作者：周敏  

2026年7月6日21时，横州市娘山水库水位已达88.15米且仍在上涨。当地水利部门直到深夜23时才发布通告，决定于次日凌晨2时30分开闸泄洪，要求横州镇、莲塘镇通知涉及的村庄群众远离河道。从发现险情到实际泄洪，留给下游群众的反应时间三个半小时，这短短200分钟是在深夜——手机信号本就容易在暴雨中中断，村民酣睡，村干部逐户敲门通知的效率可想而知。

而在几十公里外的六蓝水库，情况更为凶险。据下游村民梁微转述，8点通知泄洪，9点多堤坝就塌了——从通知到决堤，几十分钟而已。何等仓促。

六蓝水库坝体最终出现两处溃口，总长约50米，洪水直冲云表镇下辖的邓圩、振兴、黄榃、亚陂等多个村庄，部分区域水淹二楼。据南宁市应急管理局通报，此次洪涝灾害受灾人口约5.5万，转移疏散4.8万，另据多方消息，横州因灾死亡、失联人数持续攀升。

同时发生了两个灾难：泄洪与决堤。前者是有预案、有闸门控制的主动排水，后者是坝体本身失去控制的被动溃决。横州这次的实际情况，是官方通报的”漫顶、决堤”——也就是说，是一次没能及时调洪、最终酿成溃坝的险情。把泄洪当作事后解释的挡箭牌，掩盖不了大坝本身管理和预警的失败。

六蓝水库其实早已被官方点名列为问题水库，且有村民反映，险情发生前一天水库已经蓄满了。一座已知存在隐患的中型水库，承担着十几万亩农田灌溉任务、下游又是人口密集的乡镇村落，为什么没有在汛期来临前完成除险加固？为何直到坝体开裂、水位逼近极限，才仓促决定开闸？这些问题比半夜泄洪更值得深究——失责不单单发生在泄洪的那一瞬，而是隐藏在此前数月甚至数年的隐患排查与投入之中。

近些年，从郑州7·20地铁隧道进水，到涿州为保京畿而分洪淹没大片农村，几乎每一次极端降雨引发的重大伤亡，事后复盘都会指向同一组问题：预警发布滞后、基层通知不到位、工程设施带病运行、决策链条冗长。郑州事故后，国务院调查组认定其为重大生产安全责任事故，处理了数十名相关责任人。但公众普遍的感觉却是，处理停留在行政处分层面，没有对决策链条更高层级的追问，也未能转化为可验证、可监督的制度改进。涿州分洪的争议则更进一步：为保护上游城市而牺牲下游农村的决策逻辑本身有其防洪工程学上的合理性，但决策过程是否充分论证？受灾群众的知情权和补偿是否到位？至今缺乏透明的公开说明。

这次横州，同样的剧本，同样的家园尽失、生灵涂炭。

古代治水同样依赖人力和经验，技术条件远远逊于今日，但问责机制要直接得多。比如明代，对河防官员实行严格的岁修考核，堤防失事导致伤亡，主管官员依情节轻重要被杖责、革职乃至更重的惩处。且这种问责通常发生在事故之后不久，责任人是谁、受了什么处分，是可以公开追溯的。这套制度当然有其时代局限，但它至少建立了一条清晰的因果链：谁主管、谁负责、出了事谁承担。

今天的问责机制在技术能力上远胜古人——遥感监测、水位预警系统、应急指挥体系一应俱全。但在责任落地这一环，却比古代更加模糊。灾害发生后，通报中充斥着”迅速启动应急响应”、”全力搜救转移”这类程序性表述，偏偏很少看到对”为何险情发现到预警发布之间存在数小时空窗”、”为何已知隐患水库未能提前加固”等具体问题的正面回应。技术进步了，但透明度和问责的具体性没有同步提升，这才是问题。

这次横州灾情中，还有一个值得记录的现象是：大量最初的求救信息，来自身处异地、通过社交平台为家乡父母求助的普通网民，而不是官方来主动通报。云表镇一位居民的父母被困在了两层自建房顶楼，是女儿在社交平台发帖求助后，才联系上了应急部门，得到的回应却是：救援行动已经展开，但由于洪水涨势过猛，呼吁市民先自行前往高处——这是不是把最初的自救责任推回给了受灾群众本身？与此同时，本地民间自发组成的救援队伍，在断网断电、道路被淹的情况下持续搜救转移群众，成为实际救援力量的重要补充。

