Estimation of Heating-Up Effecting in the Yellow River Basin Based on MODIS Data

: Using MODIS land surface temperature data, air temperature data and elevation 9 data from 2000 to 2015 in the Yellow River Basin. The GWR analysis method with high 10 accuracy was chosen to establish the regression model of plateau air temperature, land surface 11 temperature and altitude. In the 12-month GWR regression model, the determination 12 coefficient (Adjusted R 2) was above 0.95 or more (0.959-0.980) and the root-mean-square error 13 (RMSE)was between 0.740 and 1.029 °C . Depending on the model, the air temperature of the 14 Yellow River Basin is estimated and the accuracy is verified. On this basis, the average 15 monthly air temperature in the basin is converted to altitudes of 4500m and 5000m, and the 16 heating-up effects of various shapes in the basin are compared and discussed. The results 17 show that: (1) Using the GWR method, combined with the observation data of the ground 18 station, the accuracy of the air temperature estimation in the Yellow River Basin can be 19 increased to 0.740 °C ; (2) According to the estimated annual variation of the spatial distribution 20 of the 12-month average temperature, in the upper of the Tibet Plateau, the Huangshui Valley 21 and the Gannan Plateau have lower annual air temperatures and less spatial distribution. 22 While the air temperature in the northeast of the upstream Inner Mongolia plateau was 23 higher, which was related to the rapid drying temperature rise near the desert. The change of 24 mean monthly temperature in the middle and lower reaches is relatively high and the change 25 is small, which is closely linked to the fact that it is located in the low-elevation area of the 26 basin plain and has perennial light and heat.(3) The heating-up effect in the Yellow River Basin 27 is outstanding. It is preliminaries estimated that at the same altitude, the Tibet Plateau is about 28 1.5~8 °C higher than the Loess Plateau, and about 6~13 °C higher than the North China Plain. 29

peripheral area or coastal. The phenomenon of high regional distribution is the main factor 42 affecting the distribution of altitudinal zones on a large scale, and is an important mechanism 43 for the spatial pattern and dynamic change of the geographic surface system [1][2][3][4]. The 44 thermal effect of the mountain produces a temperature space pattern that the internal 45 temperature of the mountain is higher than the outside at the same altitude, which the 46 essence is that the mountain or plateau uplift has increased the temperature effect on the 47 surrounding atmosphere [5][6][7]. The Yellow River Basin has a large geographical span and 48 flows through most of northwest and north China, which has a major impact on the Asian

64
while studied the trend of air temperature variation in the typical region of the Yellow River 65 Basin [12]; These studies all focused on the use of statistical methods to study the interannual 66 variation of temperature in the Yellow River Basin, which the results can reflect the temporal 67 variation of temperature and the overall variation of the region to a certain extent, but it 68 cannot accurately reflect the detailed spatial distribution of air temperature [13][14][15].

69
At present, the most common way to get the air temperature of a certain spatial location 70 is to obtain the data from the existing meteorological stations [16]. In the Yellow River Basin, 71 on the one hand, due to the scarcity of meteorological stations, on the other hand, because the 72 plateau types and plains are interlaced in the basin and the types of landforms are diverse, 73 the spatial position of these stations cannot fully reflect the spatial variability of air 74 temperature [17,18]. Yao Yonghui et al. conducted a detailed discussion on the variation 75 characteristics and regional differences of annual average temperature and precipitation from 76 1960 to 2008 in the Hengduan Mountains, with selecting the 27 weather stations with the 77 longest measured data sequence in the region. It is proposed that the air temperature in the 78 Hengduan Mountains showed a statistically warming trend in the past 50 years [19]. In order 79 to obtain more accurate interpolation results, Hwang et al. (2005)

130
Studying the heating-up effect of the mountain and estimating the difference between 131 the mountains at the same altitude is the key to quantifying the mountain effect. The LST data 132 provided by the MODIS sensor has medium spatial and temporal resolution, which has been 133 widely used in station temperature estimation for sparse mountainous areas. Although many 134 studies have estimated the air temperature of different regional types, however, the accuracy 135 needs to be improved. Based on Modis data and air temperature data, this paper estimates

176
The air temperature data used in this paper are the average monthly temperature data

183
ASTERGDEM data is downloaded from http://www.gdem.aster.ersdac.or.jp. The spatial 184 resolution is 30 m. In order to estimate the temperature together with the MODIS land surface temperature data, re-sampling is performed in a grid unit of 1 km x 1 km.

209
The seasonal trend can be expressed as: where s is the seasonal duration (such as the number of observations per year); , is the 212 influence factor of the i-th period. When time t is the i-th observation, , =1, otherwise 0.

