Complex spatial patterns of leaf phenology within semi‐arid African savanna landscapes driven by rainfall and topo‐edaphic variation revealed using satellite remote sensing

Abstract

Leaf phenology, the timing of seasonal leaf life‐cycle events from bud‐burst to leaf‐drop, is sensitive to climate change, so essential to monitor. The regularly‐repeated measurements over broad scales in satellite remote sensing (RS) approaches are useful to this end. However, the phenological traits of African savanna species and communities remain to be fully described, and there remains uncertainty regarding how to employ RS data to identify these. This research sought to address both issues, using south‐central African study landscapes.
Plant functional types (PFTs) are essential to broad‐scale ecosystem characterisation, and a
robust set of leaf phenological types (PFlpTs) would facilitate RS monitoring. However, PFlpTs are poorly defined in south‐central Africa. Woody and grassy elements co‐dominate savannas, and, while grass phenology tracks rainfall events, woody plants, which have deeper roots and can store water internally, can retain leaves well into the dry season and leaf prior to the start of the rainy season. There is wide variation, though, and, here, nine PFlpTs and five well‐defined sub‐types were identified using field measurements of traits in common woody species. Major vegetation types supported specific kinds of PFlpT and turnover occurred at fine spatial scale, indicating high beta diversity. To fully characterise leaf phenology in African savannas, RS monitoring needs to identify leafing events (loss and gain) in the late dry season and early rainy season with at least half‐monthly temporal resolution and account for complex, fine resolution spatial pattern in the woody layer.
Moderate Resolution Imaging Spectroradiometer (MODIS) RS data have been widely
employed in broad‐scale studies, but, for savannas, green‐up dates in MODIS time series do not always correlate with field observations. In thinly wooded savannas, leafing events can be patchy and localised within the relatively coarse resolution MODIS 250 m pixels and possibly too weak to identify in time series, especially given inherent imaging error. Earliest greening dates in each of 225 sample stands across a semi‐arid savanna landscape with wide natural variation in woody cover identified using finer resolution European Space Agency (ESA) Sentinel‐2 10 m and Planetscope PS2 3 m imagery was compared with dates estimated using daily MODIS time series composited to 16‐day periods. Analysis‐ready MODIS vegetation index products overestimated earliest greening, which was much reduced by manual compositing. Thirty‐five different pre‐processing treatments combining approaches to limiting imaging error were tested. When earliest greening was widespread across stands, estimated dates were within one compositing period (16 days) of the dates observed. However, early greening in thinly wooded savannas was missed altogether.
The spatial patterns of green‐down and green‐up in the late dry season and early rainy
season were modelled across a diverse 482 km2 semi‐arid landscape using Sentinel‐2 10 m data and found to be complex relative to topography, geology, and soil texture. Greening prior to the start of the rainy season is of interest as recent MODIS RS studies identifying it as ubiquitous across south-central Africa contradict field observations in semi‐arid savannas, in particular. Daily rainfall records from a network of 22 gauges were used to model the start of the rainy season and pre‐rain greening was found to be rare, patchy, and highly localised. Furthermore, green‐up was lagged across much of the landscape, even more than a month after rain first fell, reflecting fine‐scale variation in plant moisture availability relative to relative to cumulative rainfall he beginning of the rainy season and soil texture. Precipitation is redistributed by gravity and stored into the dry season in drainage lines and, as expected, late dry season greenness prevailed in drainage lines, except in the sandstone hills, where it occurred on interfluves at the highest available positions, favouring unusual moisture
absorption via leaves. While early greening occurred in drainage lines on gneisses, very late greendown typified drainage lines on heavier soils on the Basalt Plain.

Date of Award	3 May 2024
Original language	English
Awarding Institution	Edge Hill University
Supervisor	PAUL APLIN (Director of Studies), JOAQUIN ALBERTO CORTES CARRILLO (Director of Studies), ANNE OXBROUGH (Supervisor) & Bruce Clegg (Supervisor)