Temporal resolution trumps spectral resolution in UAV-based monitoring of canopy senescence dynamics

Abstract

Senescence is a dynamic process that is affected by many environmental, genetic, and physiological factors. Quantifying this process is important for breeding wheat varieties with high yield and of high quality. We present a method that allows up-scaling of the state of the art method - visual scoring - by using image sequences acquired from Unmanned Aerial Vehicles (UAV). This reduces measurement time and environmental changes during the measurement as well as rater bias. We compared the potential of a widely used multispectral sensor and a cheaper high-resolution RGB camera to track the dynamics of senescence. A UAV each was equipped with one of these sensors and used to measure canopy reflectance throughout the senescence process that lasted several weeks, for more than 400 winter wheat cultivars across three field seasons. Multiple spectral and RGB indices were calculated at the experimental plot level and used to model the dynamics of senescence. Model fits were further processed to extract key time points of the senescence phase. By comparing the results of the two sensors with each other and with the visual evaluation, respectively, we show that both sensors allow monitoring of senescence dynamics and measure key time points of the phase with a precision close to that of more sophisticated proximal sensing approaches. Optimal timing of measurements proved to be more important than the choice of sensor, confirming that timely and frequent measurements should be prioritized over more expensive sensors that provide a higher spectral resolution.