Waterways Are Not Vacant Land: The Decline of Blue Space in Khulna and the Strategic Blind Spot of Secondary Coastal Cities Worldwide

一、一个被反复计算、却很少被追问的选择

在孟加拉国库尔纳市，1994年至2024年的三十年间，地表水体从4.20平方公里缩减到2.27平方公里，建成区占研究区域的比例从25.9%上升到53.3%。这组数据出自一项基于多时相Landsat影像的研究：1994—2004年使用Landsat 5 TM，2014—2024年使用Landsat 8 OLI，结合归一化植被指数（NDVI）、归一化建成区指数（NDBI）与归一化水体指数（NDWI），配合随机森林分类与分类后变化检测，总体精度介于84.5%至89.3%，Kappa系数在0.78至0.86之间。

如果只把它读成一条生态退化新闻，就浪费了这组数据真正的分析价值。

更值得追问的是：为什么在城市发展的每一个阶段，水面都是最先被让渡的那一类空间？答案并不神秘，也不完全来自人口压力的物理必然。它来自一套非常具体的价值识别机制——建成区可以被征税、被抵押、被计入增长，可以被法律确权，可以被资本定价；而一条运河、一片湿地、一个滞洪塘，在城市账本上通常既没有租金，也没有清晰的产权主体。它们是“尚未被使用的土地”，而不是“正在提供服务的基础设施”。

一旦这个定义成立，填埋水面就不是短视，而是一种在既有制度框架内高度理性的行为。库尔纳案例的重要性，正在于它把这种“理性”的长期后果完整地呈现出来。

二、两阶段演进：不可逆性是怎样积累的

研究的时序分析显示出一条清晰的路径：早期阶段以混合性的土地再开发为主，后期阶段则明显转向水体直接转为建成区。同时，运河填埋、湿地丧失与河岸改造的热点集中出现在中部与南部市政分区。

这两个观察放在一起，指向一个常被忽略的机制问题——水文系统的损失不是线性递减，而是连通性递减。

单个池塘被填埋，城市或许还能承受；但当填埋从零星行为变成主导模式，排水网络的节点被逐个移除，剩余水体之间的水力联系被切断，系统就跨过了一个门槛。越过门槛之后，损失不再表现为“少了一点水”，而表现为暴雨后积水范围扩大、退水时间延长、内涝频率上升。此时再去恢复，成本不再等于当年填埋所节省的土地价值，而要以整个片区的排水改造、泵站建设与土地回购来计价。This is precisely the structural predicament facing Khulna as a climate-sensitive coastal city: it simultaneously bears the external pressure of sea-level rise and the self-weakening caused by the erosion of its internal drainage capacity. The two curves move in opposite directions, yet when risks compound, the relationship is multiplicative.

III. Attention Mismatch: Where Is the True Frontier of the Global Urban System?

International urban research, climate finance, and infrastructure investment have long focused on a handful of megacities. This has its rationale: scale brings visibility, and visibility brings capital and policy attention. But the geographic distribution of risk does not obey the distribution of attention.

Khulna is explicitly positioned in the study as a “secondary coastal city.” The common characteristics of such cities are quite consistent: urbanization outpaces the development of planning capacity, fiscal autonomy is limited, the technocratic ranks are thin, and at the same time they are located in geographic positions with the highest climate exposure. They lack the bargaining chips of megacities, yet bear a considerable share of global coastal adaptation pressure.

This constitutes a mismatch in global urban governance that has yet to be squarely confronted: the cities that most need spatial governance capacity are precisely those that most lack spatial governance tools. The annual urban population growth rate of about 3.07% in Bangladesh mentioned in the study, placed in this context, is no longer merely a demographic statistic, but an indicator of the race between institutional carrying capacity and population growth.

From this perspective, Khulna is not local news, but a replicable sample. A large number of small and medium-sized coastal cities in the Global South are repeating a similar trajectory at different speeds.

