【遥感测绘】【2020.05】稀缺数据河流环境下可通航面积估算——Chindwin案例研究

本文为荷兰代尔夫特理工大学（作者：F.S. Laurens）的硕士论文，共97页。

缅甸西北部贫困地区的季节性Chindwin河是该地区商业和运输的中心动脉。不幸的是，结合枯水期的低水位、动态形态、陈旧的船舶设备和有限的监测，每年会造成大量船舶搁浅，导致很多人员伤亡和经济损失。

2019年初启动了一个试点项目，以使Chindwin河的航行更加安全。该试验项目包括为数量有限的在Chindwin河上航行的商船配备CoVadem技术。在最先进的大数据技术支持下，CoVadem通过收集商业船队的龙骨下水深测量值来绘制最新水深图。该项目旨在通过与商船共享有关安全路线和预测水深的信息，提高Chindwin河航行的安全性。然而，使用CoVadem技术的附加值易受参与船只数量的影响。

形态变化和CoVadem数据典型空间扩展（有限的横截面覆盖）的结合限制了从数据中推算出导航通道的范围。由于只知道航道的一小部分，CoVadem技术在航行安全和效率方面的全部潜力还没有达到：尚不清楚在哪里有可能通过其他船只，哪里因瓶颈等原因应调整速度，哪里有较短的航线。

本研究的目的是提供更多有关CoVadem船舶可航区的资料，以协助船长完成航行任务。在这项研究中，已经开发了两个基于物理的模型，它们能够独立地执行这项任务。第一个模型名为soilmodel，它将CoVadem数据与河床中预期的最大坡度（给定河床的已知物理特性）结合起来。第二个模型，轴对称模型用于计算CoVadem数据周围的河床坡度，以确定可通航区域。河床坡度采用轴对称解计算，轴对称解是估算曲流河断面比例尺水深的已知解析解。此外，关于河岸和沙丘存在（和位置）的假设也包含在轴对称模型中。轴对称模型并不适用于Chindwin河的每个区域。

已经开发了两个可靠性指标来指示在何处可以安全地使用轴对称模型（以及在何处不能）。第一个可靠性指标被称为通道稳定性得分，这是一个衡量低流量通道对准稳定性的指标。分数是根据多年的卫星图像计算出来的，这些图像是在低流量的月份获得的。Watermask被检测出来，并与其他年份的数据相结合。稳定性得分是根据Watermask的可变性计算出来的。假设轴对称模型对于河道走向更稳定的河流区域更可靠。第二个可靠性指标是河流的曲率比：假设轴对称模型对河流弯曲部分更可靠。

本文对位于Chindwin河沿岸的四个研究案例的物理模型性能和可靠性指标假设进行了测试和评价。这些模型估计了一个可通航的区域。为了判断通航面积估算的质量，提出了两个无量纲的性能指标：安全评分和航道覆盖评分。第一个性能指标与通航面积估计的可靠性有关，后者与估计的宽度有关。研究结果很有希望。从这项研究可以看出，在稀缺的CoVadem船舶航迹数据周围进行航道估计，可以从基于物理模型的应用中获得实质性的好处。soil-model是非常可靠的，但在可通航面积估计的宽度上只有很小的改善。轴对称模型显著提高了单共面轨道线的通航宽度估计值，精度可以达到O(100m)。然而，该模型并非100%可靠：大约2%的通航面积估计是错误的。当仅应用于可靠性指标（航道稳定性得分和/或航道曲率），将轴对称模型标记为可靠区域时，轴对称模型的可靠性可以大大提高。因此，关于可靠性指标的两个假设都得到了研究结果的支持。

这项研究将大数据、基于物理的模型和遥感结合在一起，其方式尚未得到早期的完整证实：以专门为可通航面积估计而设计的模型和以测量数据为出发点。此外，轴对称模型的可靠性与遥感/曲率之间的相关性是以前从未研究过的。最后，本文提出的两个性能指标对通航面积估算的评估具有很大的应用前景。因此，这项研究增加了（开放访问）跨平台数据积累和数据使用的最新进展。

