Development of a novel UAV based approach for assessing the severity of spring frost and hail damages in vineyards

Abstract

A solid feature of climate change is that the frequency and severity of weather extremes are increasing. Ranking European countries for the number of crop failures related to extreme events reports France on top followed by Italy and Spain (COM 2021). While the severity and extent of a late frost damage is dependent upon several factors (e.g. altitude, flat vs sloped sites, distance of the vine from the ground, type of floor management, etc.), the increase of frost risk is primarily due to earlier bud burst which widens the time frame within which the probability of incurring freezing temperatures is greater. Both events share the need of appropriate damage assessment at either close time distance from the event (i.e. a few days) and later in the season for vines yield recovery estimation. Insurance agents often provide crop hail damage estimates based on their personal experience and field samples which are not always representative of the investigated field’s spatial variability. For this reason, farmers and the insurance market itself ask for a reliable, objective and less labor-intensive method to determine crop frost and hail losses. Remote sensing technologies offer significant potential to meet these needs. However, existing vegetation index-based solutions are often insufficiently specific or accurate. To address these limitations, the FROSTVINE-UAV project, funded by the European Union under the NextGenerationEU framework, aims to develop a novel methodology using 3D reconstruction of young shoots from ultra-high-resolution imagery captured by low-altitude unmanned aerial vehicles (UAVs). During the 2024 season, the first year of the project, UAV monitoring campaigns were conducted across four vineyards in Italy. These campaigns focused on developing an innovative protocol for detecting and estimating grapevine canopy damage caused by late spring frost and hail, both immediately after the events and over time to assess vine recovery and yield. Remote sensing data were validated through multitemporal ground-based observations and measurements. Preliminary results highlight a robust methodological workflow, from field monitoring to data analysis, capable of producing risk incidence maps within days of the event with an accuracy exceeding 85%. The UAV-based approach proposed in this research demonstrated significant potential to reduce operational costs for damage monitoring while enhancing the efficiency of viticulture practices.