UAV-Based Remote sensing technique to detect and analyze Ink disease in a chestnut orchard using high resolution multispectral imagery

High-resolution images acquired by unmanned aerial vehicles (UAVs) have been used for several years for the analysis of vegetation cover in order to derive both spectral and structural variables. The recent technological development which has concerned the image acquisition sector has led to the commercialization of multispectral cameras characterized by 4-5 aligned acquisition channels and with a very high resolution (0.8 cm/pixel at 120 m). The availability of these images allows undertaking new types of object-based analysis (at crown level), through which it is possible to carry out analyses. Starting from the spectral characteristics of every single pixel it is possible to deduce the spectral behavior of the entire tree crown. For several years, vegetation indices have been widely used to study, extract and model biophysical variables and vegetation dynamics and define the ecological processes in progress. In the scientific literature, it is possible to find that Vegetation Indices (IVs), that can be used for this purpose, are numerous. However, in this study the most commonly used spectral index was analyzed, the Normalized Difference Vegetation Index (NDVI), because it is closely related to several ecological parameters as it is considered a measure of photosynthetic biomass. The NDVI can be considered as an indirect method of measurement of photosynthetic activity and due to its ability to be sensitive to changes in vegetation condition, it can be used with good results to detect plant symptoms related to a certain disease. In the study area, a mature chestnut orchard for fruit production, located in Castagno d'Andrea, San Godenzo (FI) affected by ink disease (Phytophthora cambivora) from decades, showing plants having different degrees of decline (severity of symptomatology), was monitored. Starting from the multispectral images acquired through UAV, an image of NDVI has been created. For a sample of chestnut plants a visual classification of ink disease symptoms was performed, the position with GNSS instrument was acquired, and then the crown has been delimited through video photo interpretation. All the features relative to crowns, through an operation of spatial overlay, have been correlated with the NDVI map and for each crown have been derived the frequency distributions of the pixels that define it. The frequency distributions obtained have been analyzed in order to investigate the different spectral behavior that characterizes the plants in relation to their symptomatology (symptomatic vs asymptomatic). This technique allowed to discriminate asymptomatic plants from symptomatic ones based on the average spectral characteristics of the tree crowns (Fig. 1).