Accurate assessment of climate risk is essential for informing effective adaptation strategies, particularly in regions with high exposure and socioeconomic vulnerability. However, climate risk estimates can vary depending on the conceptual frameworks and data sources used. This study examines how different approaches influence the assessment of climate risk and vulnerability across Kenya at the county level. Using a geospatial analytical framework, we integrate historical climate data (Standardized Precipitation Evapotranspiration Index, SPEI), downscaled population data, and subnational sociodemographic indicators to generate spatially explicit risk profiles. We compare estimates derived from two different IPCC supported risk and vulnerability frameworks. Additionally, we construct and compare climate risk indices based on global and locally derived social vulnerability data. Our findings indicate that while both frameworks identify a similar set of high-risk counties, the magnitude and spatial distribution of risk and vulnerability differ substantially. Climate risk estimates show greater variability across counties, whereas vulnerability-based assessments appear more uniform. In addition, risk rankings vary depending on whether global or locally derived datasets are used, underscoring the sensitivity of results to input data. Locally derived vulnerability index captures greater subnational variation in climate risk. This study contributes to the limited body of research applying the IPCC risk framework in Africa and is among the first to systematically compare risk and vulnerability frameworks in the African context within a consistent analytical approach. These findings demonstrate that climate risk is not a fixed measure but is shaped by conceptual frameworks and data inputs, with important implications for climate adaptation planning, resource allocation, and policy prioritization in Kenya and similar contexts.
Satellite land surface temperature (LST) can be linked to public facilities even when the underlying thermal observation footprint is coarser than those facilities. We examine 19,834 delivered-grid–year rows across 415 pseudonymous school identifiers (152 Barcelona; 263 Madrid), using five-fold school-grouped prediction and separate raw and within-city-relative 2014–2024 LST trends. For the available analytical panel, fixed-parameter Extra Trees produced mean held-out-school R² = 0.648 (fold SD 0.070), RMSE = 1.970 °C and MAE = 1.572 °C; other ensembles achieved similar performance. Under the declared two-sided, tie-corrected Mann–Kendall test and within-city BH-FDR adjustment, 52 Madrid-labelled raw series and no Barcelona raw series met q < 0.05, whereas one Madrid city-relative series did so. Alternative Kendall inference yielded 4/2 Barcelona and 58/15 Madrid discoveries (raw/relative); a conservative lag-1 variance-floor sensitivity retained the main counts. These computations are conditional on the supplied analytical dataset; underlying school-register and Landsat-extraction provenance remains unverified. Because native Landsat thermal support is approximately 100 m, campus-indexed summaries do not measure within-courtyard thermal variation. The study offers a bounded framework for satellite thermal screening rather than causal or fine-scale thermometric claims.
Heatwaves are intensifying globally, posing significant risks to climate-sensitive regions such as Western Maharashtra, India. This study develops an integrated analytical framework to characterize and forecast heatwave magnitude using the Heatwave Magnitude Index daily (HWMId). Annual HWMId series (1955–2025) derived from IMD gridded temperature data were analysed across six districts. Trend analyses revealed a consistent increasing trend in heatwave magnitude, particularly in Kolhapur and Pune districts. Change-point analysis using Pettit and Buishand tests identified significant structural breaks during the late 1990s to early 2000s, indicating shifts in heatwave behaviour. Pre-modelling diagnostics confirmed significant autocorrelation (Box-Pierce test) and nonlinear dependence (BDS test), justifying the use of both linear and nonlinear modelling approaches. Accordingly, ARIMA, Interrupted ARIMA, Artificial Neural Network (ANN), and Interrupted ANN models were developed within a unified framework incorporating pulse intervention variables. Additionally, hierarchical clustering based on ITA indicators was applied to identify homogeneous groups of districts with similar heatwave characteristics. Comparative evaluation indicated district-specific forecasting performance, with the Interrupted ANN model showing improved prediction accuracy for Kolhapur, while ARIMA-based models performed better for Pune. The Diebold-Mariano test supported district-specific differences in predictive performance, with stronger evidence of improvement observed for Kolhapur. The results highlight the importance of integrating structural break analysis with nonlinear modelling to capture complex heatwave dynamics. The proposed hybrid framework provides a robust and scalable approach for improving heatwave forecasting and offers valuable inputs for climate risk management, agricultural planning, and the development of early warning systems in vulnerable regions.
Urban outdoor spaces in contemporary cities face increasing challenges related to thermal habitability. These conditions are particularly relevant in high-altitude environments, characterized by intense solar radiation and considerable daily thermal variability, where outdoor thermal comfort remains insufficiently explored. In Quito, recent vertical densification has increased urban canyon height-to-width ratios and modified radiative exchanges at pedestrian level. This study evaluated the effects of densification and urban canyon morphology on outdoor thermal comfort in Quito’s hypercenter using ENVI-met and the Physiological Equivalent Temperature (PET) index. Three scenarios were compared: the current scenario (E1), the maximum buildability scenario (E2), and the climate-sensitive design scenario (E3). Simulations represented the rainy period on 23 March and the dry period on 22 September 2023, at 09:00 and 16:00. Mean radiant temperature showed the strongest association with PET, while wind speed had a secondary negative association. Under high solar exposure, PET was 5–12 °C above air temperature, particularly in east–west canyons. Vertical densification did not produce a uniform thermal response: increased shading improved some locations, while reduced airflow affected others. In E3, vegetation, shading, and complementary spatial interventions produced localized PET reductions of 3.0–3.7 °C and MRT reductions of approximately 7–9 °C at 16:00, while expanding areas within the neutral comfort range. Under the simulated conditions, the thermal performance of densification in Quito depended on interactions among morphology, solar exposure, ventilation, vegetation, and surface properties, and may inform outdoor thermal comfort planning in comparable high-altitude contexts.
Accurate forecasting of reference evapotranspiration (ETo) is vital for predicting cropland water demand and managing water resources. Public weather forecasts, however, often omit the solar radiation (Rs) required by the FAO-56 Penman–Monteith equation (PM-ETo). This study evaluated a recently developed sunshine-duration conversion method (nimp), which combines forecast temperature and weather type, by coupling its Rs estimate (Rs_nimp) with PM-ETo (ETo_Rs_nimp). Performance was compared with PM-ETo driven by a temperature-based Rs estimate (ETo_Rs_T) and by the conventional weather-type-to-sunshine-duration conversion (ETo_Rs_ncom). The evaluation used observations and archived 1–7-day public forecasts for 83 radiation stations across China during 2015–2019. Model performance was assessed using the regression slope (b), correlation coefficient (R), index of agreement (dIA), root mean square error (RMSE), and a global performance indicator (GPI). Relative to ETo_Rs_ncom, ETo_Rs_nimp increased mean R and dIA and reduced mean RMSE across lead times, and GPI rankings placed ETo_Rs_nimp first at 31
