Land degradation and desertification threaten ecosystems, biodiversity, and human well-being, especially where anthropogenic pressures and climate change converge, often driven by unsustainable land use and environmental mismanagement. Given its cascading effects on soil fertility, water resources, and vegetation - essential ecosystem services for human livelihoods, including socioeconomic impact, addressing desertification is one of the main critical issues by reversing the Land Degradation phenomena through regeneration actions to help achieve Land Restoration and reach the Land Degradation Neutrality (LDN) Global objectives. Regenerative actions are placed within a complex context made by interconnected phenomenon requiring a dynamic, multiscale and multi-temporal approach to capture constantly evolving processes that derive from trials-and-errors and progressive ecosystem adaptation. To govern this complexity, the primary research objective of this thesis has been to develop a methodological workflow to shift from global Land Degradation Assessments toward G-local Land Restoration Monitoring, to support the effectiveness of landscape regeneration strategies at a local level, taking in account the specificity of interrelated variables, to improve MultiActor decision making and guide regional up-scalability. Land Use Change is becoming increasingly crucial because it contributes to decarbonization policies under the LUC/LULUCF (Land Use Change/Land Use, Land Use Change, and Forestry) framework, which uses Earth Observation data to monitor global trends such as deforestation rates, carbon stock changes, and greenhouse gas emissions. The thesis aims to complement LULUCF global monitoring by implementing a finer-scale monitoring approach - farm by farm, hectare by hectare - to support regenerative actions. The expected return is at a Global scale, as IPCC guidelines (2006, 2013, 2019; AR6 WGIII 2022) indicate measurable ecosystems benefits deriving from correctly implementing regenerative practices - such as minimal tillage or perennial crop succession - impacting soil quality, vegetation cover, and overall resilience. The successfulness depends on local decisions and actions within a global framework. The Thesis proposes a Shifting from Regional Land Degradation Assessments to Local Regenerative Agriculture Monitoring uniting top-down global objectives with bottom-up local data components (G-Local). To achieve this primary objective, the research is grounded in the Mediterranean Desertification and Land Use (MEDALUS) methodology and Land Use, Land-Use Change, and Forestry (LULUCF) framework and aims to develop a scientifically grounded methodological workflow to transition the global framework for assessing land degradation into a localized, context-sensitive system for monitoring land restoration activities. To achieve this, the study builds upon, and advances, two foundational technical baselines identified: Baseline 1 - the Geospatial Data Management System (GDMS), and Baseline 2 - Earth Observation (EO). The refinement of these baselines streams from local context relevance and analytical precision, identifying cross-relations, lacking points and possible enhancement. By uniting these baselines, the system overcomes the limitations of LULUCF automated classification - particularly in heterogeneous, mixed-use landscapes - by integrating a Local Land Use Classification (LoLUC) through Phenological (EO) and Local-Scale Analysis (GDMS). The overall data stream structures resulted in a Local Land Use Change Regenerating Earth “LoLUC-REarth” an integrated MultiTemporal and MultiDimensional System, to monitor complex micro-scale granular trends and regenerative context-based scenarios. Instead of relying on broad global classifications, the methodology introduces “real” polygons for micro-experimental plots, accurately delineated from on-site surveys collected over three years. By examining fractions of a hectare under differing management regimes - traditional, organic, and regenerative - while systematically integrating ground-truth data, the approach captures a highly granular picture of land-use changes and evolving ecological processes. LoLUC-REarth has been tested across different climatic, morphological, and geographical settings with varying degrees of desertification, ensuring its adaptability and robustness. The study’s first applied demonstration of the system’s potential adaptability takes place in Basilicata (Italy) (first case study), a region characterized by diverse topographical and hydrological conditions that compound risks associated with land degradation. The second case study application tested the methodology in Murcia Region, Spain. Specifically, in Altiplano Camp, part of Ecosystem Regeneration Community Network. The Camp is an experimental area where place-based regeneration actions (crop rotation, perennial rotation) allowed to monitor the results deriving from ecosystem restoration, meanwhile controlling them with on-site soil experts (Regeneration Academy). The third case study is in Nicosia District, Cyprus and represents a riverine infrastructure, the Klirou-Malounta-Akaki Dam. The Klirou case represents a typical case of river watershed renaturation, specifically of buffer zones along river streams, providing an opportunity to test, analyze and monitor the effects of dam construction on spontaneous ecosystems. Collectively, these case studies facilitated the implementation of each specific research enhancement and contributed to progressively structure the LoLUC-REarth workflow, thus enabling the test and refinement of possible interactions between global tools, local EO-based monitoring and Geospatial Data Management System (GDMS). The first research Enhancement (a) derived from the application in Basilicata (Italy) and restructured the Mediterranean desertification and land use system (MEDALUS) GDMS to facilitate transitioning from a desertification sensitivity analysis to a local-scale data-driven regeneration monitoring infrastructure. For this purpose, a geospatial system has been implemented: accordingly, the system’s indicators (Soil Quality, Vegetation Quality, Management