Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning Framework for High-Resolution Air Temperature Downscaling Using LiDAR-Derived Urban Morphological Features

Fatemeh Chajaei, Hossein Bagheri

TL;DR

The paper tackles the challenge of obtaining high-resolution urban air temperature data by integrating LiDAR-derived urban morphology with ML-based downscaling of UrbClim outputs from 100 m to 5 m. It constructs a 3D building model (LOD1) using deep learning-based footprint extraction and transfer learning, extracts morphologic features, and trains multiple regression models, with LightGBM delivering the best absolute and relative accuracy (RMSE ≈ 0.352 K, MAE ≈ 0.215 K, R^2 ≈ 0.997). The framework yields detailed street-level temperature maps and demonstrates potential for a digital-twin urban climate model, while maintaining computational efficiency compared to physics-based downscaling. The approach, validated in Amsterdam with GitHub source code, highlights the value of fusing LiDAR-derived morphology, GIS data, and ML for practical urban climate applications and planning.

Abstract

Climate models lack the necessary resolution for urban climate studies, requiring computationally intensive processes to estimate high resolution air temperatures. In contrast, Data-driven approaches offer faster and more accurate air temperature downscaling. This study presents a data-driven framework for downscaling air temperature using publicly available outputs from urban climate models, specifically datasets generated by UrbClim. The proposed framework utilized morphological features extracted from LiDAR data. To extract urban morphological features, first a three-dimensional building model was created using LiDAR data and deep learning models. Then, these features were integrated with meteorological parameters such as wind, humidity, etc., to downscale air temperature using machine learning algorithms. The results demonstrated that the developed framework effectively extracted urban morphological features from LiDAR data. Deep learning algorithms played a crucial role in generating three-dimensional models for extracting the aforementioned features. Also, the evaluation of air temperature downscaling results using various machine learning models indicated that the LightGBM model had the best performance with an RMSE of 0.352°K and MAE of 0.215°K. Furthermore, the examination of final air temperature maps derived from downscaling showed that the developed framework successfully estimated air temperatures at higher resolutions, enabling the identification of local air temperature patterns at street level. The corresponding source codes are available on GitHub: https://github.com/FatemehCh97/Air-Temperature-Downscaling.

Machine Learning Framework for High-Resolution Air Temperature Downscaling Using LiDAR-Derived Urban Morphological Features

TL;DR

The paper tackles the challenge of obtaining high-resolution urban air temperature data by integrating LiDAR-derived urban morphology with ML-based downscaling of UrbClim outputs from 100 m to 5 m. It constructs a 3D building model (LOD1) using deep learning-based footprint extraction and transfer learning, extracts morphologic features, and trains multiple regression models, with LightGBM delivering the best absolute and relative accuracy (RMSE ≈ 0.352 K, MAE ≈ 0.215 K, R^2 ≈ 0.997). The framework yields detailed street-level temperature maps and demonstrates potential for a digital-twin urban climate model, while maintaining computational efficiency compared to physics-based downscaling. The approach, validated in Amsterdam with GitHub source code, highlights the value of fusing LiDAR-derived morphology, GIS data, and ML for practical urban climate applications and planning.

Abstract

Climate models lack the necessary resolution for urban climate studies, requiring computationally intensive processes to estimate high resolution air temperatures. In contrast, Data-driven approaches offer faster and more accurate air temperature downscaling. This study presents a data-driven framework for downscaling air temperature using publicly available outputs from urban climate models, specifically datasets generated by UrbClim. The proposed framework utilized morphological features extracted from LiDAR data. To extract urban morphological features, first a three-dimensional building model was created using LiDAR data and deep learning models. Then, these features were integrated with meteorological parameters such as wind, humidity, etc., to downscale air temperature using machine learning algorithms. The results demonstrated that the developed framework effectively extracted urban morphological features from LiDAR data. Deep learning algorithms played a crucial role in generating three-dimensional models for extracting the aforementioned features. Also, the evaluation of air temperature downscaling results using various machine learning models indicated that the LightGBM model had the best performance with an RMSE of 0.352°K and MAE of 0.215°K. Furthermore, the examination of final air temperature maps derived from downscaling showed that the developed framework successfully estimated air temperatures at higher resolutions, enabling the identification of local air temperature patterns at street level. The corresponding source codes are available on GitHub: https://github.com/FatemehCh97/Air-Temperature-Downscaling.
Paper Structure (31 sections, 4 equations, 24 figures, 7 tables)

This paper contains 31 sections, 4 equations, 24 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Study Area
  4. Datasets
  5. Urban Climate Data
  6. LiDAR Data
  7. OpenStreetMap Data
  8. Methods
  9. Data Preparation
  10. Building Footprints Extraction
  11. Deep Semantic Segmentation
  12. Transfer Learning
  13. Model Setup and Performance Evaluation
  14. Postprocessing of Detected Building Footprints
  15. Generating 3D Buildings Model
  16. ...and 16 more sections

Figures (24)

  • Figure 1: A visualization of the study area, Amsterdam, the capital of the Netherlands. The red boundary represents the city limits, the yellow rectangle represents the study area, and the blue rectangles represent the training areas.
  • Figure 2: The framework designed for downscaling of air temperature over the study area
  • Figure 3: An example of samples to show discrepancies between OSM building Layer and LiDAR ground truth. OSM building footprints are depicted as yellow polygons and LiDAR DSM is visualized as a base map.
  • Figure 4: Postprocessing on the output of segmentation to improve the quality of building footprint extraction
  • Figure 5: 3D city model generation at LOD1 using LiDAR point clouds and building footprints
  • ...and 19 more figures