Monitoring spring phenology of individual tree crowns using drone‐acquired NDVI data

Abstract

Quantifying the timing of vegetation phenology is critical for monitoring and modelling ecosystem responses to environmental change. Phenological processes have been studied from landscape to global scales using Earth observing satellite data, and at local scale by in situ surveys of individual plants. Now, data acquired from multi-spectral sensors on drone platforms provide flexible opportunities for monitoring phenology from individual plants to small ecosystem scales efficiently, allowing community and species level information to be derived. We captured a time-series of drone-acquired normalized difference vegetation index (NDVI) data with a multi-spectral sensor (Parrot Sequoia, (Parrot, France)) over a highly heterogeneous ecosystem in Cornwall, UK, during a period of spring green-up. We monitored NDVI trajectories at the individual crown and species' level. For deciduous crowns, we derived metrics representative of spring phenological stages: Start-of-spring (SOS), middle-ofspring green-up (MOG) and start-of-peak greenness (SOP) using a logistic function. While the exact timing of SOS, MOG and SOP appeared susceptible to understorey effects and saturation of the NDVI, relative timing of green-up for a subset of species was plausible in relation to phenological observations from an extended geographic region and in situ plant area index (PAI) measurements. In evergreen vegetation (Pinus spp.) subtle changes were also detected through the growing season. The impact of illumination differences was analysed for image pairs during leaf-off and leaf-on conditions. While significant, these effects were small (mean absolute NDVI deviation of up to 0.034 for leaf-off, 0.013 for leaf-on conditions), meaning that data captured under both constant direct and diffuse irradiance conditions can be used together and that cloudy conditions should not lead to data gaps. We conclude that the capability of drone-mounted multi-spectral instruments for spatio-temporal characterization of crown-level phenology shows great promise for improving the understanding of intra-and inter-species differences in strategy, and offers an efficient means of doing so over areas of a few hectares.

Detection of aspen in conifer-dominated boreal forests with seasonal multispectral drone image point clouds

“…The period during which aspen has no leaves but birch does is short and is probably under the influence of annual and geographic variation. Consequently, phenological monitoring via, for example, cameras mounted in place (Zeng et al 2020) or drones (Fawcett et al 2020) is required to not miss the optimal time for inventories.…”

“…Leaf flush, leaf expansion and maturation, coloration and senescence patterns alter the structural and spectral values of deciduous tree leaves and canopies (Boyer et al 1988;Blackburn and Milton 1995), including aspen (Keskitalo et al 2005). In addition to intraspecific patterns of leaf and canopy characteristics, there are also interspecific seasonal ecophysiological differences between deciduous tree species (Key et al 2001;Fawcett et al 2020). For example, in boreal Europe the timing of leaf flush in birches is earlier than that of aspen (Heide 1993) which can help in their separability (Dymond et al 2002).…”

“…The use of drones appears to be especially promising for the collection of time-series data for species classification. Drones can provide a high temporal resolution and present a high capability of detecting phenological dynamics from the level of individual trees to that of small ecosystems (Berra et al 2019;Fawcett et al 2020). Multispectral drone image point clouds have been frequently employed for species classification (Nevalainen et al 2017;Franklin and Ahmed 2018;Xu et al 2020), but are often limited to Red, Green, and Blue band (RGB) images with one near infrared channel (Lisein et al 2015;Michez et al 2016).…”

Remote Sensing Analysis of the Changes in Vegetated Coastland: The Case of Southeastern Europe

“…In situ surveys of individual plants can be used, besides Landsat and EO satellite imagery. LiDAR data, national cadastral forest data and multi-spectral sensors data on drone platforms (e.g., Fawcett et al 2020). Such data can be considered for this study, if these data are available.…”

Mask R-CNN and OBIA Fusion Improves the Segmentation of Scattered Vegetation in Very High-Resolution Optical Sensors

“…Remote sensing studies based on very high-resolution images have increased in the last years (e.g., [ 24 , 25 , 26 , 27 ]), partly because of the availability of Google Earth images worldwide [ 28 , 29 , 30 ] and the popularization of unmanned aerial vehicles (UAV). Although these images have shown a high potential for vegetation mapping and monitoring [ 31 , 32 , 33 ], two main problems arise when they are used.…”