The research, led by Satti J. R. Kamenan of the National Polytechnic Institute Félix Houphouët-Boigny in Yamoussoukro, is the first attempt at near real-time rainfall prediction in any of Ivory Coast’s river basins. The team built a forecasting framework around Long Short-Term Memory neural networks, a deep learning architecture specifically designed to remember patterns across long sequences of data. LSTM networks, first proposed by Sepp Hochreiter and Jürgen Schmidhuber in 1997, use internal gates—forget, input and output gates—that regulate what information is stored, discarded or passed forward through the network’s memory cells. This structure allows them to capture the long-term temporal dependencies and nonlinear behavior that characterize daily rainfall, something conventional statistical models have long struggled with.

Feeding the neural network required an ambitious data pipeline. The researchers combined ground-based rainfall records from SODEXAM, the Ivorian meteorological agency, covering ten stations in the Marahoué watershed from 1961 to 2022, with two NASA satellite products: GPM IMERG, a high-resolution precipitation dataset with a spatial resolution of 0.1 degrees and roughly 30-minute temporal resolution, and MERRA-2, a reanalysis that blends satellite observations, ground measurements and numerical weather models into consistent atmospheric fields including temperature, wind, pressure and humidity. All of this was processed on Google Earth Engine, the cloud-based platform that has become a workhorse for large-scale environmental data analysis. Before the satellite data could be trusted, the team validated it against rain gauge observations from 2000 to 2017, finding correlation coefficients of 0.80 at Bouaflé and 0.83 at Zuénoula—meaning the satellite product captured more than 80 percent of the observed rainfall variability.

Quality control was a critical and methodical step. The historical gauge records contained gaps of 7.29 to 8.04 percent, and raw observations can harbor errors that would corrupt any machine learning model. The team applied a two-stage screening protocol: first, the interquartile range method flagged values that deviated unusually far from each station’s typical distribution; second, a spatial consistency check using the network-wide median and median absolute deviation determined whether a flagged value was physically plausible or simply wrong. Crucially, extreme rainfall events that were spatially coherent across the network were retained even when they exceeded statistical thresholds, preventing the accidental deletion of genuine meteorological extremes. Only 689 of 29,510 observations—2.34 percent—were ultimately identified as outliers, and erroneous values were corrected using the validated GPM IMERG estimates as an auxiliary source.

Selecting the right predictors was equally deliberate. From twenty-one candidate atmospheric variables, the researchers used a Pearson correlation matrix to eliminate multicollinearity, keeping only variables with pairwise correlations below 0.60. They then ran a Random Forest model to rank the remaining candidates by importance, retaining twelve predictors—including seven-day antecedent rainfall, vertical atmospheric motion at 500 hectopascals, specific humidity at multiple pressure levels, and wind components—that each contributed meaningfully to predictive accuracy. The final dataset spanned 15,094 daily time steps from January 1984 to April 2024, with data through January 2020 used for training and the final four years reserved entirely for independent validation, a rigorous test of whether the models had truly learned generalizable patterns rather than memorizing the past.

Three separate LSTM models were trained, one for each forecast horizon: one day ahead, three days ahead and seven days ahead. The architecture settled on three layers of 100 neurons each, a learning rate of 0.001 with automatic reduction, a batch size of 32, and an early stopping mechanism that halted training when validation loss stopped improving. The team then pitted the LSTM against three popular machine learning benchmarks—Random Forest, Extra Trees and XGBoost—using standard metrics including the coefficient of determination, Nash-Sutcliffe Efficiency, normalized root mean square error and mean absolute error. The results were striking. At the one-day horizon, the LSTM achieved R-squared and NSE values exceeding 95 percent at both stations, with a normalized RMSE below 10 percent and a mean absolute error under 1 millimeter, placing it firmly in the ‘excellent’ category by conventional hydrological standards.

The gap between the deep learning model and its competitors widened as the forecast horizon stretched. At three days ahead, the LSTM still maintained R-squared values between 88 and 90 percent with a normalized RMSE below 10 percent, while the tree-based models saw their accuracy slip, with R-squared often dropping below 65 percent. By seven days ahead, the LSTM was the only model still producing statistically reliable forecasts, with R-squared values between 70 and 81 percent and errors remaining controlled, whereas Random Forest and Extra Trees deteriorated dramatically—mean absolute errors exceeding 7 millimeters and NSE values sometimes falling below 40 percent—and XGBoost performed worst of all. The explanation lies in fundamental architecture: tree-based ensembles treat observations largely as static feature sets, while the LSTM’s recurrent memory cells explicitly model the sequential structure and long-range dependencies inherent in daily precipitation series.

