Spatial and temporal analysis of vegetation dynamics on desertification hotspots of Ceará, Northeastern Brazil

Abstract

Seasonally dry tropical forests (SDTFs) have a distinctive seasonality. The Brazilian Caatinga, the largest SDTF in South America, faces three main types of disturbance: abrupt change, chronic disturbance, and species invasion. Furthermore, droughts and rainfall seasonality have made Caatinga at risk of desertification.

Remote sensing has brought new insight into the detection and monitoring of vegetation dynamics. This thesis advanced the use of satellite time series data from 2002-2018 for phenological pattern extraction, abrupt and gradual change detection, and resilience characterization in Caatinga.

Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) time series have been shown effectiveness in characterizing phenological patterns as well as abrupt changes and long-term trends in vegetation cover. Further analysis was implemented for breakpoint detection and long-term gradual change with the application of the Breaks for Additive Season and Trend (BFAST) algorithm, with particular attention to ‘desertification nuclei’, areas identified as at risk of desertification.

BFAST analysis using MODIS NDVI and Enhanced Vegetation Index (EVI) found that 2012 was the year with the most breakpoints. The derived ecological resilience suggested that most areas in the desertification nuclei are not recovering from desertification.

On a finer scale, BFAST application on Landsat time series found that 1) Landsat’s increased spatial resolution and observation frequency improves vegetation disturbance detection over MODIS’s; 2) increased spatial resolution does not affect the number of breakpoints detected, but it affects the detection of their timing. NDVI was found to be the with the great accuracy in Caatinga monitoring. A case study in grazing exclusion plots has emphasized the positive effects of grazing exclusion for the restoration of Caatinga vegetation.

Overall, this work has demonstrated the effectiveness of time series analysis using satellite time series on the detection and monitoring of vegetation dynamics in Caatinga with potential utility for conservation and management.

Date of Award	3 Dec 2024
Original language	English
Awarding Institution	University of St Andrews
Supervisor	Thomas Robert Meagher (Supervisor), Eimear Nic Lughadha (Supervisor) & Justin Moat (Supervisor)