Diagnosing hazelnut vegetation phenology and health from satellites

<p>Satellite observations represent a valuable tool for studying vegetation dynamics because they provide consistent and timely views of Earth's surface with time and space continuity, allowing for detecting changes in phenological patterns. The temporal evolution of remotely sensed (RS) vegetation indices spans the distinct plant growth stages observed from the ground, lending itself to extracting phenological metrics (RS phenometrics), such as the timing of the start, the end, and the duration of the growing season. Finding a meaningful correlation between RS phenometrics and ground observations paves the way to build a near real-time phenological monitoring system where a large amount of RS data are processed over large areas and at a low cost. We developed an analytic method to extract an ensemble of hazelnut phenological metrics from the MODIS 8-day Enhanced Vegetation Index (EVI) using the phenofit R package and to correlate them with ground observations collected during weekly surveys in 2019-2022 on 20 hazelnut orchards in Turkey (> 1500 observations). A process-based phenological model driven by chilling, forcing, and photoperiod has been calibrated using ground and RS phenometrics as reference data. Model performances in hindcast mode (2000-2019) have been evaluated concerning the reproduction of RS spring phenology over the Turkish hazelnut growing area.</p><p>Results showed a marked temporal gradient among the extracted RS phenometrics in different moments of spring vegetative development. Significant correlations have been detected considering early buds break phases and nuts cluster appearance, showing the potential of the method for predicting hazelnut vegetation phases in near-real time. The Greenup, Upturning Date, and Start of Season metrics occurred at the beginning of the onset process, in association with BBCH phases referred to female flowering, generally observed before early bud break. On the contrary, Stabilization Date, Maturity, and Peak of Season metrics marked the end of spring vegetative phenology, the former occurring in proximity to the visible nuts cluster phase. The 50% threshold and Derivative Start of Season metrics described the steepest portion of the EVI curve, i.e., the greening and thickening process of the orchard canopy, in correspondence with nuts ovary enlargement and leaves expansion. Key findings were that female flowers blooming and ending occur before the 20% development of the hazelnut vegetative process. In contrast, RS phenometrics referred to the Start of Season well aligned with leaf emergence, and the nuts cluster appearance marked the Maturity of the growing season. The calibration of the phenological model led to the accurate reproduction of ground and RS phenometrics, and its spatially distributed application revealed temporal and local phenological patterns over the main Turkish hazelnut cultivation area.</p><p>Our findings have significant implications for improving the interpretation of RS hazelnut phenology, distinguishing the timing of canopy dynamics, understanding the impact of environmental cues, and evaluating climate-forcing effects on hazelnut vegetation. The implementation of the analytical method in a near-real-time monitoring system will be presented, as well as its potential to analyze hazelnut health in response to abiotic (cold, frost, heat) and biotic (pests, diseases) stresses in sensitive phenological phases from satellite observations.</p>