1 — Introduction
Typhoid fever, caused by Salmonella enterica serovar Typhi, remains a persistent public health challenge in low- and middle-income countries, where inadequate water, sanitation, and hygiene (WASH) infrastructure facilitates fecal–oral transmission (World Health Organization, 2023). Globally, the disease causes an estimated 9.2 million cases and 133,000 deaths annually, with the highest incidence concentrated in South and Southeast Asia (GBD 2023 Causes of Death Collaborators, 2025). Despite the availability of an effective typhoid conjugate vaccine (TCV), seasonal outbreaks continue to burden healthcare systems in endemic settings, disproportionately affecting children under 15 years and socioeconomically marginalized communities (Antillón et al., 2025; Mogasale et al., 2016).
Nepal is a landlocked Himalayan country with diverse ecological zones, ranging from the flood-prone Terai plains to high-altitude mountain regions vulnerable to landslides. In this setting, seasonal monsoons, rapid urbanization, and socioeconomic inequalities contribute to recurring typhoid outbreaks that strain limited health resources and especially affect children under 15 years (60–70% of cases) and marginalized communities in low-income districts (Karkey et al., 2018; Ministry of Health and Population, Government of Nepal, 2023). Nepal records approximately 80,000–100,000 clinically diagnosed typhoid cases annually through its Health Management Information System (HMIS), with incidence rates as high as 351 per 100,000 in urban Kathmandu (Coalition Against Typhoid, 2024). Climate change is intensifying these pressures: Nepal is warming at approximately 0.056 °C per year — nearly twice the global average — and is experiencing more frequent and severe monsoon floods (Intergovernmental Panel on Climate Change, 2022; Ministry of Forests and Environment, Government of Nepal, 2021). Flooding disrupts WASH systems by contaminating drinking water sources and breaching sanitation infrastructure, creating environmental conditions conducive to Salmonella Typhi persistence and transmission (Bhandari et al., 2022; Levy et al., 2023).
While the epidemiology of typhoid in Nepal has been documented through hospital-based surveillance and vaccine effectiveness studies (Karkey et al., 2018; Shakya et al., 2019), the influence of climate-induced extreme events on spatiotemporal disease patterns at the district level remains poorly characterized. Prior studies have primarily focused on urban settings or national-level estimates, with limited integration of high-resolution climate data, flood records, or subnational heterogeneity (Bhatt et al., 2019; Pitzer et al., 2025). Traditional statistical approaches, including autoregressive models and generalized linear models, struggle to capture the nonlinear and lagged relationships between climate variability and disease incidence that are characteristic of monsoon-driven epidemiology (Choi et al., 2023; Hess et al., 2020).
Advances in machine learning (ML) offer new opportunities to address these limitations. Tree-based ensemble methods such as Random Forest and XGBoost, as well as recurrent neural networks such as LSTM, can simultaneously model nonlinear interactions, lagged predictor effects, and spatiotemporal heterogeneity in a way that conventional regression approaches cannot (Choi et al., 2023; Dixon et al., 2023). However, very few studies have applied ML to typhoid forecasting in Himalayan or flood-prone settings, and none have combined HMIS district-level surveillance with satellite-derived climate metrics and disaster event records to generate forward-looking predictions for Nepal (Go, 2025; Ogunleye et al., 2023).
Objectives
This study addresses these gaps by examining the spatiotemporal relationship between climate-induced flooding and typhoid fever incidence across all 77 districts of Nepal from 2015 to 2023. The specific objectives are:
- To describe the spatiotemporal distribution of typhoid cases and flood events across ecological zones.
- To quantify associations between typhoid incidence and hydro-meteorological variables, including lagged flood effects.
- To develop and compare ML predictive models (Random Forest, XGBoost, LSTM, and an ensemble).
- To generate scenario-based projections of typhoid burden under the SSP2-4.5 and SSP5-8.5 climate pathways through 2050.
Research question
To what extent do climate-induced flood events, together with precipitation anomalies and relative humidity, influence the spatiotemporal patterns of typhoid fever incidence across Nepal?
This question is especially pressing given that Nepal’s TCV rollout and ongoing WASH investments may be undermined by climate-exacerbated floods, which could amplify national cases by 25–40% by 2050 under high-emissions scenarios (Intergovernmental Panel on Climate Change, 2022; Ministry of Forests and Environment, Government of Nepal, 2021). The Terai lowlands — home to over 50% of Nepal’s population and disproportionately affected by both flooding and poverty — face the greatest structural deficits in safe water access and improved sanitation, making them the highest-priority area for evidence-based climate-health intervention (Asian Development Bank, 2024; Central Bureau of Statistics and UNICEF, 2019). The findings contribute to Nepal’s National Adaptation Plan (2021–2050) and WHO’s typhoid control strategy by providing district-level, climate-informed risk estimates that can be operationalized within existing digital health platforms such as DHIS2.