Esposito, Marco (2023) Use of new weed management approaches and artificial intelligence to obtain neutral weed communities. [Tesi di dottorato]

Anteprima

Testo Esposito_Marco_35.pdf Download (2MB) | Anteprima

Abstract

It is time to identify new weed management solutions to reduce the reliance on herbicides and intensive soil tillage. The number of herbicide-resistant weed species and the environmental risks associated with using herbicide and soil tillage are increasing. Understanding the ecological and biological interactions between crops, weeds, and other biological agents is critical to point out new strategies for managing weed communities. This Ph.D. dissertation is divided into four chapters. In the first chapter, we propose the concept of neutral weed communities, which are weed communities that can coexist with the crops without significantly reducing the crop yield and quality compared to weed-free plots. We provide scientific evidence showing the presence of neutral weed communities in many agricultural contexts and the ecological principles explaining why weed communities can be neutral. Finally, we provide two approaches to obtaining neutral weed communities. The first approach aims to maximize weed biodiversity to decrease the intensity of the niche overlap and the functional diversity between weeds and crops. The second approach relies on artificial intelligence, machine vision, and robots able to discriminate and selectively manage the growth of specific weed species. The second approach aims to establish weed communities shaped based on the ecology of the crops. The second chapter shows the effect of different nutritional treatments (low, optimal, and surplus) on the formation and differentiation of weed communities and their impact on wheat yield and kernel quality. Indeed, out of the four weed communities identified under optimal and low nutrient conditions, two communities were neutral. Only one detrimental weed community was identified under surplus nutrition. The highest weed species richness under optimal and low nutrition might have significantly increased the differentiation grade of weed communities. In this chapter, we highlight the necessity to rationalize using fertilizers to shape weed community composition and increase their neutrality. The third chapter shows the effect of weed competition in combination with salt stress on the agronomical and physiological performances of green beans. When green bean plants were subjected to weed competition and salt stress showed higher yield losses than those found under the single stress (salinity or weed). The leaf morphological and physiological adaptations under weed competition compromised their response to salt stress, negatively affecting crop productivity at the end of the growth cycle. The fourth chapter gives an overview of sensor technology and the use of drones in agriculture to manage weed communities, decreasing the chemical inputs. Unmanned aerial and terrestrial vehicles may be an essential tool to obtain neutral weed communities shaped based on the ecology of the crops. Indeed, artificial intelligence algorithms may process weed images provided by drones, allowing weed species-specific recognition. This procedure significantly reduces herbicide use in the cropping systems, decreasing the probability of more competitive and resistant weed species and environmental risk. More research is needed to improve the sensors' ability to acquire weed spectral information and the artificial intelligence algorithm's capacity to recognize weed species at different phenological stages.

Downloads

Actions (login required)

Modifica documento