低空遥感无人机影像反演河道流量

河道流量在维持水圈系统稳定性、估算国家水能资源可开发量等方面具有重要作用。卫星遥感受其分辨率限制很难准确反演中小河流流量,近地面遥感流量计算方法及传统水文测流方法技术复杂、设备昂贵、测算效率低,限制了其在无资料区、灾害突发事件非接触式应急监测等方面的广泛应用。为此,在充分吸收国内外遥感反演河道流量方法优点的基础上,基于低空遥感无人机（UAV）影像,提出了一种适用于各类尺度河流的高效、非接触、简易快速反演河道流量的方法。该方法提供了有、无地面实测大断面两类情况下流量反演途径,通过无人机影像生成点云和表面高程（DSM）,基于点云和DSM获取水面宽、糙率、水面比降以及水上大断面信息,采用水力学方法计算河道流量。并根据地面336组野外站点实测数据验证了方法的精度,进一步分析了无地面实测大断面情况下的流量计算误差。结果表明,反演流量在高值区略高于实测流量,可以满足灾害应急监测流量精度需求（R ²= 0.997,RMSE = 4.55 m ³/s）;无地面实测大断面资料而进行概化时,流量计算误差随水位升高、河宽增大而减小,最大累积误差为最大过水流量的8.28%,误差主要来自于水位低、河宽小、流量小的过水断面底部。考虑到研究区大断面多样性受限,而人类活动影响下的河底断面复杂多样,未来尚需进一步研究提高近河底处大断面概化精度,以提高无地面实测大断面情况下的流量反演精度。本文利用无人机遥感影像反演河道流量的思路可为灾害应急监测提供快速流量监测的新途径,也可为无资料地区遥感水文测站的建立提供重要参考依据。

Measuring streamflow with low-altitude UAV imagery

Stream flows are of great importance in maintaining a stable hydrosphere and assessing available water resources of a nation. However, previous satellite-methods are difficult to retrieve stream flows for middle- or small-scale rivers due to the satellite course spatial resolution whereas near-ground measuring methods have too complex procedure, requirement of expensive apparatus, or low-efficiency. These shortcomings hindered them to be used widely in non-gauged areas and situations needing non-contact measurement, e.g., accidental pollution events. This paper presented a novel, non-contact, fast method to calculate streamflow using UAV images which can be easily applied to rivers with different scales of width. Using this method, stream flows can be calculated with or without ground-measured cross-section data. With UAV images it produced point-cloud and DSM (digital surface model) which were then used to calculate values of river-width, roughness, longitudinal water-surface slope and cross-section above water surface. With all these values, the hydraulic method was finally adopted to calculate stream flows. Results show that the method has a satisfactory performance with modelled streamflow values slightly higher than observed ones at high-flow periods (R ²= 0.997, RMSE = 4.55 m ³/s) with ground-observed cross-section data. When the cross-section data were absent, the cross-section under water can be generalized with the UAV measured above-water cross-section data. Errors in estimating stream flows induced by cross-section generalization decreased with increment of water-level and water-width. The maximum accumulated errors accounted for 8.28% of the bankfull streamflow. The errors were resulted from the generalization of river bottom with un-regular cross-sections. All the results and methodologies could be of great help in streamflow measurement in accidental pollution events and in ungauged areas across the globe.