Floods, waterlogging, flash floods, landslides, storms, and droughts are pervasive nationwide, with 91.5% of Brazil’s 5,570 municipalities reporting at least one such disaster. Notably, 1,814 cities faced incidents linked to three or more of these factors, and 270 cities suffered from all six. Regionally, the Northeast bore the brunt in terms of affected municipalities, followed by the Southeast, South, North, and Midwest. Fatalities and economic consequences vary geographically, with the Southeast experiencing the most deaths from floods and landslides, the South most storm fatalities, and the Northeast the highest drought-related mortality.

The study highlights critical distinctions among disaster types: flooding stems from river overflow; waterlogging results from overwhelmed drainage systems; and flash floods occur due to intense rainfall over short periods. Economic damages, inclusive of direct property loss and indirect economic disruptions, are most severe in the South for floods and flash floods, the Northeast for landslides and droughts, and the Southeast for storms. Historical events illustrate these impacts vividly: in São Sebastião, São Paulo, record rainfall during Carnival 2023 caused over 60 deaths and infrastructural collapse, while catastrophic storms in Rio Grande do Sul in May 2024 affected 2.3 million people and caused more than 180 fatalities.

Despite Brazil’s recent advances in disaster data collection through platforms like S2iD and the Digital Atlas of Disasters, the study notes critical limitations. Data are self-reported by municipalities seeking federal aid, which can lead to underreporting where local agencies manage events autonomously or lack reporting infrastructure. Approximately 1,660 municipalities reportedly lack organized civil defense agencies, and many existing bodies need professionalization and resources to effectively respond to climate risks. Additional data gaps include the absence of multi-risk recording—events triggered by one disaster type often overshadow linked phenomena—and inaccuracies in attributing causes of death.

Efforts are underway to upgrade disaster databases. SEDEC’s forthcoming revisions to the S2iD platform and Disaster Atlas will adopt multi-risk frameworks and enable disaggregated data collection by age and gender, facilitating targeted responses. These enhancements, paired with technical training for local disaster managers, aim to improve data reliability and narrow disparities among civil defense agencies across Brazil.

Experts emphasize shifting focus from disaster response to proactive loss reduction. Sociologist Victor Marchezini, co-author of the study, stresses that these disasters reflect a chronic crisis intertwined with socioenvironmental vulnerabilities rather than isolated incidents. “We must move beyond accepting economic losses as inevitable,” he says, urging investment in mechanisms that reduce impacts rather than relying primarily on reconstruction efforts.

The research further integrates models predicting urban expansion, land use changes, and hydrodynamics, offering novel geographic risk assessments for flooding in urban areas driven by extreme rainfall. This fusion of climate science and urban planning provides critical tools for identifying high-risk zones within cities, informing more precise and effective mitigation policies.

By illuminating the profound intersection of climate dynamics and social management failures, this study calls for a scientific approach to disaster preparedness and public policy reform in Brazil—transforming data into decisive action to safeguard vulnerable populations and national resilience in the face of escalating extreme weather events.

Subject of Research: Hydrogeological disasters and their impacts in Brazil from 1991 to 2024
Article Title: Water-related disasters in Brazil: an assessment from 1991 to 2024
News Publication Date: 10-Apr-2026
Web References: https://iopscience.iop.org/article/10.1088/1748-9326/ae5991
Keywords: Floods, Climate change, Public policy, Natural disasters, Drought, Landslides, Storms

The increasing frequency and severity of weather-related calamities such as cyclones, torrential rains, and floods underscore an urgent need for advanced intervention methodologies. Despite these pressing circumstances, modeling potential weather interventions remains an exquisitely complex challenge. Numerical Weather Prediction (NWP) models attempt to simulate atmospheric dynamics, but because weather systems are inherently nonlinear and multi-scale, their accurate representation requires immense computational power. This complexity limits both the practicality and scalability of deploying weather control designs based purely on traditional simulation approaches.

To overcome these limitations, researchers led by Professor Masaki Ogura of Hiroshima University’s Graduate School of Advanced Science and Engineering have explored the integration of black-box optimization techniques with meteorological simulation. Black-box optimization methods treat the weather simulation models as opaque systems — where only inputs and outputs are accessible — circumventing the need for explicit gradient or internal system knowledge. This paradigm is particularly attractive for weather modeling, given the inscrutable nature of atmospheric physics and the prohibitive costs of extensive simulation runs.

