The year 2026 has been declared “International Year of Rangelands and Pastoralists” by the FAO in order to raise awareness about the fundamental role that pasture and shepherd play for both the environment and our culture. Despite the important socioeconomic role of these lands, the latest changes in the livestock sector and the demographic dynamics have caused a phenomenon of abandonment/conversion (into arable lands or building areas) in the whole Italian territory. Mitigation strategies are needed to stop this alarming negative trend, including livestock farmer’s training and, above all, the identification of innovative technological solutions aimed at promoting the rational and sustainable use of rangelands. New portable kind of NIRS readers stand out among other technologies since capable of providing a real time overview of the chemical composition of grass. This work focused on the creation of a NIRS calibration for grass (GRASS_NIRS) to be used in a next-generation NIRS reader. The model was created from 125 pasture grass samples collected between May and October 2025 on six farms within the PASCOLOVE project (Intervention SRG01 “Support for PEI-AGRI Operational Groups - Implementation Phase of the Operational Groups” funded by the Veneto Region) and two external structures. Samples were processed using chemical and spectral analyses on both fresh and dried matter. The calibrations were developed using dedicated UCAL 3.0 software (Unity Scientific, USA), which applies the Partial Least Square Regression (PLSR) method for developing predictive models. To amplify the absorbances related to the chemical composition, a first derivative treatment calculated directly by the UCAL software was applied. The prediction performance of the PLSR model on fresh grass varied significantly for the parameters considered (CP, NDF, uNDF240, UFL), with cross-validation R2 ranging from 0,386 to 0,555. Consistent with previous studies, the analysis of the same samples after drying resulted in an increase in cross-validation R2 (ranging from 0,496 to 0,777), confirming that water interferes with NIRS measurements and accuracy. These results suggest the need to identify low-cost technological solutions to enable rapid sample drying on-site before spectrum acquisition.
Il 2026 è stato nominato dalla FAO “Anno internazionale dei pascoli e dei pastori” al fine di aumentare la consapevolezza nei confronti di risorse e professioni frutto di una tradizione secolare, fondamentali per il territorio e per le culture locali. Nonostante l’importante ruolo socioeconomico dei pascoli, in Italia è da più di vent’anni che queste superfici subiscono profondi mutamenti (tra cui l’abbandono o la conversione) dettati sia da trasformazioni interne al settore zootecnico che da dinamiche demografiche. Per arginare questo preoccupante trend negativo, sono necessarie strategie di contrasto come l’individuazione di innovative soluzioni tecnologiche volte a favorire un utilizzo razionale e sostenibile delle superfici a pascolo. Tra queste spicca la possibilità di sviluppo di specifiche applicazioni NIRS da campo in grado di fornire in tempo reale una predizione della composizione chimica dell’erba. Il presente lavoro si è concentrato sulla creazione di una calibrazione NIRS dell’erba del pascolo (GRASS_NIRS) destinata ad un prototipo di strumentazione di nuova generazione. Una prima calibrazione è stata creata a partire da 125 campioni di erba del pascolo raccolti nel periodo Maggio-Ottobre 2025 in 6 aziende interne al progetto PASCOLOVE (Intervento SRG01 “Sostegno gruppi operativi PEI-AGRI – Fase di attuazione dei Gruppi Operativi” finanziato dalla Regione Veneto) e in 2 strutture esterne al progetto. I campioni sono stati sottoposti ad analisi chimiche e spettrali sia sul fresco che dopo essiccazione. Le calibrazioni sono state sviluppate con software dedicato UCAL 3.0 (Unity Scientific, USA) che applica il metodo delle Partial Least Square Regression (PLSR) per lo sviluppo dei modelli predittivi. Per amplificare le assorbanze legate alla composizione chimica, si è applicato un trattamento di derivata prima calcolato direttamente dal software UCAL. Utilizzando i campioni freschi, le performance di predizione del modello PLSR sono risultate notevolmente variabili per i parametri presi in considerazione (CP, NDF, uNDF240, UFL) con R2 di cross validazione da 0,386 a 0,555. Coerentemente con precedenti studi, l’analisi degli stessi campioni dopo essiccazione ha visto un aumento dei valori di R2 di cross validazione (da 0,496 a 0,777), evidenziando l’effetto dell’acqua nell’interferire e ridurre l’accuratezza delle letture NIRS. Questi risultati suggeriscono la necessità di identificare delle soluzioni tecnologiche a basso costo per permettere la rapida essiccazione del campione in azienda prima dell’acquisizione dello spettro.
La tecnologia a supporto dell’allevamento al pascolo. Creazione di una calibrazione NIRS per la predizione delle caratteristiche nutrizionali dell’erba
PIETROBON, MARTA
2025/2026
Abstract
The year 2026 has been declared “International Year of Rangelands and Pastoralists” by the FAO in order to raise awareness about the fundamental role that pasture and shepherd play for both the environment and our culture. Despite the important socioeconomic role of these lands, the latest changes in the livestock sector and the demographic dynamics have caused a phenomenon of abandonment/conversion (into arable lands or building areas) in the whole Italian territory. Mitigation strategies are needed to stop this alarming negative trend, including livestock farmer’s training and, above all, the identification of innovative technological solutions aimed at promoting the rational and sustainable use of rangelands. New portable kind of NIRS readers stand out among other technologies since capable of providing a real time overview of the chemical composition of grass. This work focused on the creation of a NIRS calibration for grass (GRASS_NIRS) to be used in a next-generation NIRS reader. The model was created from 125 pasture grass samples collected between May and October 2025 on six farms within the PASCOLOVE project (Intervention SRG01 “Support for PEI-AGRI Operational Groups - Implementation Phase of the Operational Groups” funded by the Veneto Region) and two external structures. Samples were processed using chemical and spectral analyses on both fresh and dried matter. The calibrations were developed using dedicated UCAL 3.0 software (Unity Scientific, USA), which applies the Partial Least Square Regression (PLSR) method for developing predictive models. To amplify the absorbances related to the chemical composition, a first derivative treatment calculated directly by the UCAL software was applied. The prediction performance of the PLSR model on fresh grass varied significantly for the parameters considered (CP, NDF, uNDF240, UFL), with cross-validation R2 ranging from 0,386 to 0,555. Consistent with previous studies, the analysis of the same samples after drying resulted in an increase in cross-validation R2 (ranging from 0,496 to 0,777), confirming that water interferes with NIRS measurements and accuracy. These results suggest the need to identify low-cost technological solutions to enable rapid sample drying on-site before spectrum acquisition.| File | Dimensione | Formato | |
|---|---|---|---|
|
Pietrobon_Marta.pdf
accesso aperto
Dimensione
3.88 MB
Formato
Adobe PDF
|
3.88 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/110217