temperature at 0 cm were lower than 1.5℃,and MRE(mean relative error)of soil temperature at 0 cm was less than 9%;MAE and RMSE of soil temperature at 5-80 cm were ranged from 0.096-1.642℃,and MRE of soil temperature at 5-80 cm was less than 10%,inferring that the simulation results were reliable.(3)Spatio-temporal dynamic changes of simulated soil temperature in the study areaUsing the spatial prediction model of soil temperature with the help of ArcGIS,spatial distribution of soil temperature from 2000 to 2012 with the depth of 0-80 cm through the whole study area was achieved.Soil surface temperature in Liangshan and Panzhihua region was higher in the southern part and river valley as well as was lower in the northern part and plateau.The lowest soil temperature occurred in January,which ranged from-7.6-17.49℃.The highest temperature appeared during the period May to July,which ranged from 16.2-31.2℃.During 2000-2012,74.25%of the soil temperatures across the study area had no significant trend,23.18%of them had significant increasing trend(P<0.05),2.02%of them had extremely significant increasing trend(P<0.01),and 0.55%of them had significant decreasing trend(P<0.05)or extremely significant decreasing trend(P<0.01).In addition,Panzhihua had an stronger warming trend than Liangshan Yi Autonomous Prefecture.On vertical space,the mean mensal temperature varied from 1.8 to 4.5℃,whereas the mean annual temperature was 1.4℃.Based on the regime of observed soil temperature with RS and GIS technology,this study rationally chose the influence factors of spatial distribution of soil temperature in the complex terrain,and used multiple stepwise regression method and the exponential function to build a simulation model for soil temperature prediction in a complex topography region,and finally obtained a high accurate spatio-temporal information of soil temperature.Using the spatial simulation model could obtain soil temperature,through the whole study area,from 2000 to 2012 with the depth of 0-80 cm.In addition,the space resolution of predicted soil temperature way 1km.This study provides basic information for local climate research and agricultural production,and also provides an efficient approach to explore soil thermal physics and study the climate system. |
PDF Full Text Request