Phenological Dynamics Characterization of Alignment Trees with Sentinel-2 Imagery: A Vegetation Indices Time Series Reconstruction Methodology Adapted to Urban Areas

Carlos Granero-Belinchon,Xavier Briottet,Karine Adeline,Aude Lemonsu

doi:10.3390/rs12040639

Abstract

This article presents a novel methodology for the characterization of tree vegetation phenology, based on vegetation indices time series reconstruction and adapted to urban areas. The methodology is based on a pixel by pixel curve fitting classification, together with a subsequent Savitzky–Golay filtering of raw phenological curves from pixels classified as vegetation. Moreover, the new method is conceived to face specificities of urban environments such as: the high heterogeneity of impervious/natural elements, the 3D structure of the city inducing shadows, the restricted spatial extent of individual tree crowns and the strong biodiversity of urban vegetation. Three vegetation indices have been studied: Normalized Difference Vegetation Index (NDVI) and Normalized Difference Red Edge Index 1 (NDRE1), which are mainly linked to chlorophyll content and leaf density and Normalized Burn Ratio (NBR) mostly correlated to water content and leaf density. The methodology has been designed to allow the analysis of annual and intra-annual vegetation phenological dynamics. Then, different annual and intra-annual criteria for phenology characterization are proposed and criticized. To show the applicability of the methodology, this article focuses on Sentinel-2 (S-2) imagery covering 2018 and the study of groups of London planes in an alignment structure in the French city of Toulouse. Results showed that the new method allows the ability to 1) describe the heterogeneity of phenologies from London planes exposed to different environmental conditions (urban canyons, proximity with a source of water) and 2) to detect intra-annual phenological dynamics linked to changes in meteorological conditions.

Highlights

The presence of vegetation improves the responsible and sustainable development of urban areas by offering several ecosystem services [1,2,3] such as: vegetation–climate interactions [4,5], energetic consumption and CO2 track reductions [6], biodiversity conservation [6], human well-being [7], ambiance and socio-cultural benefits [7]
The objective of this article is to propose a new methodology adapted to urban areas to characterize tree phenology by using Vegetation Indices (VIs) time series reconstruction from Sentinel-2 imagery
The results obtained with the S-G filtered time series are compared to those obtained with a fitting function time series reconstruction method applied on the pixels classified as vegetation

Summary

Introduction

The presence of vegetation improves the responsible and sustainable development of urban areas by offering several ecosystem services [1,2,3] such as: vegetation–climate interactions (air quality increase, urban cool island, rain water management and thermal comfort) [4,5], energetic consumption and CO2 track reductions [6], biodiversity conservation [6], human well-being [7], ambiance and socio-cultural benefits [7] Among this urban vegetation, alignment trees create shade contributing to reduce the urban heat islands and fit the geometrical urban planning [8]. Most of these previous studies work on a limited number of acquisitions along the year (often only one date) and as a consequence they cannot properly differentiate normal vegetation dynamics from a change in health condition

Objectives

Methods

Results

Discussion

Conclusion