Spatially-optimized urban greening for reduction of population exposure to land surface temperature extremes

Over half the world's population resides in cities, concentrated in hotspots covering less than 3% of Earth's land surface. This proportion is projected to increase significantly by 2050. The urbanization process dramatically alters local environments and climates, replacing natural land cover with impervious materials, which reduces albedo, heat capacity, and evaporative cooling. This leads to the formation of urban heat islands (UHIs). The combination of global warming and urbanization is exacerbating human exposure to dangerous heat, which has tripled in recent decades. This has profound impacts on human health and well-being, ranging from direct effects like heat stroke to indirect effects such as increased prevalence of infectious and allergic diseases. Effective urban planning is essential for mitigating these risks, and this requires understanding the causes of extreme heat and establishing monitoring and modeling systems at suitable spatial and temporal scales. While previous research focused on the Urban Heat Island (UHI) effect, particularly the Surface Urban Heat Island (SUHI), defining SUHI as the difference between urban and rural LST, this study aims to develop city-specific approaches for evaluating and designing urban adaptation plans to reduce population exposure to LST extremes. The SUHI metric can be influenced by the surrounding rural vegetation and is not a direct measure of urban thermal comfort. The lack of appropriate data has hampered the development of effective methods to quantify and profile these factors; however, increased availability of time series thermal remote sensing data offers an opportunity to address this gap.

Existing literature extensively explores the Urban Heat Island (UHI) phenomenon, focusing on the temperature difference between urban and rural areas. Studies have shown the relationship between urban greening and reduced LST, highlighting vegetation's role in mitigating peak summer temperatures through shading and evapotranspiration. However, tools for designing city-specific plans based on required intervention levels for achieving climate targets are limited. While the benefits of proximity to green spaces are acknowledged, challenges include competition for space and resources in urban environments. Microscale models simulating the impact of urban design and land use on the microclimate are needed. Research also indicates that other strategies such as increasing surface albedo also have positive effects on mitigating urban heat. While urban greening effectively reduces LST, tradeoffs with water resources need to be considered.

This research employs a spatial regression model to predict population exposure to LST extremes in 200 cities across diverse climates. The model uses remote sensing data, including Land Surface Temperature (LST) from MODIS, and Normalized Difference Vegetation Index (NDVI) to represent greenness. Data were aggregated at a 1km spatial resolution for the three warmest months of each year from 2010 to 2019. Exposure is calculated as the number of days/nights when LST exceeds a threshold (defined as the 90th percentile of the LST distribution over 20 years for each city) multiplied by the exposed population. A spatial lag model (SLM), accounting for spatial autocorrelation, is used due to the spatial correlation of variables. The SLM uses NDVI and distance to water bodies as predictors. Model performance is evaluated using R-squared and Mean Absolute Error (MAE) through k-fold cross-validation. The model is used to estimate the required vegetation increment for reducing exposure. Two scenarios are compared: 1) uniform NDVI increase across the entire city, and 2) focused NDVI increase in the most densely populated areas to achieve the same exposure reduction. The Köppen-Geiger climate classification system is used for climate zone analysis and data aggregation.

The analysis of 200 cities across various climate zones (Arid, Continental, Temperate, Tropical) reveals a significant increase in population exposure to LST extremes between 2010 and 2020, primarily due to urbanization. The spatial lag model shows high accuracy (R-squared ≥ 0.8) in predicting both daytime and nighttime exposure. The model indicates that increased green space and proximity to water bodies significantly reduce LST exposure. The coefficients for NDVI and distance to water bodies are negative, indicating that increases in green cover and water bodies lead to lower exposure. Simulations demonstrate that targeting high-exposure areas (most populated areas in this study) for urban greening is far more efficient than uniform treatment. For instance, in Paris, a 12% reduction in exposure requires a 44% increase in overall NDVI with uniform treatment, while targeting densely populated areas achieves the same reduction with only a 14% NDVI increment. Across all cities, increasing NDVI by 0.3 leads to average exposure reductions of 12%, 13%, 16%, and 32% for temperate, arid, continental, and tropical regions, respectively. Significantly, targeting specific areas allows for up to 70% reduction in the overall required NDVI increase compared to a uniform approach.

The findings highlight the significant role of spatially optimized urban greening in mitigating population exposure to LST extremes. The high accuracy of the spatial lag model validates its use for predicting exposure and informing urban planning strategies. The substantial efficiency gains from targeting high-exposure areas suggest a more effective and resource-conscious approach to urban greening initiatives. These results offer concrete guidance for decision-makers in developing climate adaptation strategies. The model's adaptability to any city worldwide makes it a valuable tool for urban planners and policymakers globally.

This study provides a robust methodology for assessing and predicting urban population exposure to LST extremes, demonstrating the significant benefits of spatially-optimized urban greening. The high accuracy of the spatial lag model and the efficiency gains from targeting specific areas provide concrete guidance for urban planning and climate adaptation strategies. Future research should investigate the combined effects of urban greening with other mitigation strategies (e.g., increased albedo, water bodies), explore different targeting strategies beyond the most densely populated areas, and incorporate socio-economic factors to ensure equitable distribution of green infrastructure.

The study uses LST as a proxy for population exposure to heat, acknowledging that LST may not fully capture the thermal environment experienced by individuals. The urban canyon effect and the complex relationship between emissivity and urban materials/vegetation may introduce some uncertainty. The study focuses on the 3 warmest months of the year and averages values over a ten-year period; higher temporal and spatial resolution could provide finer-grained insights. Additionally, other mitigation strategies besides greening are not considered in this study.