The 2024 floods and the responsibility to revise cities’ hydrological references

Updating a reference in the face of new evidence is not technical insecurity; it is quality control of the model.

The diagnostic coordinated by the ANA (Brazil's National Water and Basic Sanitation Agency) described the episode as the greatest natural disaster in the history of Rio Grande do Sul and recorded failures of distinct natures, among them ruptures in floodgates, backflow in culverts, dikes below their design elevation, and infrastructure fragility. This evidence demands technical examination, but it does not authorize shortcuts: a climate percentage cannot be applied indiscriminately to any basin, and an observed maximum does not, on its own, become a new design elevation. Brazilian legislation reinforces prevention, mapping, monitoring, and periodic review; abroad, regulators and large engineering firms have been combining climate scenarios, gray, green, and blue measures, and communication of residual risk. The result is not a new "universal elevation," but a more defensible process for making decisions in a territory that has changed.

What 2024 changed, and what it did not

The publication by the Technical Advisory Group coordinated by ANA brought together universities, research institutions, and public agencies to reconstruct the event. The report points to roughly 2.4 million people affected across 478 municipalities and 183 deaths, and treats the flood as the result of a combination: exceptional precipitation, antecedent conditions, basin response, urbanization, the performance of dikes, floodgates, and culverts, and the exposure of people and assets.

The event added rare observations to series that, in many locations, are short or discontinuous, which may alter a model's calibration or reveal that its conceptual structure is incomplete. Even so, an observed maximum does not become a "new design flood" without analysis: it is necessary to know the quality of the record, the vertical datum, and the mechanism that generated that level. Nor should a local trend be inferred from a single episode alone: the flood is decisive evidence for testing the system, but future frequency requires a broader set of data, hypotheses, and scenarios.

A hydrological reference is not a single elevation

In public debate, observed high-water mark, flood extent, regulatory elevation, design level, and protection standard often appear as if they were synonyms. They are not, and the distinction is indispensable for the review to produce a technically coherent decision:

• Observed flood mark: empirical evidence that the water reached a given position at a certain moment; it requires survey, time, datum, and quality control.
• Mapped flood extent: the extent interpreted from field data, imagery, or a model, which may combine different moments and uncertainties of classification and terrain.
• Design flood: discharge, hydrograph, or level associated with a probability criterion, scenario, and design purpose, calculated with a declared method.
• Implementation or safety elevation: an engineering and regulatory decision that may incorporate design level, freeboard, waves, settlement, uncertainty, and consequence of failure.
• Residual risk: risk that remains after dikes, reservoirs, drainage, floor elevation, or other measures, including through exceedance, failure, or inadequate operation.

Hydrology estimates how precipitation and basin characteristics convert into discharge and volume; hydraulics calculates how that flow interacts with channels, floodplains, bridges, and structures, producing levels and flood extents. Updating only the design rainfall does not correct an outdated topography; revising the channel geometry does not resolve a poorly estimated frequency. For this reason, the review needs to locate the origin of the discrepancy.

Return period is probability, not a calendar

Under the usual stationary formulation, a 100-year return-period event corresponds to a 1% probability of exceedance in each year, not to an event scheduled to occur once per century. Assuming constant probabilities and independent years, the chance of at least one exceedance over 30 years is approximately 26%; two events close together in time are not, in themselves, a statistical contradiction.

The difficulty arises when the distribution can no longer be treated as invariant. Urbanization, straightening of watercourses, new dams, and global warming may alter magnitude, frequency, or the rainfall-runoff relationship, a condition the literature calls nonstationarity. Milly et al. drew attention to the limitation of using the historical record as an immutable envelope; Salas and Obeysekera demonstrated that, in this context, risk and return period need to be formulated over the service life, with probabilities that may vary over time. This does not render history useless nor require abandoning stationary methods; the prudent response combines observed series, trend tests, scenarios, and engineering judgment, documenting what each method can or cannot support.

Why rainfall alone does not explain the elevation

The same depth of precipitation can produce different responses depending on antecedent moisture, permeability, slope, and land use. In urban areas, drainage capacity and surface overflow routes become decisive; in fluvial and lacustrine systems, downstream levels, wind, tide, and reservoir operation may control the water line. The joint technical note by INMET, Cemaden, Cenad, and INPE recorded the influence of multiple basins and of conditions capable of hindering flow toward Lagoa dos Patos, while the ANA report documented protection failures. Thus, an extreme elevation may result from a rare discharge, an adverse boundary condition, a system failure, or the combination of these factors.

