Abstract:Urban forests are important for effectively mitigating urban heat island (UHI) effects. However, thorough investigations into how the three-dimensional (3D) structures of urban forests influences urban thermal conditions collectively and individually are limited. In this study, voxel-based landscape indices were innovatively extracted from UAV LiDAR data, and high-precision land surface temperature (LST) data were obtained using thermal infrared sensors mounted on a UAV. These were combined with a random forest (RF) model to analyze the relative influences and marginal effects of urban forest three-dimensional (3D) structure on LST. Our results showed the following: (1) The voxel-based landscape index exhibits a stronger capability to interpret LST than both the 2D landscape index and the gradient-based landscape index, with significant enhancements in model accuracy across all dimensions (an increase in R of 0.17-0.25 and a decrease in RMSE by 0.39-1.59°C). (2) Considering the vertical stratification of tree canopies, which voxel-based landscape index has the greatest LST fitting precision (R = 0.75, RMSE = 3.11°C). Including the canopy's vertical layers in analyses is pivotal, with the upper canopy layers exerting the most significant influence on reducing LSTs. (3) The scale of the grid impacts the accuracy of LST fitting, showing a trend where accuracy increases and then decreases with increasing grid scale; at the 40-m scale, the landscape indices demonstrate their highest explanatory capacity for LST (2D landscape index R=0.43, RMSE=4.65°C; gradient-based landscape index R=0.56, RMSE=4.07°C; voxel-based landscape index R=0.68, RMSE=3.94°C; vertical stratification (VS) voxel-based landscape index R=0.75, RMSE=3.30°C.). (4) Volume, proportion of volume, surface area, and diversity represent the parameters that most significantly influence variations in LST. Notably, volume, proportion of volume, and surface area exhibit a significant negative correlation with temperature, whereas diversity displays a distinct positive correlation. For the whole canopy at the optimal scale of 40 meters, a volume within 4200 m3, proportion of volume within 0.8, and a surface area within 18000 m2 are associated with a cooling effect. For the upper canopy, volume within 1200 m3, proportion of volume within 0.22, and surface area within 2000 m2 are associated with a cooling effect. This study unequivocally confirms the feasibility of using drones with LiDAR and thermal infrared sensors to analyze small-scale UHI issues. This approach is beneficial for describing the 3D structure of a forest and fitting surface temperature. Urban planners can utilize these findings in practical applications by prioritizing forest configurations with optimal 3D structures in their planning efforts to effectively mitigate UHI effects. This research provides groundbreaking methods and highly reliable data to significantly deepen our understanding of the mechanisms behind the UHI effect.

DeepLCZChange: A Remote Sensing Deep Learning Model Architecture for Urban Climate Resilience

Urban surface classification using self-supervised domain adaptive deep learning models and its application in urban environment studies

High Spatial-Resolution Classification of Urban Surfaces Using a Deep Learning Method

A WRF-UCM-SOLWEIG Framework for Mapping Thermal Comfort and Quantifying Urban Climate Drivers: Advancing Spatial and Temporal Resolutions at City Scale

Machine learning-assisted mapping of city-scale air temperature: Using sparse meteorological data for urban climate modeling and adaptation

Deep learning-based urban morphology for city-scale environmental modeling

The Impact of Urban Forest Landscape on Thermal Environment Based on Deep Learning: A Case of Three Main Cities in Southeastern China

Mapping urban form into local climate zones for the continental US from 1986–2020

The Heterogeneous Effects of Microscale-Built Environments on Land Surface Temperature Based on Machine Learning and Street View Images

Hyper‐Local Temperature Prediction Using Detailed Urban Climate Informatics

Local climate zone mapping using remote sensing: a synergetic use of daytime multi-view Ziyuan-3 stereo imageries and Luojia-1 nighttime light data

Leveraging Diffusion Modeling for Remote Sensing Change Detection in Built-Up Urban Areas

Neural Network-Based Urban Change Monitoring with Deep-Temporal Multispectral and SAR Remote Sensing Data

Enhanced Interpretation of Green Space Surface for Land Surface Temperature through a Novel Voxel-based Landscape Index from UAV LiDAR

AutoLCZ: Towards Automatized Local Climate Zone Mapping from Rule-Based Remote Sensing

Deep learning-based local climate zone classification using Sentinel-1 SAR and Sentinel-2 multispectral imagery

Spatial Optimization of Land Use and Carbon Storage Prediction in Urban Agglomerations under Climate Change: Different Scenarios and Multiscale Perspectives of CMIP6

A cellular automata model coupled with partitioning CNN-LSTM and PLUS models for urban land change simulation

Long-Term Annual Mapping Of Four Cities On Different Continents By Applying A Deep Information Learning Method To Landsat Data

Climate-conscious spatial morphology optimization strategy using a method combining local climate zone parameterization concept and urban canopy layer model

DeepCL: Deep Change Feature Learning on Remote Sensing Images in the Metric Space