Hyperspectral Analysis for Protected Agriculture land cover mapping: A Remote Sensing Approach

Spaceborne and airborne remote sensing data serve as powerful tools for the analysis and
monitoring of both urban and agricultural territories, with diverse applications contingent upon spatial
resolution. In recent years, remote sensing imagery has been utilized for the recognition of protected
agriculture landcovers, such as greenhouses and mulch. Various studies in the scientific literature have
focused on satellite sensors like Sentinel-2 and WorldView-3, mapping the presence of protected
agriculture surfaces and implementing specific indices for recognition.
A recurrent limitation in these studies lies in the often insufficient spatial resolution of the sensors,
particularly for identifying smaller-sized greenhouses. Additionally, spectral resolution is crucial. While
some laboratory studies analyse the spectral characteristics of plastic surfaces typical of protected
agriculture, they often neglect the issue of mixed pixels inherent in satellite or aerial detection.
The aim of this study is to analyze images from the AVIRIS airborne sensor over the agricultural area of
Salerno in southern Italy. AVIRIS, a hyperspectral sensor with over 400 bands covering the visible (VIS) to
the shortwave infrared (SWIR) region, provided images with a spatial resolution of 1m and 3m. We
scrutinize these images to discern the spectral signatures of different types of greenhouses in the study
area, subsequently comparing them with other land cover classes. For this, we employ supportive tools,
including specific spectral indices and transformations such as Tasselled Cap and Principal Components
Analysis (PCA). We implement the Region of Interest (ROI) separability technique to identify distinctive
spectral features in the signatures of protected agriculture coverings that differentiate them from other
surfaces. Finally, the spectral signatures obtained from AVIRIS offer the opportunity to simulate spectral
responses of other satellite sensors with lower spatial and/or spectral resolutions, assessing the suitability
of currently available data for recognizing this specific type of surface.