但民间救援长期面临的困境是：缺乏统一协调、没有官方授权往往寸步难行，装备、信息、人力都要自行解决，一旦深入险情区域出现意外，责任归属也不明确，这使得很多本可以更快抵达现场的民间力量，不得不在”救不救、怎么救”之间反复权衡。一个健康的应急体系，理应是官方力量与民间力量的协同，而不是让民间在审批空白与责任真空中自行摸索。如果政府救援迟缓，又不能为民间救援提供便利通道和明确规则，受灾群众实际上是两头落空。

批评这类事件，不是简单地情绪宣泄，而是要把几个具体问题铺开在桌子上，倾听一个清晰的答案：

六蓝水库作为已知的”问题水库”，此前的除险加固计划执行到了什么程度？谁对此负责？

从险情预警到实际泄洪/决堤之间的时间窗口，是否符合应急预案的最低标准？如果不符合，是谁在哪个环节延误了？

民间救援力量在此次灾害中发挥了怎样的作用？官方是否应当、以及将如何为民间救援提供更明确的协作机制？

此次灾害的死亡、失联人数最终核实结果如何？相关责任人是否会被问责，问责结果是否会公开？

这些问题的答案，悬而未决。灾情在笔者打字的同时还在发展中。一些说法（包括社交平台上流传的、尚未获得官方或主流媒体证实的很多求救、遇难细节）仍需要时间和更多信源去核实。但可以确定的是，如果这次事件的问责最终仍旧止步于几份谨小慎微的通报，那么下一次暴雨时，会不会又一个深夜发出的泄洪通知，把同样的悲剧再来一遍。

编辑：李晶  校对：周敏 翻译：周敏

Guangxi Hengzhou Reservoir Dam Failure Flood — Repeated Dereliction of Duty on a Land of Suffering

Author: Zhou Min

Abstract: The Hengzhou reservoir dam failure incident in Guangxi once again exposes the low social governance and emergency response capabilities of the CCP authoritarian government. Power without accountability and supervision will inevitably disregard the life and death of the people.

At 21:00 on July 6, 2026, the water level of Niangshan Reservoir in Hengzhou City had reached 88.15 meters and was still rising. The local water conservancy department did not issue a notice until 23:00 late at night, deciding to open the gates to discharge floodwaters at 2:30 the next morning, requiring Hengzhou Town and Liantang Town to notify the residents of the involved villages to stay away from the river course. From the discovery of danger to the actual flood discharge, the reaction time left for the downstream masses was three and a half hours. This short 200 minutes was in the dead of night — mobile phone signals were already prone to interruption during heavy rain, villagers were fast asleep, and the efficiency of village cadres knocking on doors house by house to notify them can well be imagined.

In the Liulan Reservoir dozens of kilometers away, the situation was even more dangerous. According to the narration of downstream villager Liang Wei, the notice to discharge floodwaters came at 8 o’clock, and the dam collapsed at past 9 o’clock — from the notice to the breach, it was only a matter of tens of minutes. How hasty. The dam body of Liulan Reservoir eventually sustained two breaches, with a total length of about 50 meters. The flood rushed straight towards several villages under the jurisdiction of Yunbiao Town, including Dengxu, Zhenxing, Huangtan, and Yabei, with some areas flooded up to the second floor. According to a briefing by the Nanning Municipal Emergency Management Bureau, this flood disaster affected a population of about 55,000 and evacuated 48,000. In addition, according to multiple sources, the number of dead and missing people in Hengzhou due to the disaster continues to climb.

Two disasters occurred simultaneously: flood discharge and dam breach. The former is an active drainage with contingency plans and controlled by gates; the latter is a passive collapse where the dam body itself loses control. The actual situation in Hengzhou this time is the “overtopping and dam breach” as reported in the official briefing — that is to say, it was a danger that failed to regulate the flood in time and ultimately resulted in a dam failure. Using flood discharge as a shield for post-event explanations cannot cover up the failure of the management and early warning of the dam itself.

Liulan Reservoir had in fact long been explicitly named and listed by officials as a problem reservoir, and some villagers reported that the reservoir was already full the day before the danger occurred. For a medium-sized reservoir known to have hidden dangers, which bears the irrigation task of hundreds of thousands of mu of farmland and has densely populated towns and villages downstream, why was the elimination of dangers and reinforcement not completed before the arrival of the flood season? Why was the decision to open the gates made so hastily only when the dam body cracked and the water level approached the limit? These questions deserve deeper investigation than the midnight flood discharge — the dereliction of duty does not merely occur at the moment of flood discharge, but is hidden in the hidden danger inspection and investment during the preceding months or even years.