IV. The Real Contribution of Remote Sensing and Machine Learning: The Institutionalization of Visibility

The methodological significance of this study deserves to be evaluated separately from its conclusions.

Combining NDVI, NDBI, and NDWI with random forest classification, and maintaining the reliability of the water body category in classification results across multiple periods, has the practical effect of transforming a change long on the margins of administrative vision into traceable, comparable, and verifiable time-series evidence.

This changes the epistemological basis of urban governance. In the past, the filling of water surfaces was often dispersed, gradual, and lacking unified records; when disputes arose, each side stuck to its own account. But when changes can be mapped year by year, the attribution of responsibility shifts from a contest of narratives to a contest of data. This is a precondition for transparency in governance.

But restraint must be maintained at the same time: measurable does not mean implementable. Satellites can prove that a canal has disappeared, but they cannot bring a canal back. The links after mapping—planning permits, land registration, enforcement capacity, fiscal incentives—are where outcomes are determined. Technology solves the problem of “not seeing”; it does not solve the problem of “not wanting to manage.” Without supporting institutions, remote sensing monitoring can easily degenerate into an archive that is updated every year and shelved every year.

V. Concepts Need Localization: Blue-Green Infrastructure Cannot Just Be a Traveling Concept

The study advocates treating urban blue space as critical infrastructure and promoting integrated management, natural drainage solutions, and routine remote sensing monitoring. This proposition is in line with recent international discussions around frameworks such as “nature-based solutions,” blue-green infrastructure, and sponge cities.The problem is that most of these frameworks were born in contexts with strong planning capacity, ample fiscal resources, and clear land tenure. When they are transplanted to secondary cities, often only the terminology remains, without the supporting conditions for implementation. At that point, “treating wetlands as infrastructure” stays at the level of planning documents, while real decisions remain dominated by the logic of land finance.

A more realistic path is not necessarily to introduce more advanced concepts, but to adjust the most basic incentive structure:

• Incorporate the ecosystem service value of water bodies and wetlands into formal asset accounting, so that they have a comparable “price” in development decisions rather than a default of zero;
• Establish a high-frequency, public, and verifiable monitoring system for changes in water surfaces, so that infilling bears visible administrative costs;
• Position natural drainage systems as insurance expenditure rather than landscaping expenditure, and assess their necessity by the expected value of disaster losses rather than construction cost;
• Establish cross-municipal coordination mechanisms at the scale of watersheds and drainage units, because water systems never respect administrative boundaries.

None of these measures is new, nor are they highly technical. The reason they are difficult to implement is precisely that they touch on definitional power and incentive structures, not engineering capacity.

VI. Long-Term Judgment: Water’s Shift from an Amenity Issue to a Security Issue

Over the next ten to twenty years, the position of urban water systems in the global urban agenda is likely to shift: it is moving from environmental and amenity issues to security and sovereignty issues.

This shift has three drivers. The first is rising climate exposure, which makes the drainage capacity of coastal and delta cities directly relevant to habitability. The second is the changing logic of capital: as climate risk is incorporated into asset pricing, land that cannot drain will face revaluation, and the “built-up areas gained by filling in water surfaces” may turn from assets into liabilities. The third is the localization of governance authority: national governments negotiate emissions reductions and climate finance, while whether a canal is retained is ultimately decided at the municipal level. In this silent way, cities are becoming the de facto implementers of global climate governance.

From this, a relatively restrained long-term judgment can be offered: what determines the success or failure of global climate adaptation may well not be the mega-cities that are repeatedly studied, heavily invested in, and frequently reported on, but the numerous, financially constrained, data-poor secondary coastal cities that nevertheless sit on the front lines of risk. Their choices will not be written into the communiqués of climate summits, but they will determine the real-world substance of those communiqués’ goals.

Khulna’s thirty years of data are therefore more like an early indicator than an isolated case. What they record is not only the disappearance of a water surface, but also that an old consensus about “what urban infrastructure is” is failing.

The order in which water surfaces are filled in is, in essence, the order in which a city ranks its own future risks.