Vegetation monitoring using multispectral sensors — best practices and lessons learned from high latitudes

Rapid technological advances have dramatically increased affordability and accessibility of unmanned aerial vehicles (UAVs) and associated sensors. Compact multispectral drone sensors capture high-resolution imagery in visible and near-infrared parts of the electromagnetic spectrum, allowing for the calculation of vegetation indices, such as the normalised difference vegetation index (NDVI) for productivity estimates and vegetation classification. Despite the technological advances, challenges remain in capturing high-quality data, highlighting the need for standardized workflows. Here, we discuss challenges, technical aspects, and practical considerations of vegetation monitoring using multispectral drone sensors and propose a workflow based on remote sensing principles and our field experience in high-latitude environments, using the Parrot Sequoia (Pairs, France) sensor as an example. We focus on the key error sources associated with solar angle, weather conditions, geolocation, and radiometric calibration and estimate their relative contributions that can lead to uncertainty of more than ±10% in peak season NDVI estimates of our tundra field site. Our findings show that these errors can be accounted for by improved flight planning, metadata collection, ground control point deployment, use of reflectance targets, and quality control. With standardized best practice, multispectral sensors can provide meaningful spatial data that is reproducible and comparable across space and time.