In recent years, from the Zhengzhou July 20 subway tunnel flooding to the Zhuozhou flood diversion to protect the capital region which inundated vast rural areas, almost every major casualty caused by extreme rainfall points to the same set of problems in post-event reviews: delayed issuance of early warnings, inadequate notification at the grassroots level, facilities running with defects, and lengthy decision-making chains. After the Zhengzhou accident, the State Council investigation team identified it as a major production safety responsibility accident and penalized dozens of relevant responsible persons. However, the general feeling among the public was that the handling remained at the level of administrative sanctions, lacking questioning of higher levels of the decision-making chain, and failing to transform into verifiable and supervisable institutional improvements. The controversy over flood diversion in Zhuozhou went a step further: the decision-making logic of sacrificing downstream rural areas to protect upstream cities has its own rationality in flood control engineering, but was the decision-making process fully demonstrated and debated? Were the affected masses’ right to know and compensation in place? To this day, a transparent public explanation is lacking.

This time in Hengzhou, it is the same script, the same loss of homes, and the same devastation of life.

Ancient flood control also relied on human labor and experience, with technical conditions far inferior to those of today, but the accountability mechanism was much more direct. For example, in the Ming Dynasty, strict annual maintenance assessments were implemented for river defense officials. If a dike failure resulted in casualties, the chief official would be flogged, dismissed, or even subjected to harsher punishments depending on the severity of the circumstances. Moreover, such accountability usually occurred shortly after the accident; who the responsible person was and what punishment they received could be publicly traced. This system certainly had its historical limitations, but it at least established a clear chain of causality: whoever is in charge is responsible, and whoever is responsible bears the consequences if something goes wrong.

Today’s accountability mechanism is far superior to that of the ancients in terms of technical capability — remote sensing monitoring, water level early warning systems, and emergency command systems are all available. However, in the link where responsibility lands, it is more ambiguous than in ancient times. After a disaster occurs, briefings are filled with procedural expressions such as “rapidly initiated emergency response” and “sparing no effort to search, rescue, and transfer,” yet they rarely see direct responses to specific questions such as “why there was a gap of several hours between the discovery of danger and the release of early warnings” and “why the known hidden-danger reservoir failed to be reinforced in advance.” Technology has progressed, but transparency and the specificity of accountability have not improved simultaneously, which is the problem.

In this Hengzhou disaster, another phenomenon worth recording is: a large amount of the initial information seeking help came from ordinary netizens who were located elsewhere and sought help for their parents in their hometown through social platforms, rather than being actively reported by officials. The parents of a resident in Yunbiao Town were trapped on the top floor of a two-story self-built house. It was only after the daughter posted a plea for help on a social platform that contact was made with the emergency department. The response received was: rescue operations had already commenced, but due to the rapid rise of the flood, citizens were urged to first head to high ground on their own — is this pushing the initial responsibility of self-rescue back to the affected masses themselves? At the same time, local civil rescue teams spontaneously formed, continuously searching, rescuing, and transferring the masses under conditions of cut-off network and power, and flooded roads, becoming an important supplement to the actual rescue forces.

However, the long-term dilemma faced by civil rescue is: lacking unified coordination and without official authorization, they often find it difficult to move a single step; equipment, information, and manpower must all be resolved by themselves. Once they go deep into dangerous areas and accidents occur, the attribution of responsibility is also unclear. This makes many civil forces, which could have arrived at the scene faster, repeatedly weigh between “whether to rescue and how to rescue.” A healthy emergency system should rationally be the synergy between official forces and civil forces, rather than letting the civil sector grope on their own in administrative blanks and responsibility vacuums. If government rescue is slow and cannot provide convenient channels and clear rules for civil rescue, the affected masses actually fall between two stools.

Criticizing this type of event is not simple emotional venting, but rather laying out several specific questions on the table to listen for a clear answer:

To what extent had the previous elimination of dangers and reinforcement plan for Liulan Reservoir, as a known “problem reservoir,” been executed? Who is responsible for this?

Does the time window between the danger warning and the actual flood discharge/dam breach comply with the minimum standards of the emergency plan? If it does not comply, who delayed in which link?

What kind of role did civil rescue forces play in this disaster? Should the authorities, and how will they, provide a clearer collaboration mechanism for civil rescue?

What is the final verified result of the number of deaths and missing persons in this disaster? Will the relevant responsible persons be held accountable, and will the results of accountability be made public?

The answers to these questions hang in the balance. The disaster situation is still developing while the author is typing. Some statements (including many details of seeking help and casualties circulating on social platforms that have not been confirmed by official or mainstream media) still require time and more sources of information to verify. But what is certain is that if the accountability for this incident ultimately stops at a few cautious briefings, then during the next heavy rain, there might be another midnight flood discharge notice repeating the same tragedy all over again.

Editor: Li Jing Proofreader: Zhou Min Translator: Zhou Min