Green Roof Plants Monitoring via Hyperspectral Sensors and Machine Learning Algorithms

Green roofs play a key role in urban stormwater management and environmental sustainability, and their performance strongly depends on vegetation health. Hyperspectral sensing provides detailed spectral information for monitoring vegetation conditions, but its high dimensionality poses challenges for data analysis. This study investigates the use of machine learning techniques to classify vegetation health status from hyperspectral data acquired during a ground-based monitoring campaign on a full-scale green roof located at the University of Calabria (Italy). Two complementary dimensionality reduction approaches were considered: feature selection based on minimum redundancy maximum relevance and feature extraction based on principal component analysis. These methods were integrated with four supervised classifiers—linear discriminant analysis, k -nearest neighbors, support vector machines, and random forest—and evaluated under two data partitioning strategies: a spatial split with training and testing data collected at distinct locations and a random split. The results show that support vector machines and random forest provide the most robust and accurate classification performance across all plant species. In particular, high accuracy values (up to approximately 0.95) were achieved using a limited number of features (15–20 selected wavelengths) or only three principal components, demonstrating that most of the relevant information is concentrated in a reduced feature space. The spatial split configuration highlights the ability of the models to generalize under varying environmental conditions, while the random split yields higher overall accuracy due to more homogeneous data distribution. The findings confirm that machine learning approaches significantly improve vegetation health classification compared to traditional vegetation indices by exploiting the full hyperspectral signature. Moreover, the study provides practical insights into the application of machine learning techniques for real-time monitoring of green roof systems under Mediterranean climate conditions.

Authors

Institutions

Publication Details

Journal
Journal of Water Resources Planning and Management
Published
2026-11-18
DOI
https://doi.org/10.1061/jwrmd5.wreng-7238
Primary Topic
Remote-Sensing Image Classification
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Green Roof Plants Monitoring via Hyperspectral Sensors and Machine Learning Algorithms

Patrizia Piro, Michele Turco, Monica Moroni
3 links
Journal of Water Resources Planning and Management
Remote-Sensing Image Classification
article

Green Roof Plants Monitoring via Hyperspectral Sensors and Machine Learning Algorithms

Patrizia Piro, Michele Turco, Monica Moroni
article en

Abstract

Green roofs play a key role in urban stormwater management and environmental sustainability, and their performance strongly depends on vegetation health. Hyperspectral sensing provides detailed spectral information for monitoring vegetation conditions, but its high dimensionality poses challenges for data analysis. This study investigates the use of machine learning techniques to classify vegetation health status from hyperspectral data acquired during a ground-based monitoring campaign on a full-scale green roof located at the University of Calabria (Italy). Two complementary dimensionality reduction approaches were considered: feature selection based on minimum redundancy maximum relevance and feature extraction based on principal component analysis. These methods were integrated with four supervised classifiers—linear discriminant analysis, k -nearest neighbors, support vector machines, and random forest—and evaluated under two data partitioning strategies: a spatial split with training and testing data collected at distinct locations and a random split. The results show that support vector machines and random forest provide the most robust and accurate classification performance across all plant species. In particular, high accuracy values (up to approximately 0.95) were achieved using a limited number of features (15–20 selected wavelengths) or only three principal components, demonstrating that most of the relevant information is concentrated in a reduced feature space. The spatial split configuration highlights the ability of the models to generalize under varying environmental conditions, while the random split yields higher overall accuracy due to more homogeneous data distribution. The findings confirm that machine learning approaches significantly improve vegetation health classification compared to traditional vegetation indices by exploiting the full hyperspectral signature. Moreover, the study provides practical insights into the application of machine learning techniques for real-time monitoring of green roof systems under Mediterranean climate conditions.

Journal of Water Resources Planning and ManagementVol. 153(2)
University of Calabria (IT), Sapienza University of Rome (IT)
Remote-Sensing Image Classification
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon