Retrieval Of Land Surface Temperature In Mountainous Area Of Manasi River In Central Tianshan Mountains

Temperature is a dominant factor that affects snowmelt,as well as the most intuitive factor characterizing climate change.Therefore,the spatial-temporal information of temperature is of great importance to snowmelt runoff simulation in cold-high mountainous,water resource management and local climate change research in arid and semi-arid regions.Field-measured temperature at the measured point is accurate,while it is hard to obtain field-measured temperature for large areas,especially in cold-high and inaccessible mountains with bad weather condition and hard working environment,which results in the shortage of field-measured temperature data in cold-high mountains.Along with the development of thermal remote sensing,its advantage in quickly obtaining large areas temperature,make it becomes the only effective approach to obtain temperature data in cold-high mountains.Currently,commonly used temperature retrieval methods are successful in retrieving temperature at flat regions;however,more efforts should be focused on discussing and improving their applicability in mountains,with complex terrain and land cover types.Additionally,analysis about the spatial-temporal characteristic of temperature in cold-high mountains using limited field-measured temperature,as well as comparison analysis between field-measured and retrieved temperature are still two urgent issues.This study is an important component of National Science and Technology Major Project "snow and ice monitoring and its evaluation based on high-resolution remote sensing data in Central Tianshan mountains in Xinjiang"(Grant No.:95-Y40B02-9001-13/15-04).According to its demands in snowmelt process simulation,water resource management and local climate change research,mountains of Manasi River in Central Tianshan Mountain,Xinjiang Province was selected as the study area.The spatial-temporal characteristics of field-measure temperature was analysed firstly.Secondly,factors that affect temperature retrieval in mountains was discussed based on JM model(a signle channel temperature retrieval model),for the purpose determining parameters that suitable for temperature retrieval in mountains.Then,temperature retrieval was done combined with Landsat-8 TIRS data and the modified JM model.Finally,field-measured temperature was used to validate the quality of rereieved temprature.The main contents and results of this study are as follows:(1)The spatial-temporal characteristics of field-measured temperature.Times series data of field-measured temperature,obtained through three scientific investigations carried out at April,2014,April 2015 and June,2016,were used to analyze variation trend of temperature at three scales,i.e.day,month and year.The results of 'daily temperature variation analysis',based on hourly average temperature,show that variation trend of temperature for different land cover types and aspects are very different;during snow accumulation period,the highest temperature of mountainous grass belt appears at 13:00,which is one hour earlier than that of forest belt;however,during snowmelt period,the former appears at 14:00,and which is two hours earlier than that of the later;daily temperature differences at different aspects are more obvious during snowmelt period than that during snow accumulation period,the time that temperature rise to peak at the sunny slope is three hours earlier than that at the semi-shady slope,and temperature of the sunny slope is about 10?15? higher than that of the semi-shady and the shady slopes when satellite passing(about 13:00)the study area.The results of‘Monthly Temperature Change Analysis',based on daily average temperature,show that the dates that temperature drops below 0? or rises above 0? are not the same at different altitudinal belts,during the cooling phase,temperature of snow-ice belt and alpine meadow belt will drop below 0 ? in late October,however temperature of forest belt and mountainous grass belt is still fluctuating near 0? during November and will drop below 0? until early December;during the warming phase,temperature of mountainous grass belt and forest belt will rise above 0? during the period from late February to early March,while it will rise above 0? until the middle April for alpine meadow and snow-ice belt.The results of‘Annual Temperature Change Analysis' based on monthly average temperature show that the highest temperature of snow-ice belt occurs in August,while it occurs in July for alpine meadow belt,forest belt and mountainous grass belt;the lowest temperature of snow-ice belt and alpine meadow belt occurs in December,while it occurs in January for forest belt and mountainous grass belt;during snow accumulation period,temperature inversion is very obvious at mountainous grass and forest belt,temperature gradient is about 0.54?0.61?/100 m from December to February of the next year,and this phenomenon will gradually disappear after March,temperature gradient of March is-0.06?/100 m,and it is about-0.48?-0.61?/100 m from April to June.Besides,temperature variation curve with elevation and temperature gradient at six specified dates were built using field-measured temperature data,for the purpose of assessing temperature retrieval results.