The seasonal Chindwin river in the poornorth-western part of Myanmar forms the central artery for business andtransport in the region. Unfortunately, a combination of low water levelsduring the dry period, the dynamic morphology, archaic boat equipment andlimited monitoring, yearly results in large numbers of grounding ships. Everyyear this causes many injuries and deaths as well as economic losses. A pilotproject was initiated in the beginning of 2019 to benefit safer navigation onthe Chindwin. The pilot project involves equipping a limited number ofcommercial ships sailing on the Chindwin with CoVadem technology. Supported bystate of the art Big Data technology, CoVadem charts the most up to date waterdepths by collecting under keel clearance measurements from the commercialfleet. The pilot project aims to increase safety on the Chindwin throughsharing information about safe routes and forecasted water depths with captainssailing the river. The added value of using CoVadem technology, however, isvulnerable to the number of participating vessels. A combination ofmorphological changes and the typical spatial spread (with limitedcross-sectional coverage) of CoVadem data limits the extent of the navigationchannel that can be derived from the data. With only a small part of thenavigation channel known, the full potential of CoVadem technology fornavigational safety and efficiency is not reached: it is unclear where passingof other vessels is possible, where speeds should be adjusted due to e.g.,bottlenecks and where shorter routes are present. The objective of thisresearch is to provide more information about the navigable area around CoVademship track data in order to assist captains with navigation. During thisresearch, two physics-based models have been developed that are independentlyable to carry out this task. The first model, named the soilmodel, combinesCoVadem data with the maximum slope that can be expected in the river bed(given known properties of the bed material). The second model, theaxi-symmetric model, calculates the bed slope around CoVadem data to determinenavigable areas. The bed slope is calculated with the axi-symmetric solution, aknown analytical solution to estimate the cross-section scale bathymetry formeandering rivers. In addition to this, assumptions about the presence (andlocation) of river banks and sand dunes are included in the axi-symmetricmodel. The axi-symmetric model is not suitable for every part of the Chindwin.Two reliability indicators have been developed to indicate where theaxi-symmetric model can be safely used (and where not). The first reliabilityindicator is named the channel stability score, a measure for how stable thelow discharge channel alignment is. The score is calculated based on multipleyears of satellite imagery, acquired during months with low discharge. Watermasks are detected and combined with those of other years. The stability scoreis calculated from the variability in the water masks. It was hypothesised thatthe axi-symmetric model is more reliable for parts of the river with a morestable channel alignment. The second reliability indicator is the curvatureratio of the river: it was hypothesised that the axi-symmetric model is morereliable for curved parts of the river. The performance of the physics-basedmodels and the hypotheses about the reliability indicators have been tested andevaluated for four study cases located along the Chindwin river. The modelsestimate a navigable area. To judge the quality of a navigable area estimate,two dimensionless performance indicators have been developed: the safety scoreand the channel coverage score. The first performance indicator is related tothe reliability of a navigable area estimate, the latter to the width of theestimate. The results are promising. It follows from this research thatnavigation channel estimates around scarce CoVadem ship track data cansubstantially benefit from application of a physics-based model. The soil-modelis very reliable, but boasts only a small improvement in the width of thenavigable area estimate. The axisymmetric model increases the navigable widthestimate from a single CoVadem trackline significantly O(100 m). However, themodel is not 100% reliable: around 2% of the estimated navigable area waswrong. The reliability of the axi-symmetric model can be substantially improvedwhen it is only applied to areas where the reliability indicators (channelstability score and/or channel curvature) mark the axi-symmetric model asreliable. Hence both the hypotheses about the reliability indicators aresupported by the results.

This research combines Big Data,physics-based models and remote sensing in a not early demonstrated way: withmodels tailored for navigable area estimates and with measured data as thestarting point. The correlation between axi-symmetric model reliability andremote sensing/curvature is, moreover, something that has not been demonstratedbefore. Finally, the two developed performance indicators show great promisefor the evaluation of navigable area estimates. As such, this research adds tocurrent advancements in (open-access) cross-platform data accumulation andutilisation.

1. 引言
2. 研究背景
3. 研究方法
4. 结果与评估
5. 讨论
6. 结论与建议

更多精彩文章请关注公众号：