This framing also matters for accountability: it is not technically correct to attribute every divergence to "climate" before testing data, works, maintenance, and land use, nor is it correct to ignore climate change when there is evidence of a change in the hazard. The analysis must separate causes, contributions, and uncertainties.

Five layers of a technically defensible review

Reviewing does not mean merely running the same file again with a larger rainfall. It means auditing the chain of evidence and decision in connected layers:

1. Evidence and series: verify the consistency of rain gauges and stream gauges, gaps, station changes, rating curves, flood marks, timing, vertical datum, and metadata.
2. Territory and system: update the digital terrain model, bathymetry, roughness, channels, dikes, culverts, pumps, floodgates, land use, and sedimentation.
3. Hydrology: revise basin delineation, losses, response times, IDF relationships, regionalization, rainfall-runoff, frequency, trends, and climate scenarios.
4. Hydraulics: test boundary conditions, backwater, coupling between network and surface, routing, overflow routes, and performance in 1D, 2D, or combined models according to scale.
5. Decision and risk: relate probability, consequence, service life, regulatory criteria, freeboard, uncertainty, residual risk, monitoring, and adaptation triggers.

Calibrating with 2024 without forcing the model

Field-surveyed marks, level and discharge series, Cemaden flood polygons, and mappings by the Geological Survey of Brazil constitute a valuable base, but before comparing them it is necessary to reconcile vertical references, timing, and uncertainties: a datum difference may look like an elevation error. Where a model already existed, 2024 can serve as an independent test; when the event is used for calibration, other episodes should support the validation. Adjusting roughness or boundary conditions until a single mark is reproduced, without testing other locations or events, produces overfitting: the model replicates the chosen past but may fail in extrapolation.

The Flood Risk Atlas of Rio Grande do Sul, published in 2025 by ANA together with state agencies and the Geological Survey of Brazil, illustrates an institutional update: it revised the 2014 national atlas with new methodology and data from the last decade. The lesson is not that every map must be redone after any rainfall, but that the validity of a reference depends on the currency of the data and the materiality of the new evidence.

Climate change is not a magic percentage

The attribution analysis by World Weather Attribution, with the participation of Brazilian scientists, estimated that human-induced climate change made the episode's rainfall approximately twice as likely and increased its intensity by between 6% and 9%; El Niño also made a relevant contribution. These results concern the event, the region, and the models analyzed: they are not equivalent to a new municipal IDF equation, nor do they justify applying that percentage directly to any design discharge.

The IPCC indicates that intense precipitation and urbanization amplify flood risk, and the World Meteorological Organization (WMO) recorded 2024 as a year of high-impact hydrometeorological extremes in Latin America and the Caribbean. These assessments support the need to test future conditions, but the translation into design must consider basin, horizon, scenario, and uncertainty.

International regulatory agencies offer useful methodological references, such as climate allowances that vary by basin, season, and percentile, working with central, upper, and sensitivity scenarios. The value for Brazil lies in the logic: regionalize, make scenarios explicit, and test sensitivity, and not in the direct import of foreign percentages.

Resilience is not reduced to a larger structure

Increasing the design return period or the cross-section of a culvert may be necessary, but it does not, on its own, resolve a system with power failure, backflow, floodplain occupation, or lack of maintenance. The World Bank recommends combining gray, green, and blue infrastructure; the IPCC highlights the role of retention areas and sustainable drainage, without eliminating the need for conventional protection. Large international consultancies apply this integration by combining floodable parks, engineered defenses with natural basin management, and hydrodynamic forecasting systems; these examples demonstrate that data, operation, and urban uses need to function as a system.

Nature-based solutions also have limits of area, soil, and performance in extreme events, and their contribution must be quantified at the relevant scale. Resilience is the capacity to reduce damage, maintain essential functions, and adapt, not the promise of zero risk.

