| 摘要: |
| 冒泡作为湖库等淡水生态系统中温室气体CH4排放的主要方式,存在着高度的时空异质性。水库库尾浅水区沉积物易产生甲烷,并以冒泡的形式释放,但是在开放水体中很难捕捉到CH4冒泡的排放热区。冰封水库在冰的生长过程中会冻结捕捉由沉积物产生并上浮富含CH4的气泡,在冰面上形成一定规模的“冰封气泡”,这使得基于冰封气泡特征识别CH4冒泡排放热区的规模、空间分布成为可能。本研究以我国东北地区大型水库-东风水库为调查对象,基于冬季水库冰面的无人机影像与现场冰封气泡特征调查,结合面向对象的分类方法构建了冰封气泡影像提取分割方法,利用空间自相关分析探究冰封气泡空间分布模式,识别冰封气泡聚集的热点区域。研究结果表明：(1)东风水库库尾冰封气泡直径的范围为1~10cm,无人机15m飞行高度空间分辨率为0.4cm,能够满足识别冰封气泡的质量要求,形态学顶帽变换方法能够有效改善图像光照不均现象,显著提高正射影地图的合成质量；(2)利用气泡、冰裂纹在长宽比、亮度、密度等特征值的差异性能够有效进行冰封气泡的提取分类,受调查的3个气泡区分类总体精度均为0.8以上,显示了较高的冰封气泡提取精度；(3)3个气泡区冰封气泡面积占比顺序为R2>R1>R3,总气泡面积占库尾的0.24%,但3个区域内部10个调查样带的气泡面积占比存在一定差异,分布范围为2.6%~7.8%,冰封气泡空间分布均呈现显著的聚集模式(Moran"s I >0.8, Z>80),被划分为冒泡排放热点的“高-高”聚类占库尾总面积的21.5%。基于无人机影像识别冰封水库CH4冒泡排放热区可以进一步指导开放水体中CH4排放规模研究。 |
| 关键词: 冰封水库 无人机 甲烷冒泡 排放热区 面向对象分类 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金项目（面上项目，重点项目，重大项目）（42077348、42377412） |
|
| An effective method to identify the methane ebullition hotspots in ice-covered reservoir with image data analysis by UAV* |
|
HU Tianchao,jinkang,jinye,chen xiaoqiang,sunuan,lizhijun,xvshiguo
|
|
Dalian University of Technology
|
| Abstract: |
| Ebullition is a major pathway for methane emissions in freshwater ecosystems such as lakes and reservoirs, characterized by high spatiotemporal heterogeneity that makes it difficult to capture ebullition hotspots. exhibiting high spatiotemporal heterogeneity. In shallow zones near the reservoir tail, sediments tend to produce methane that is released in bubble form. However, capturing CH4 ebullition hotspots in open waters is challenging. During ice cover formation, methane-rich bubbles produced by sediments can become trapped and frozen in the ice, forming visible "ice-trapped bubbles" on the ice surface. This phenomenon allows for the identification of CH4 ebullition hotspots based on the characteristics of ice-trapped bubbles. This study investigates the large Dongfeng Reservoir in northeastern China. Using UAV imagery of the winter reservoir ice surface and field surveys of ice-trapped bubble characteristics, an object-oriented classification method was developed to extract and segment images of ice-trapped bubbles. Spatial autocorrelation analysis was employed to explore the spatial distribution patterns of ice-trapped bubbles and identify hotspot areas of bubble aggregation. The results indicate: (1) The diameter of ice-trapped bubbles in the Dongfeng Reservoir tail ranges from 1 to 10 cm. The spatial resolution of UAV imagery at a flight height of 15 m is 0.4 cm, sufficient for identifying ice-trapped bubbles. The morphological top-hat transformation method effectively improves image illumination uniformity and significantly enhances the quality of orthophoto mosaics.(2) Classification and extraction of ice-trapped bubbles based on differences in aspect ratio, brightness, and density between bubbles and ice cracks show high precision, with overall classification accuracies above 0.8 in the three surveyed bubble zones.(3) The proportion of ice-trapped bubble area among the three zones follows the order R2 > R1 > R3, with the total bubble area accounting for 0.24% of the reservoir tail. However, the bubble area proportions within the ten surveyed transects in the three zones vary between 2.6% and 7.8%. The spatial distribution of ice-trapped bubbles exhibits significant clustering (Moran"s I > 0.8, Z > 80), with "high-high" clusters identified as ebullition hotspots covering 21.5% of the reservoir tail area. Using UAV imagery to identify CH4 ebullition hotspots in ice-covered reservoirs can further guide studies on CH4 emission scales in open waters. |
| Key words: Ice-Covered Reservoir UAV Methane ebullition Hotspots Object-Based Classification |