(2)Land surface temperature retrieval using thermal infrared data.Aims at determining factors that affect land surface temperature retrieval in cold-high mountains,sensitivity analysis of two major parameters(i.e.land surface emissivity and atmospheric water vapor content)used in JM model were analyzed firstly,the results show that retrieval result is very sensitive to land surface emissivity,increase of land surface emissivity by 0.02 will bring in 1 K error to retrieval result,when atmospheric water vapor content is 0.5 g/cm2;the sensitivity of retrieval result on atmospheric water vapor content is not significant,in the case that land surface emissivity is 0.97,retrieval result will decrease by 0.65 K when atmospheric water vapor ranges from 0.1 g/cm2to 0.5 g/cm2,and it will decrease by 1.59 K when atmospheric water vapor content ranges from 0.5 g/cm2 to 1.0 g/cm2;consider that atmospheric water vapor content of the study area varies from 0.1 g/cm2 to 1.0 g/cm2 during snow accumulation and snowmelt period,and in most cases it is below 0.5 g/cm2,retrieval error caused by atmospheric water vapor content is under 1 K in most cases,which means that retrieval result can meet the accuracy requirement of land surface temperature retrieval.Simulation of the directional emissivity of snow cover,based on a semi-empirical snow directional emissivity model(Hori model),show that terrain has significant effect on snow emissivity,emissivity of medium-grained snow will decrease by 0.08 when the local view angle ranges from 60°to 90°,which will result in an error more than 4 K for temperature retrieval result,therefore,terrain is a key factor should be taken into consideration for land cover temperature retrieval in mountains,especially for surface temperature retrieval of snow cover.In this study,local view angle was introduced into JM method as a parameter describing terrain condition,then thermal infrared data of Landsat-8 TIRS was used to retrieve land surface temperature of the study area at six different dates.Land surface temperature retrieval results show that,during snow accumulation period,temperature in the middle mountain zones is higher than that of low mountain zone and alpine zone,elevation of the highest temperature is about 2500 m,and temperature between 3000?3200 m is approximately equal to that between 1200?1500 m;overall,temperature of the study area decreases gradually along with the increase of elevation during snowmelt period.In addition,temperature variation curve with elevation and temperature gradient at six specified dates were built,in order to perform comparison analysis with field-measured temperature in Chapter five.(3)Comparison of retrieved and field-measured temperature.It should be noted that field-measured temperature is air temperature 10 cm off the ground,while retrieved temperature is "skin temperature" of the ground,which make it is impossible to quantitatively validate the accuracy of retrieved temperature using field-measured temperature.Therefore,comparison analysis,focus on trend consistency comparison,was carried out at three spatial scales,comparison of temperature at measured point and corresponding pixel,comparison of vertical variation trend between retrieved and field-measured temperature,comparison of spatial variation trend between retrieved and field-measured temperature.The results show that retrieved temperature is higher than field-measured temperature,when the ground is coverd by snow,the averaged retrieved temperature is about 12.8? lower than field measured temperature during snow accumulation period,and the former is about 3.1? lower that the latter during snowmelt period;however,the averaged retrieved temperature is about 3.6? higher than field-measured during snow accumulation period when the land cover is barren land,and the former is about 12.6? higher than the latter druing snowmelt period.The vertical variation trends of retrieved and field-measured temperature are generally consistent,i.e.,temperature of mountainous grass belt and forest belt increase gradually with the rising of elevation during snow accumulation period;temperature gradient of field-measured temperature at Dec.5,2015 and Feb.1,2016 are 1.06?/100 m and 0.95?/100 m,respectively,and temperature gradient of retrieved temperature are 0.88 ?/100 m and 0.81?/100 m,respectively;overall,temperature gradient built on field-measured temperature is slightly larger than that built on retrieved temperature during snow accumulation period;during snow melt period,the decrease of temperature with the rising of elevation is very obvious,temperatures gradient of field-measured temperature and retrieved temperature are-0.49?/100 m and-0.83?/100 m,respectively,at May 18,2014.The spatial variation characteristics of temperature trend surface,built on retrieved temperature,are generally same to that built on field-measured temperature,the highest and the lowest temperatures of two temperature trend surfaces all appear in the same regions.Overall,vertical variation trend and spatial variation of retrieved temperature is generally aggrement with that of field-measured temperature,which means the quality of retrieval result is reliable.