东天山长时序遥感冰川形态参数变化特征研究

Regional climate change and allocation of regional water resources urgently need wide range and long time-series researches on glacier change mechanism of eastern Tianshan Mountains. However, since there are massive remote sensing imageries from which we haven’t mined sufficient glacier change information, and there is not an effective glacier change time-series analysis method, these reasons limit the study to explore glacier change mechanism on eastern Tianshan Mountains from different views. The purpose of this project is to study the change characteristics of glacier key morphological parameters, and this project focus on two points: one is to improve glacier boundary delineation method, the other is to establish long time-series analysis model of glacier morphological parameters. Firstly, we establish a glacier boundary delineation method based on knowledge using multitemporal remote sensing data to the test area eastern Tianshan Mountains, aiming at reducing the interference of clouds, shadows and snows, so that to improve glacier boundary delineation automaticity and accuracy. Secondly, we build the long time-series change models of typical glacier morphological parameters, to compute morphological information of glaciers which are lack of efficient observations. Finally, on the background of climate changes and different time scales, this project will acquire glacier morphological change mechanism, after analyzing changes of different time scale, glacier sizes, elevations, orientations. Moreover, this project can provide references to the quantitative analyses of response relationships among glaciers, rivers, climates and other environmental factors.

区域气候变化和区域水资源配置迫切需要覆盖全、时序长的冰川变化规律研究。然而，由于对已有大量遥感数据中冰川变化信息挖掘不充分，以及缺乏有效的冰川变化时序分析方法，限制了从不同角度探求长时序东天山冰川变化机理的研究。本项目以研究东天山冰川关键形态参数的变化特征为切入点，重点在改进现有遥感冰川边界提取方法，并构建形态参数变化分析的长时序模型。主要研究内容包括三部分：首先，以东天山为研究区，利用多时相遥感数据建立基于知识冰川边界提取方法，减小或消除云、阴影、积雪对冰川边界的干扰，提高冰川边界提取的自动化程度和精度；然后，以典型单条冰川或流域范围冰川为对象，构建长时序冰川形态参数变化模型，有效计算缺观测资料年份冰川形态信息；最后，分析气候变化条件下，不同时间尺度、不同规模、高程、朝向、坡度因素作用下的东天山冰川形态变化机理。进一步，为定量化分析冰川与气候、河流水量等环境要素响应过程提供参考。

在气候变化加剧背景下，项目以多光谱、合成孔径雷达、激光雷达为数据源，利用格网化空间分析策略、GMM (Gaussian Mixture Model )、LFM (Logarithmic Fitting Model)、ARIMA (Autoregressive Integrated Moving Average model)，分析研究区冰川平面和高程参数变化。项目主要研究内容如下所述。①项目通过期望最大算法对局部冰川区域NDSI (Normalized Difference Snow Index)建立GMM，自动计算出区域内的冰川提取阈值。基于GMM的冰川阈值计算方法，有效在解决了GMM曲线模拟过程中冰川类、非冰川类占比不平衡的问题。②项目利用0.5°×0.5°网格化空间分析策略，分析天山地区1962-2016年冰川时空变化，发现1990-2005年期间气温升高速度加快，这是导致冰川加速退缩的主要原因。③构建ARIMA模型对28条冰川的面积进行未来5年的预测，得出博格达山冰川部分冰川波动前进，总体呈退缩趋势。④利用2017年、2018年的Sentinel-1A数据，采用重复轨道InSAR (Interferometric Synthetic Aperture Radar)技术，结合LFM得到博格达山、哈尔里克山表面高程变化情况。博格达山2000-2017年冰川高程变化量为-11.6±1.3 m。2000-2018年哈尔里克山冰川高程变化量为-8.74±0.14 m。⑤研究SBAS(Small Baseline Subsets) InSAR与冰川运动作用原理，构建哈尔里克山2017-2018年冰川形变变化密集时间序列，得出该时间段内该区域冰川LOS向形变为-11.77±9.73mm，夏季降水丰富是主导哈尔里克冰川形变增加的主要因素，冬季降水较少，温度偏低，冰川形变不明显。