A machine learning approach to generate daily synthetic NDVI and LST products for urban scale SUHI and heatwaves assessment

Isaac Buo, Valentina Sagris, Jaak Jaagus

Zenodo (CERN European Organization for Nuclear Research) · 01 January 2021

In one sentence. Random forests generate gap-free daily 100 m and 500 m NDVI and land surface temperature products to support urban heatwave and heat-island assessment.

Abstract

The frequency of extreme climate events like heatwaves is increasing because of the global climate warming. Heatwaves are devastating and create discomfort. Advances in remote sensing provide an avenue to monitor and assess the impact of heatwaves. Land surface temperature (LST) is an important parameter in studying heatwaves and urban climates. The normalized difference vegetation index (NDVI) is also a key parameter in studying urban heat islands. Researchers acquire these parameters mainly from remote sensing imagery. However, some limitations make the available data not suitable for the heatwave studies at an urban scale. Such limitations are poor temporal resolution, coarse spatial resolution and cloud contamination. Heatwaves occur within a short period, and it is important to have data of adequate temporal and spatial resolution to monitor and study it at urban scales. This paper implements a machine learning approach to generate daily NDVI and LST for assessing surface urban heat islands and heatwaves at two suitable resolutions (100 m and 500 m). The proposed approach models the NDVI of MODIS and Landsat 8 at one kilometer with the random forest (RF) algorithm using land use and temporal features. Also, we modeled the LST with the random forest algorithm using land surface types and spatio-temporal features. Further on, we applied models to predict the daily MODIS-like and Landsat 8-like NDVI and LST products at desired resolutions. We validated the outputs from the models using products from Landsat 8 from two days. The generated products are free of gaps and are useful for studying temporal and spatial variability of LST during a heatwave.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.