基于半变异函数的重庆市地表温度空间异质性建模及多尺度特征分析
作者简介:陈 昭(1993-),男,湖北仙桃人,硕士生,主要从事遥感技术应用及城市热岛等研究。E-mail: 506556163@qq.com
收稿日期: 2018-09-05
要求修回日期: 2019-03-19
网络出版日期: 2019-07-25
基金资助
国家自然科学基金项目(41871226)
重庆市博士后特别资助项目(Xm2016081)
重庆市气象局开放基金项目(KFJJ201602)
重庆市应用开发计划重点项目(cstc2014yykfB30003)
中国气象局省所科技创新发展专项(SSCX201917)
Modeling and Multi-Scale Analysis of the Spatial Heterogeneity of Land Surface Temperature in Chongqing based on Semi-Variogram
Received date: 2018-09-05
Request revised date: 2019-03-19
Online published: 2019-07-25
Supported by
National Natural Science Foundation of China, No.41871226
Chongqing Postdoctoral Special Funding Project, No.Xm2016081
Chongqing Meteorological Bureau Open Fund Project, No.KFJJ201602
Chongqing Application Development Plan Key Project, No.cstc2014yykfB30003
China Meteorological Administration Provincial Science and Technology Innovation Development Project., No.SSCX201917
Copyright
城市地表温度空间异质性的研究对理解城市地表温度空间结构有重要意义。本文利用大气校正法反演地表温度,基于半变异函数构建城市地表温度空间异质性模型,并进一步分析不同空间尺度下地表温度空间异质性结构参数的变化规律。以2013年6月16日的Landsat 8为数据源,以重庆为研究区开展实验,研究结果表明:① 不同空间尺度下重庆地表温度空间异质性均呈现指数模型分布特征;② 在30 m空间尺度下,地表温度空间异质性主要是由空间结构引起,但随机因素引起的空间变异占比为0.45,呈现出明显的块金效应,表明该尺度下随机因素引起的空间变异不可忽略;③ 从空间尺度(30~1500 m)整体变化上看,地表温度空间异质性主要由空间结构引起,同时表现出明显的尺度效应;随着空间尺度增大,块金值(C0)、偏基台值(C)、基台值(C0+C)以及块基比(C0/(C0+C))逐渐减小,表明地表温度空间异质性逐渐减弱但空间自相关性逐渐增强。变程(A)逐渐增大,表示空间自相关性范围逐渐扩大;④ 随机因素引起的空间变异占比为0.23~0.46,呈现出波动变化,这是因为地表温度在像元内部也存在空间异质性。空间结构引起的空间变异相对平缓,这是因为空间尺度的变化不会改变地形结构;⑤ 从尺度域来看,基台值与块金值在尺度域(690 m,1500 m)内呈现出较大幅度波动变化状态,且变化趋势相似,表明地表温度空间异质性的变化与随机因素有较大关联。综上所述,分析地表温度空间结构需要选取合适的空间尺度,尺度较小时,容易受到随机因素干扰,从而影响地表温度在空间结构上的空间变异性;尺度较大时,地表温度空间异质性较弱且变化不稳定。
陈昭 , 罗小波 , 高阳华 , 叶勤玉 , 王书敏 . 基于半变异函数的重庆市地表温度空间异质性建模及多尺度特征分析[J]. 地球信息科学学报, 2019 , 21(7) : 1051 -1060 . DOI: 10.12082/dqxxkx.2019.180611
Analyzing the spatial heterogeneity of land surface temperature (LST) is important for understanding the spatial structure of LST. This study retrieved LST by the atmospheric correction method, and constructed a spatial heterogeneity model of LST by using the semi-variogram function. It then took a multi-scale perspective to discuss LST’s spatial heterogeneity in the study area of Chongqing. A Landsat 8 OLI imagery in June 16, 2013 was the primary data source. Results show that: ① The LST’s spatial heterogeneity was exponentially distributed at different spatial scales. ② At the 30 m spatial scale, the spatial heterogeneity was mainly caused by spatial structure, though the proportion of spatial variation caused by random factors accounted for 0.45, showing obvious nugget effect; thus, random factors cannot be ignored at this scale. ③ On the whole spatial scale (30~1500 m), the spatial heterogeneity was mainly caused by spatial structure, and showed obvious spatial scale effect. As the spatial scale increases, the nugget (C0), the partial sill (C), the sill (C0+C), and the nugget-sill ratios (C0/(C0+C)) gradually decreased, indicating that the spatial heterogeneity declined and the spatial autocorrelation gradually increased. Meanwhile, the range (A) gradually increased, indicating that spatially autocorrelated regions gradually enlarged. ④ On one hand, the proportion of spatial variation caused by random factors ranged from 0.23 to 0.46, showing obvious volatility, because the LST also had spatial heterogeneity within each pixel. On the other hand, the spatial variability caused by spatial structure was relatively flat, because the change of spatial scale did not affect the topographic structure. ⑤ From the scale effect perspective, both sill and nugget showed large fluctuations, and the trend was similar from 690 m to 1500 m, indicating that the change of the LST's spatial heterogeneity was related to random factors. In summary, choosing the appropriate spatial scale is very important for analyzing the spatial structure of LST. When the scale is small, the spatial distribution of LST is easily disturbed by random factors, which affects the variability in spatial structure. When the scale is large, the spatial heterogeneity of LST is weak and unstable.
