Abstract

The short revisit times afforded by recently-deployed optical satellite sensors that acquire 3–30 m resolution imagery provide new opportunities to study seasonal vegetation dynamics. Previous studies demonstrated a successful retrieval of phenology with Sentinel-2 for relatively stable annual growing seasons. In semi-arid East Africa however, vegetation responds rapidly to a concentration of rainfall over short periods and consequently is subject to strong interannual variability. Obtaining a sufficient density of cloud-free acquisitions to accurately describe these short vegetation cycles is therefore challenging. The objective of this study is to evaluate if data from two satellite constellations, i.e., PlanetScope (3 m resolution) and Sentinel-2 (10 m resolution), each independently allow for accurate mapping of vegetation phenology under these challenging conditions. The study area is a rangeland with bimodal seasonality located at the 128-km2 Kapiti Farm in Machakos County, Kenya. Using all the available PlanetScope and Sentinel-2 imagery between March 2017 and February 2019, we derived temporal NDVI profiles and fitted double hyperbolic tangent models (equivalent to commonly-used logistic functions), separately for the two rainy seasons locally referred to as the short and long rains. We estimated start- and end-of-season for the series using a 50% threshold between minimum and maximum levels of the modelled time series (SOS50/EOS50). We compared our estimates against those obtained from vegetation index series from two alternative sources, i.e. a) greenness chromatic coordinate (GCC) series obtained from digital repeat photography, and b) MODIS NDVI. We found that both PlanetScope and Sentinel-2 series resulted in acceptable retrievals of phenology (RMSD of ~8 days for SOS50 and ~15 days for EOS50 when compared against GCC series) suggesting that the sensors individually provide sufficient temporal detail. However, when applying the model to the entire study area, fewer spatial artefacts occurred in the PlanetScope results. This could be explained by the higher observation frequency of PlanetScope, which becomes critical during periods of persistent cloud cover. We further illustrated that PlanetScope series could differentiate the phenology of individual trees from grassland surroundings, whereby tree green-up was found to be both earlier and later than for grass, depending on location. The spatially-detailed phenology retrievals, as achieved in this study, are expected to help in better understanding climate and degradation impacts on rangeland vegetation, particularly for heterogeneous rangeland systems with large interannual variability in phenology and productivity.

Highlights

  • Rangelands are the land use with the largest spatial extent globally, and comprise various land cover types such as savannahs, grasslands, prairies, steppe, and shrubland

  • The most striking phenological difference between vegetation communities was for EOS50, which was substantially later for shrubs and trees as compared to grasses for all three seasons considered

  • It is worth noting that the trees and grass in the field of view of camera C had a short secondary green-up phase after the main green-up phase in Sep­ tember 2017–1 March 2018 (SR2017), caused by an approximate 43-day dry period in late November and December 2017 followed with substantial rainfall on 2 and 3 January 2018, and in LR2018 caused by a small rainfall event on 31 July 2018 (Fig. 4)

Read more

Summary

Introduction

Rangelands are the land use with the largest spatial extent globally, and comprise various land cover types such as savannahs, grasslands, prairies, steppe, and shrubland. Millions of pastoral households depend on their livestock for milk and meat production (Sayre et al, 2013). Their livelihoods are strongly affected by weather shocks, such as drought (Blackwell, 2010; Little et al, 2008). Climatic shifts have resulted in increasing rainfall variability in rangelands (Sloat et al, 2018), which in turn affects rangeland productivity (Briske et al, 2015; Knapp et al, 2008) and composition (Scheiter and Higgins, 2009), and the animals and humans that depend on the Remote Sensing of Environment 248 (2020) 112004 rangelands for forage and livelihoods. Factors such as carbon dioxide enrichment (e.g., Morgan et al, 2007), soil properties, herbivory, and fire regime influence rangeland status (Briske et al, 2005; House et al, 2003)

Objectives
Methods
Results
Discussion
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call