Monitoring the incidence of Xylella fastidiosa infection in olive orchards using ground-based evaluations, airborne imaging spectroscopy and Sentinel-2 time series through 3-D radiative transfer modelling

Abstract

The outbreaks of Xylella fastidiosa (Xf) in Europe are generating considerable economic and environmental damage, and the spread of this plant pest appears to continue. Detecting and monitoring the spatio-temporal dynamics of the symptoms of diseases caused by Xf at large scales is key to curtailing its expansion or mitigating its impacts. This study evaluates the temporal series of airborne hyperspectral and Sentinel-2 satellite images for monitoring Xf infection incidence in olive orchards integrating satellite and airborne data with radiative transfer modelling and field observations. We used time-series of Sentinel-2A images collected over a two-year period to assess the temporal trends of Xf-infected olive orchards located in the region of Apulia, Southern Italy. First, we evaluated the sensitivity of different physiological and structural vegetation indices (VIs) to the severity and incidence of Xf-induced disease observed in situ. The same relationships were then evaluated using a 3D radiative transfer model to account for the temporal variations of canopy structure, understory and soil background that affect the spectral reflectance of Sentinel-2 over a grid-planted orchard. Hyperspectral images, spanning the same 2-year period as the Sentinel-2 data collected in the Xf-infected zone in Italy, were used for validation along with field surveys comprising more than 3000 trees across disease severity (DS) classes in 16 orchards, with varying disease-incidence (DI) levels. Among a wide range of structural and physiological vegetation indices evaluated from Sentinel-2 imagery, the temporal variation of the Atmospherically Resistant Vegetation Index (ARVI) and Optimized Soil-Adjusted Vegetation Index (OSAVI) showed superior performance for DS and DI estimation (r 2 ARVI=0.74, p<0.001).We estimated the difference in the spectral reflectance within each plot between 2016 and 2017 based on the VIs calculated from model simulations accounting for the temporal variations of the understory which confirm its impact, showing a Root Mean Square Error (RMSE) three times 2 lower than without temporal understory changes simulated. This analysis demonstrates the benefit of combining 3-D radiative transfer modelling accounting for the background variations with Sentinel-2 data to assess the spatio-temporal dynamics of Xf infections in olive orchards.The systematic retrieval of DI through model inversion and Sentinel-2 imagery can form the basis for operational damage monitoring worldwide. Furthermore, interpreting temporal variations of model retrievals is a critical step to detect anomalies in vegetation health.

Estimating the epidemiology of emerging Xylella fastidiosa outbreaks in olives

“…Detecting the presence of infected plants for surveillance or control can require considerable costly resources. The problem is further compounded when the pest is not easily detectable, such as when symptoms are not visible (Mastin et al, 2019), although new remote-sensing technologies are being developed (Zarco-Tejada et al, 2018;Hornero et al, 2020). This challenge is especially important for biosecurity of imports, because long symptomless periods may result in the movement of infected materials without the knowledge of the importer or importee.…”

“…The majority of the data came from 3 years of monitoring symptom progression in 3,335 olive trees in 16 infected olive groves containing susceptible olive cultivars (mean plot area 1.86 ± 1.57 ha, mean tree density 97.9 ± 41.7 trees/ha, typical of ancient olive groves in Puglia (EFSA Panel on Plant Health, 2019a). A subset of these data has previously been used to detect infection using airborne imaging spectroscopy and thermography(Zarco-Tejada et al, 2018), and earth observation satellite imagery(Hornero et al, 2020). The data consist of visual rating of disease severity for each tree, scored on a 0-5 canopy desiccation scale (seeZarco-Tejada et al, 2018, andHornero et al, 2020, for a full description of the data).…”

“…A subset of these data has previously been used to detect infection using airborne imaging spectroscopy and thermography(Zarco-Tejada et al, 2018), and earth observation satellite imagery(Hornero et al, 2020). The data consist of visual rating of disease severity for each tree, scored on a 0-5 canopy desiccation scale (seeZarco-Tejada et al, 2018, andHornero et al, 2020, for a full description of the data). Maximum severity scores for each tree surveyed in late 2016 (June or October), 2017 (June, July, or December), and 2018 (July) were classified into compartments of the epidemiology model.…”

Guidelines for statistically sound and risk‐based surveys of Xylella fastidiosa

“…Despite this high sampling intensity, the prevalence at first detection is 0.02 (2%) of the population (red dashed line), whereas the estimated eradicable prevalence is much lower at 0.0004 (0.04%) (blue dashed line). Calculations are based on epidemic growth rate data from olive orchards in Apulia (Hornero et al;, asymptomatic period data for olive trees (EFSA PLH Panel, 2019) and a mathematical framework linking epidemiological parameters and surveillance (Parnell et al, 2017;Mastin et al, 2017Mastin et al, , 2019 In EFSA PLH Panel (2019), based on the literature review reported in the Xylella Host Plant Database (EFSA, 2018), studies were identified which quantify the time from infection to the onset of symptoms. The data were categorised by host plant and X. fastidiosa subspecies.…”

Fast Detection of Olive Trees Affected by Xylella Fastidiosa from UAVs Using Multispectral Imaging

“…The use of terrestrial or aerial drones, both manned and unmanned, equipped with multi- or hyperspectral image cameras to study the health status of plantations of various kinds is not a novelty in precision agriculture or for forestry monitoring [ 51 , 52 , 53 , 54 , 55 ]. There are also pioneering applications of drones for the detection of Xf -infected plants [ 56 , 57 , 58 , 59 ].…”