213
Therefore, when t is the seasonal zero period, , − ,0 = -1. When t is other seasonal period,

214
, − ,0 = 1. , usually considered as seasonal dummy variable, it has two available 215 values of 0 and 1 to explain the seasonal period in the regression model.

224
The iterative algorithm starts with the STL method to estimate Ŝ , where Ŝ is estimated 225 from the mean of all seasonal subsequences. After that, the iterative algorithm follows the 226 following 4 steps: the number and location (time) information of these mutation points (t1 * ,…,tm * ) can be 229 estimated from the corrected seasonal data to get Yt − Ŝ .

230
2) The trend coefficients , , in the interval j=1,…,m are calculated by the robust 231 regression equation (2)   Where dijis the distance between the sites and bis anadaptive bandwidth size that is 253 defined as the nearestneighbour distance.

254
The bi-square kernel function has a clear-cut rangewhere the weighting is non-zero. In 255 this study, the adaptivebi-square kernel function was used to describe the spatialdependence 256 of the data: The selection of an appropriate bandwidth requires careand in some instances may 259 benefit from a measure of howwell the model fits the data. An interval search was used 260 todetermine the optimal bandwidth of the adaptive bi-squarekernel function.

261
(3)The high-precision GWR model is used to estimate the average monthly air 262 temperature of the basin for 12 months. In this paper, the air temperature is converted into was 4,500m and 5000m respectively, so as to compare the differences of geomorphic units 266 within the basin: Where h is the specified altitude, in this paper 4,500 m and 5000 m; Tah and Ta

295
Although the air temperature in the basin has a linear relationship with the land surface temperature, the air temperature is still spatially different.

296
Therefore, the altitude factor is introduced into the GWR regression analysis method to estimate the air temperature. The GWR regression analysis results

297
( Table 2) show that, from the estimation, the 12-month decision coefficient (Adjusted R 2 ) is higher than 0.95 (0.959-0.980), and the RMSE of each month is 298 between 0.740 and 1.029°C, the accuracy is higher.   According to the spatial distribution of the estimated air temperature (Fig. 2), it can be 333 seen that the air temperature range in January (the coldest month) is -20.612~6.499°C, and the 334 air temperature range in July (the hottest month) is 0.906~35.704°C; The air temperature in 335 each month is generally low in the west and high in the east, followed by the north-central 336 characteristics, and the reason is that the west is located in high altitude areas such as   Therefore, it can be seen from the elevation of the isotherm that it gradually rises from the 371 east to the west of the Yellow River Basin.

319
same altitude is compared, and the air temperature difference of different geomorphic units 375 at the same altitude is calculated.The average monthly air temperature was converted to the 376 altitude of the Yellow River basin in accordance with equation (5), which was 4500m and 377 5000m, and the air temperature profile was made along the east-west direction (Fig. 5, Fig. 6).

378
From the air temperature at the same altitude in the Tibet Plateau, the Loess Plateau and the 379 North China Plain, the air temperature in the Tibet Plateau is higher than that in the Loess

445
Later, as the summer monsoon weakened, the temperature began to decline in August. Thus, 446 the annual variation of temperature is steep and symmetrical, with spring temperatures 447 slightly higher than autumn.

448
According to the time series analysis of MODIS land surface temperature, the 449 interannual variation of air temperature data and land surface temperature is basically similar.

450
Due to the influence of underlying surface information or cloud and the complex topography,

490
(2) Through the estimation of the temperature of the Yellow River Basin for 12 months, 491 the variation of temperature and time in the study area was obtained. According to the spatial 492 distribution of the estimated air temperature, the temperature in each month is generally low 493 in the west and high in the east, followed by the north-central characteristics, and the reason 494 is that the west is located in high altitude areas such as plateaus and mountains, and the 495 annual temperature is low.

512
Although the focus of this paper is on the estimation of the mean value, the method 513 proposed in this paper can also be applied to determine the daily maximum, minimum and 514 average temperature. In addition, more variables (such as NDVI, precipitation, albedo, etc.) 515 will be considered in the future to explore whether they improve the accuracy of model 516 estimation.     Study area and weather station distribution map. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
Spatial distribution of monthly mean Ta in the study area. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 3
Distribution pattern of 0°C isotherms in the key research area of the Yellow River Basin in the winter half year. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 4
Distribution pattern of 10°C isotherms in the key research area of the Yellow River Basin in the summer half year. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 5
Temperature distribution map of the 4500m altitude in the Yellow River Basin (left: January;right: July). Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors. Temperature difference in each month at an altitude of 4500m