Assessing the flow variability of headwater streams in High Mountain Asia using satellite remote sensing

Headwater streams are significant in the ecology and hydrology of river systems as they transport minerals, nutrients, and sediment downstream. Headwaters in High Mountain Asia are of particular importance as they are the primary routes that funnel glacier/snow melt to sustain continuous water supply for the downstream region. However, at present, the responses of headwater streams to glacial meltwater and their influence on downstream rivers are not well understood. Further, gauge network observations are largely unavailable in High Mountain Asia due to the extreme topographic attributes and climatic conditions. Remote sensing of downstream flow is possible but has not yet been applied for narrow headwater streams due to limitations of satellite resolutions. The temporal and spatial resolution of Planet imagery can help image these streams previously too small to see with coarser sensors. Therefore, we propose a viable approach to tracking headwater streams across High Mountain Asia using multi-temporal 3m-resolution Planet images coupled with width-based discharge inference. We formulate four research questions as the following: a) how accurately can we estimate the width of headwater streams using computer vision and Planet imagery? b) can these widths be transformed to discharge with existing methods and detect the flow variability of headwater streams in High Mountain Asia? and c) what are the hydrologic characteristics of headwater streams in the region? In this study, we take advantage of computer vision methods and the high spatiotemporal resolutions of Planet imagery to classify narrow, turbid, and shadowed headwater streams. We then perform stream width extraction from the classified images and width-based Bayesian discharge inference. From this approach, we use multi-temporal Planet imagery to evaluate the region's flow variability of headwater reaches. Finally, we analyze the patterns and characteristics of the hydrologic regime of these headwater streams and describe their relationship to the glacier meltwater and downstream waters of High Mountain Asia. Together, these approaches will help inform the applicability and opportunities for combining Planet imagery with computer vision stream detection and width-based discharge inference to estimate headwater flows in a unique and dynamic hydrologic region.