Honest scrutiny of the model’s errors revealed both strengths and limits. Residual analysis showed that predictions were largely unbiased for light to moderate rainfall, which constitutes the majority of events in the Marahoué basin, with mean residuals near zero at both stations. But the Kolmogorov-Smirnov test confirmed the residuals were not normally distributed, and a class-by-class examination exposed an intensity-dependent bias: some low-intensity events were slightly underestimated while moderate and heavy rainfall tended to be overestimated, with error dispersion growing sharply for extreme events. This is a well-documented weakness of deep learning applied to precipitation—intense rainfall is rare, irregular and heavily skewed, so models see too few examples to learn its behavior precisely, and recent studies have shown LSTM networks tend to smooth high-intensity peaks rather than reproduce them faithfully.

For the people of Bouaflé and Zuénoula, however, the practical implications are significant. Early warning systems depend less on perfectly reproducing once-a-decade extremes than on reliably tracking the ordinary rainfall regime and flagging major events in time for communities to act—and the LSTM delivers exactly that at lead times of one to three days, with usable guidance even at seven. The authors point to clear paths forward: enriching the training data with additional atmospheric predictors such as convective available potential energy, total column water vapor and sea surface temperatures over the Gulf of Guinea and ENSO regions, applying post-prediction bias correction, and exploring hybrid models that couple neural networks with physical approaches for extreme events. As climate change intensifies rainfall variability across West Africa, this study demonstrates that a developing region can build world-class forecasting capability from open satellite data, cloud computing and a well-designed neural network—no expensive radar infrastructure required.

Subject of Research: Deep learning-based daily rainfall forecasting using satellite and reanalysis data for flood early warning in the Marahoué River basin, Ivory Coast

Article Title: Intelligent daily rainfall prediction for early warning using deep learning and satellite data: application to Bouaflé and Zuénoula stations, Ivory coast

Article References: Kamenan, S. J. R., Youan, T. M., Adja, M. G., Soro, S. I., & Kouassi, A. M. (2026). Intelligent daily rainfall prediction for early warning using deep learning and satellite data: application to Bouaflé and Zuénoula stations, Ivory coast. Advances in Statistical Climatology, Meteorology and Oceanography, 12(1), 173-193. https://doi.org/10.5194/ascmo-12-173-2026

Image Credits: AI Generated

DOI: 10.5194/ascmo-12-173-2026

Keywords: deep learning, LSTM neural networks, rainfall prediction, flood early warning, GPM IMERG, satellite data, Ivory Coast, Marahoué River basin, machine learning, hydrology, MERRA-2, extreme rainfall

The research team, led by Peter Kisaakye of Makerere University in Uganda and including collaborators from institutions in Ghana and Uganda, performed a secondary analysis of data collected in 2018 and 2019. Their sample comprised 9,826 young people between the ages of 18 and 24, all drawn from surveys designed by the US Centers for Disease Control and Prevention in partnership with national governments to measure violence against children and youth. Because these surveys use household sampling frames and standardized questionnaires, they allow researchers to estimate prevalence with confidence intervals and to generalize findings to entire national populations of young adults, something that small clinic-based studies cannot do.

The headline numbers are stark. Sexual violence prevalence exceeded 20 percent in all four countries, with Ivory Coast recording the highest rate at 38.7 percent, followed by Kenya at 28.8 percent, Namibia at 23.7 percent and Lesotho at 20.5 percent. Suicide ideation, meaning serious thoughts of ending one’s own life, was highest in Kenya at 13.2 percent and lowest in Ivory Coast at 6.1 percent. Self-harm showed a different geographic pattern: it was most prevalent in Ivory Coast at 13.5 percent and least common in Lesotho at 2.8 percent. The fact that the country with the highest violence prevalence did not have the highest suicide ideation rate underscores that the relationship between victimization and suicidality is shaped by broader social and economic contexts, not by exposure alone.

Methodologically, the study relied on binary logistic regression, a statistical technique suited to modeling outcomes with two possible values, such as whether a respondent reported suicidal thoughts or not. The researchers estimated odds ratios, which express how much more likely an outcome is in one group relative to another, together with 95 percent confidence intervals that indicate the range within which the true population value likely falls. Crucially, the models were adjusted for several potential confounders, including sex, school attendance, orphanhood status, work experience, whether respondents could talk to friends, and whether they believed physical violence was justifiable. This adjustment matters because factors such as losing a parent or lacking social support are themselves linked to both violence exposure and mental health outcomes, and failing to account for them could exaggerate or distort the apparent effect of sexual violence.

After adjustment, the association between sexual violence and suicide ideation held in every country. Young people who had ever experienced sexual violence had 2.86 times higher odds of suicidal thoughts in Ivory Coast, 3.06 times higher odds in Kenya, 3.30 times higher odds in Lesotho and 2.51 times higher odds in Namibia, with confidence intervals that excluded the null value of one in all cases. The consistency of this effect across four distinct national settings is one of the study’s most significant contributions. It suggests that the psychological toll of sexual victimization is not an artifact of any single country’s data quality or cultural reporting norms, but a robust pattern that transcends borders within the region.