Fig. 1 Original remote sensing imagery for Chongqing City on June 16, 2013图1 2013年6月16日重庆市原始遥感影像 |
Fig. 2 An example curve of the semi-variogram function图2 半变异函数拟合曲线示例 |
Fig. 3 LST at the 30 m scale in Chongqing City on June 16, 2013图3 2013年6月16日重庆市30 m空间分辨率的地表温度 |
Fig. 4 A fitted curve of the semi-variogram of the LST at the 30 m scale图4 30 m空间分辨率地表温度的半变异函数值的拟合曲线 |
Tab.1 Characteristic parameters of the spatial heterogeneity表1 空间异质性的特征参数 |
空间分辨率/m | 块金值C0/(℃2) | 偏基台值C/(℃2) | 基台值C0+C/(℃2) | 块基比C0/(C0+C) | 变程A/m | 决定系数R2 | 最佳模型 | |
---|---|---|---|---|---|---|---|---|
30 | 10.22 | 12.51 | 22.73 | 0.45 | 12330 | 0.943 | 指数 |
Fig. 5 Spatial distribution of the LST at different spatial scales in Chongqing City on June 16, 2013图5 2013年6月16日重庆市不同空间尺度下地表温度的空间分布 |
Tab. 2 Characteristic parameters of the spatial heterogeneity at different spatial scales表2 不同空间尺度下的空间异质性特征参数 |
空间分辨率/m | 块金值C0/(℃2) | 偏基台值C/(℃2) | 基台值C0+C/(℃2) | 块基比C0/(C0+C) | 变程A/m | 决定系数R2 | 最佳模型 |
---|---|---|---|---|---|---|---|
30 | 10.22 | 12.51 | 22.73 | 0.450 | 12 330 | 0.943 | 指数 |
60 | 10.16 | 12.33 | 22.49 | 0.452 | 12 480 | 0.949 | 指数 |
90 | 10.01 | 12.34 | 22.35 | 0.448 | 12 570 | 0.985 | 指数 |
… | … | … | … | … | … | … | … |
480 | 6.98 | 11.12 | 18.10 | 0.386 | 14 370 | 0.955 | 指数 |
510 | 8.42 | 11.06 | 19.48 | 0.432 | 18 450 | 0.960 | 指数 |
540 | 7.61 | 10.72 | 18.33 | 0.415 | 17 070 | 0.958 | 指数 |
… | … | … | … | … | … | … | … |
1440 | 4.04 | 9.58 | 13.62 | 0.297 | 23 970 | 0.953 | 指数 |
1470 | 3.17 | 9.00 | 12.17 | 0.260 | 19 920 | 0.945 | 指数 |
1500 | 5.95 | 10.01 | 15.96 | 0.373 | 31 860 | 0.958 | 指数 |
Fig. 6 Changes in the nugget (C0), the partial sill (C), and the sill (C0+C) at different scales图6 不同空间尺度下块金值(C0)、偏基台值(C)、基台值(C0+C)的变化状态 |
Fig. 7 Change in the nugget-sill ratios (C0/(C0+C)) at different scales图7 不同尺度下块基比(C0/(C0+C))的变化状态 |
Fig. 8 Chagne in the range (A) at different scales图8 不同空间尺度下变程(A)的变化状态 |
[1] |
|
[2] |
Rao,
|
[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] |
[
|
/
〈 | 〉 |