The research team applied four distinct black-box optimization algorithms — Bayesian optimization, random search, particle swarm optimization, and genetic algorithms — in experimental settings that integrated real atmospheric data and sophisticated meteorological models. They employed the SCALE-RM (Scalable Computing for Advanced Library and Environment Regional Model), a powerful NWP tool developed to facilitate climate research and regional atmosphere modeling, as the computational backbone for their experiments. In two distinct scenarios termed the “warm bubble experiment” and the “real atmosphere experiment,” the team targeted the modulation of wind fields to reduce precipitation over designated geographic regions.

One particularly innovative aspect of their approach was the manner in which weather interventions were applied temporally and spatially. Interventions were tested as either one-step inputs at the simulation’s onset or multi-step inputs occurring every 600 to 3600 seconds, demonstrating the framework’s adaptability to varying temporal resolutions. This flexible implementation allowed the team to probe how incremental adjustments to atmospheric variables influence rainfall accumulation, potentially steering weather outcomes towards less destructive regimes.

Among the optimization techniques evaluated, Bayesian optimization demonstrated superior performance, efficiently navigating the vast search space of potential interventions and identifying candidate configurations that significantly curtailed rainfall levels. Remarkably, the method achieved meaningful reduction even within stringent computational constraints, highlighting its practical viability for real-world applications where extensive simulation budgets are untenable.

The efficacy of Bayesian optimization is attributed to its probabilistic modeling of search spaces and its strategic balance of exploration versus exploitation, which is crucial when simulation evaluations are costly and scarce. Furthermore, the study underscored the sensitivity of Bayesian optimization to hyperparameters, implying that careful tuning can enhance its adaptability to various atmospheric scenarios, thereby widening its applicability in diverse meteorological contexts.

This investigation not only advances the feasibility of designing weather interventions using computational optimization but also paves the way for integrating such frameworks into disaster prevention and climate engineering strategies. By achieving effective rainfall reduction with minimal trials, the approach offers a blueprint for sustainable intervention policies that respect computational and environmental constraints.

Nevertheless, the authors caution against overgeneralization, noting that their experimental conditions are limited and may not extrapolate seamlessly across broader weather intervention contexts. The ongoing challenge lies in extending the methodology to abundant and varied atmospheric conditions, unraveling the underlying dynamics that dictate algorithmic performance divergences.

Future research aims to deepen the understanding of the mechanistic interplay between black-box optimization outputs and atmospheric responses, striving to establish more robust and reliable computational infrastructures for weather control. The ultimate vision is to empower precise and computationally feasible designs that can meaningfully mitigate climate-induced disasters on a global scale.

The research also signifies interdisciplinary collaboration, blending atmospheric science, advanced mathematics, and computational engineering. The confluence of these fields fosters novel perspectives and potent tools for confronting one of humanity’s greatest challenges — the increasing menace of climate-exacerbated natural disasters.

In sum, this pioneering work by Yuta Higuchi, Yang Bai, Rikuto Nagai, Naoki Wakamiya, and Atsushi Okazaki orchestrates an elegant solution to a complex conundrum, employing black-box optimization frameworks to reimagine the feasibility of weather control. With meticulous experimentation and insightful analysis, the study marks a transformative step towards operational weather intervention, balancing scientific rigor, technical sophistication, and practical considerations.

This study was published on April 10, 2026, in the Journal of Computational Science, under the title “Development and Evaluation of a Black-Box Optimization Framework for Weather-Intervention Design.” It was supported by the Japan Science and Technology Agency (JST) Moonshot R&D Program, illustrating the strategic importance of this research within national innovation agendas.

Subject of Research: Weather intervention design and computational optimization for rainfall minimization.

Article Title: Development and Evaluation of a Black-Box Optimization Framework for Weather-Intervention Design

News Publication Date: 10 April 2026

Web References: Journal of Computational Science – DOI: 10.1016/j.jocs.2026.102850

Image Credits: Yuta Higuchi / Hiroshima University

Keywords: Weather intervention, black-box optimization, Bayesian optimization, numerical weather prediction, rainfall reduction, computational modeling, climate engineering, disaster mitigation, atmospheric simulation, SCALE-RM model