Hyperspectral proximal sensing and machine learning techniques to estimate wheat nutritional status for digital agriculture application

Global megatrends (climate change, population growth, technological change) have gradually caused the supply-demand balance to shift towards a not sufficient and unsustainable food production, with a potentially dramatic consequence for environmental and humanitarian aspects [1]. We are forced to "produce more with less" protecting the most important production factors (i.e. soil and water) and reducing impact to environment (e.g. pollution and greenhouse gases emission) (FAO 2016). For this reason, the development of diagnostic tools able to support farmer towards rational nitrogen (N) management, based on the actual crop requirements, is among the most challenging issues to target European Policy Frameworks and it is a central topic of the National Center for the Development of New Technologies in Agriculture (Agritech). In this context, geo-information products from Remote sensing (RS), able to quantify within-field crop status variability, are fundamental solution to support site-specific N management [1,2]. In particular, a new era of hyperspectral RS (HRS) system, on ground, from UAV platform or satellite (e.g. ASI-PRISMA mission) is opening new opportunity for quantitative crop traits monitoring (LAI, chlorophyll and Nitrogen content) [3]. Appropriate methods must be developed to fully exploit the HRS data information content to generate information useful to smart agricultural management.