The link with self-harm was even stronger in some settings. Adjusted odds ratios for self-harm among survivors of sexual violence were 1.77 in Ivory Coast, 2.59 in Kenya, 2.81 in Namibia and a striking 4.75 in Lesotho, where survivors had nearly five times the odds of engaging in deliberate self-injury compared with non-survivors. Self-harm, which includes behaviors such as intentional cutting or burning, is recognized in the clinical literature as both a marker of acute psychological distress and one of the strongest predictors of future suicide attempts. The Lesotho finding, in particular, implies that in some contexts sexual violence may be intertwined with coping mechanisms that carry immediate physical danger, amplifying the case for early identification of survivors before self-directed harm begins.

Why does sexual violence exert such a profound effect on the mental health of young adults? The authors frame sexual violence as having both immediate and long-term adverse consequences, including reduced mental well-being, substance abuse and suicidality. Developmental science offers several converging explanations. The late teens and early twenties are a period of heightened vulnerability, in which identity formation, educational attainment and intimate relationships are all in flux, and the brain’s emotion-regulation circuitry is still maturing. Trauma experienced during this window can disrupt attachment, erode self-worth and instill a sense of shame that discourages disclosure. In many communities across sub-Saharan Africa, stigma surrounding sexual victimization remains intense, and survivors, particularly young men, may fear social rejection if they come forward. Silence, in turn, prevents access to the social support that the survey data themselves suggest is protective.

The study also carries implications for how mental health services are organized in low- and middle-income countries. The authors call for prioritizing sexual violence prevention and survivor-centered response, and specifically recommend integrating routine screening for sexual violence into youth mental health services, sexual and reproductive health services and primary care. This is a practical proposal with a strong evidence logic: young people are far more likely to encounter a clinic for contraception, HIV testing or a routine illness than to seek out a specialist mental health provider, and a brief, confidential screening question about violence exposure could route survivors toward counseling and crisis support they would otherwise never receive. Screening programs of this kind have shown promise elsewhere, though they require training, referral pathways and safeguards for confidentiality to avoid re-traumatizing patients.

The reliance on the Violence Against Children Surveys gives the study unusual methodological strength, but it also imposes limits that the authors acknowledge implicitly through their design. The surveys are cross-sectional, capturing a snapshot in time, which means they can establish association but not definitive causation; it remains possible that some unmeasured factor, such as family instability or prior childhood adversity, contributes to both violence exposure and suicidality. Self-reported data on sexual violence are also sensitive to disclosure bias, and stigma may lead some survivors to remain silent, which would tend to make the observed associations conservative rather than inflated. The authors tested for multicollinearity among predictors, reported in an appendix, to ensure that the adjusted estimates were not distorted by overlapping variables, a standard but important robustness check.

For policymakers across the region, the message is difficult to ignore. Suicide is among the leading causes of death for young people worldwide, and the new evidence indicates that a substantial and preventable share of the risk in these four countries is tied to sexual violence that often occurs years before the suicidal thoughts emerge. Prevention programs that address gender norms, strengthen child protection systems and engage communities have demonstrated reductions in violence exposure in trials supported by the same CDC survey infrastructure. What the present study adds is a clear quantitative bridge between that prevention agenda and mental health outcomes: every case of sexual violence averted is not only an end to a human rights violation but potentially an intervention against a lifetime of psychological suffering. As the authors conclude, sexual violence, suicide ideation and self-harm remain relatively high among young people, and treating them as separate policy problems, rather than as connected outcomes of the same trauma, risks missing the point at which help can change a life.

Subject of Research: The association between sexual violence experience and suicide ideation and self-harm among young people aged 18 to 24 in four sub-Saharan African countries

Article Title: Experience of sexual violence is associated with suicide ideation and self-harm among young people (18–24 years): evidence from a violence against children survey in four countries

Article References: Kisaakye, P., Kyei-Arthur, F., Kyei-Gyamfi, S., Afrifa-Anane, G. F., Boadu, E. S., Miezah, D., & Bukuluki, P. (2026). Experience of sexual violence is associated with suicide ideation and self-harm among young people (18–24 years): evidence from a violence against children survey in four countries. Discover Psychology. https://doi.org/10.1007/s44202-026-00892-4

Image Credits: AI Generated

DOI: 10.1007/s44202-026-00892-4

Keywords: sexual violence, suicide ideation, self-harm, young people, sub-Saharan Africa, Violence Against Children Surveys, mental health, Ivory Coast, Kenya, Lesotho, Namibia, logistic regression