EU Science Hub

Phenology of short vegetation cycles in a Kenyan rangeland from PlanetScope and Sentinel-2


The short revisit times of recent fine-resolution optical satellite sensors, such as PlanetScope and Sentinel-2, provide new opportunities to study seasonal vegetation dynamics. A few studies demonstrated that phenology can be accurately estimated with fine spatial detail from Sentinel-2, at least for relatively long annual growing seasons of more than four months. In semi-arid rangelands, vegetation green-up and senescence occur more rapidly in response to a concentration of rainfall over short periods. Obtaining a sufficient density of cloudfree acquisitions to accurately describe these short vegetation cycles can thus be challenging. This study aims to evaluate if two types of fine-resolution imagery, i.e., PlanetScope and Sentinel-2, each independently allow for accurately mapping vegetation phenology in a semi-arid rangeland. The study is conducted in a rangeland with bimodal seasonality located at the128-km2 Kapiti Farm in Machakos County, Kenya. Using all available PlanetScope and Sentinel-2 imagery between March 2017 and February 2019, we derived temporal NDVI profiles and compared these for three locations with greenness chromatic coordinate (GCC) series obtained from digital repeat photography. Double hyperbolic tangent models were fitted to the time series, separately for the two rainy seasons locally referred to as short and long rains. We estimated start and end-of-season for camera and satellite series using a 50 percent threshold between minimum and maximum levels of the modelled time series (SOS50/EOS50). The RMSD between SOS50 derived from PlanetScope and Sentinel-2 was nine days with respect to camera-derived SOS50. For EOS50 this was 16 days for Sentinel-2 and 14 days for PlanetScope. Subsequently, the spatial variability of the phenological parameters was mapped by applying the approach to all pixels within Kapiti. Our analysis shows that both PlanetScope and Sentinel-2 provide sufficient temporal detail for accurately estimating the phenology of short vegetation cycles. The performance was somewhat better for PlanetScope, showing fewer spatial artefacts. Reduced image availability due to persistent cloud cover in some seasons remains a concern to accurately estimate phenology, as we highlighted with image frequency reduction experiments. 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 spatially-heterogeneous rangeland systems with large interannual variability in phenology and productivity.