In Brazil, reviewing is a matter of governance

The updated version of Brazilian Law No. 12.608/2012 (Brazil's National Civil Protection and Defense Policy Law) adopts the hydrographic basin as the unit of analysis, prioritizes prevention based on studies, assigns to municipalities the identification and mapping of risk areas, and states that uncertainty does not preclude preventive measures. The law includes risk analysis in environmental licensing, with preventive measures, monitoring, and contingency plans for qualifying developments.

Brazilian Law No. 14.904/2024 (Brazil's law on urban climate adaptation) determines that climate risk be considered in sectoral and land-use planning policies. Adaptation plans must be grounded in scientific evidence and undergo review every four years; this cycle refers to the adaptation plans and does not create a universal deadline for each hydrological model or private study.

The concrete obligation to update a study depends on the instrument, the authority, local legislation, and the risk of the development. The technical argument is broader: when a decision continues to rest on a premise materially contradicted by new data, there must be at least a documented screening to determine whether review is necessary.

How this changes environmental and development studies

A hydrological study or a component of an EIA (Environmental Impact Assessment), EIV (Neighborhood Impact Study), licensing, or drainage design should allow another specialist to reconstruct the logic adopted. To this end, it is advisable to organize, at a minimum:

1. Question and criterion: define whether the decision involves land use, drainage, floor elevation, protection, or risk to third parties, and record the applicable requirement.
2. Source and scale of the hazard: distinguish river flooding, pluvial flooding, flash flooding, or a combination, delineating the basin and area of influence.
3. Data and territory: list series, period, gaps, topography, and limitations, checking land use and changes upstream and downstream.
4. Calibration and validation: present events, goodness-of-fit metrics, and discrepancies that remained unexplained.
5. Scenarios and uncertainty: compare the current condition, future land use, and climate, testing parameters and making limits and residual risk explicit.
6. Update triggers: define when a new event, structure, or change of use will require screening or a full review.

The counterpoint: not every study needs to be rebuilt

A responsible review also avoids excess. A lot far from watercourses and with no material effect on drainage does not require the same effort as a critical infrastructure on a floodplain protected by a dike; depth should grow with exposure, consequence, and the project's capacity to alter the risk.

The screening may conclude that the existing data remain adequate, and this decision must also be recorded: which datasets were verified and why 2024 does not alter the relevant mechanism. Traceability protects against two opposite errors: retaining an obsolete reference out of inertia, or replacing a consistent reference in reaction to a single number.

Conclusion

The 2024 floods showed that the reliability of a hydrological reference depends not only on the sophistication of the original calculation, but on the quality of the data, the representation of the territory, and the transparency with which uncertainty and residual risk are communicated. Reviewing does not mean declaring that every prior study was wrong, but recognizing that models are representations conditioned by information, purpose, and time. Resilient cities do not work with the illusion of a definitive elevation; they work with clear criteria, monitoring, adaptation, and room for water.

Technical and framing note

This article is technical and informational in nature. It does not establish an elevation, return period, climate percentage, modeling method, or universal review frequency. The applicable requirements depend on the legislation, the competent authority, the Terms of Reference, the purpose, the scale, the risk, and the location. The probability figures presented assume annual independence; in nonstationary processes, risk must be formulated over time. Climate attribution results do not replace local hydrological analysis, and international references require calibration to the Brazilian context.

Sources consulted: ANA/GTA RS (diagnostic of the floods in Rio Grande do Sul); Flood Risk Atlas of Rio Grande do Sul (ANA, 2025); Sistema Nacional de Informações sobre Recursos Hídricos (ANA/SNIRH); Cemaden/MCTI (Nota Técnica No. 469/2024); Serviço Geológico do Brasil (emergency mapping of risk areas); INMET, Cemaden, Cenad, and INPE (technical note on geo-hydrological risks); MCTI and World Weather Attribution (climate attribution study); Brazilian Law No. 12.608/2012 and Brazilian Law No. 14.904/2024; IPCC (AR6 WGII, chapter on cities and infrastructure); World Meteorological Organization (State of the Climate in Latin America and the Caribbean 2024); World Bank (urban flood management); Milly et al. and Salas & Obeysekera (scientific literature on hydrological nonstationarity); international technical publications on urban water resilience. The text of this article is an original synthesis by LZ Ambiental.