Quality, and Climate Quality) were restructured to include micro-level data, and progressively evolved to incorporate remote sensing data, through cloud computing (Google Earth Engine). In this way, the new GDMS indicators shift the attention from land degradation, toward supporting the monitoring of local-scale LUCs deriving from regenerative practices, fostering multidisciplinary cooperation to monitor positive outcomes. To enable re-scaling to local, the research proposes a second Enhancement (b), based on Enhancement (a) – which provided the baseline GDMS infrastructure tailored to Mediterranean regions. The enhancement (b) envisions its direct implementation in operational tools and systems for monitoring Land Restoration: LoLUC-REarth. To do that, LoLUC-REarth evolves the GDMS toward an adaptive system capable of supporting Local-scale data integration and EO data streams, opening to integrate advancements from remote sensing technologies (i.e., Google Earth Engine - GEE), and local-scale survey data. To integrate remote sensing, and field data to track localized regeneration dynamics into the GDMS, the research defines the parameters for a detailed local survey methodology, the “LoLUC-REarth Survey Structure”. The LoLUC-REarth Survey Structure is based on Arena FAO Data collection Platforms principles and introduces a spatial grid system and survey validation areas, to support the integration of local Scale Indices deriving from Field Observations and Laboratory Analyses. R Studio supports Data Management and Analytics and moreover, allows to align EO metrics with on-field measurement. In this way the Survey Structure itself can support to integrate r and Google Earth Engine (rGEE). The applicative use case to structure the Survey and populate the GDMS was the living laboratory at the Altiplano Camp. In this setting, field validation data - collected by soil experts, agronomists, and local stakeholders - were combined with LULUCF EO-derived metrics to populate and refine the local-scale MEDALUS-based GDMS. This unified workflow enables point-level analyses at sub-hectare scales, an essential capability when small plot variations can vastly influence soil quality and vegetation outcomes. At the heart of Enhancement (b) lies the EO segment with phenological modeling of ecological succession to monitor phenological trends (greening, browning and no trend). Using rGEE (R-Google Earth Engine) scripts, Landsat 5, 7, 8 and Copernicus Sentinel-2 imagery are processed for phenological analyses. The phenological modelling involves large temporal resolution time-series analysis of key Vegetation Indictors such as the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) to detect start-of-season (SOS) and end-of-season (EOS) and identify transitions in vegetative cycles (ΔT) to guide on-site interventions. The system tracks multiple vegetative peaks within the same year, capturing complex coexisting plant communities. For instance, a single parcel might show overlapping signals from perennial crop succession, no-till legumes that enhance soil carbon retention, improved evaporative activities, and Mediterranean hedges designed to preserve soil moisture. By streamline workflow interconnections field validation data - combined with EO-derived metrics - are fed into the LoLUC-REarth GDMS and enable point-level sub-hectare analyses using precisely delineated micro-experimental plots. Although the thesis primarily focuses on agricultural settings, the application to contexts dominated by man-made interventions and large-scale infrastructures, such as the Klirou Dam large- highlights the potential framework’s adaptability in scenarios where spontaneous revegetation of external taxa occurs under changed hydrological patterns. Results proved that LoLUC-REarth can handle multiple forms of revegetation, suggesting the methodology’s future potential in managing watershed forestry planning and other non-cultivated domains. The research builds upon the insights and frameworks established in the initial phases and designs a system that can be scaled, adapted, and iteratively improved over time. As for that, The Research enhancement (c) proposes an Iterative refinement plan which identifies two main on-field data integration: on one side, photogrammetric surveys (UAV), which have been addressed to generate 3D landscape textured models to refine the delineation of experimental plots, detect subtle topographical features affecting runoff and infiltration; on the other side, Leaf Area Index ground measurements which are going to future iteratively refine the vegetation trends to discern thresholds of species composition. The results evidenced how LoLUC-REarth workflow allows to analyze the EO in relation to the local scale indicator, and vice versa, thus providing for each micro experimental plot the quantitative outcomes deriving from soil conservation, crop rotation efficacy, and vegetation recovery. The EO phenological model detected multiple vegetation trends and pixel-based ecological succession per year, reflecting mixed species growth peaks, deriving from implementing practices like perennial crops, no-till legumes, and soil moisture-preserving hedges. Moreover, by actively involving local stakeholders in data-gathering, the system captured knowledge about traditional, organic, and regenerative cultivation methods. This inclusive process enabled evidence-based interventions, guiding decisions about land management improvements while also deepening societal engagement. The envisioned LoLUC-REarth System creates a unified and adaptive framework for managing regenerative land restoration, aligning large-scale geospatial insights with site-specific, context-aware implementation. The system supports continuous, multi-scale tracking of land-use regeneration dynamics and restoration efforts, effectively bridging the gap between remote sensing analytics and field realities. The research stands ready for further refinements and applications, transforming traditional desertification risk assessments into an active, data-driven system, envisioning a digital twin infrastructure. The creation of a LoLUC-REarth GIS-based testing platform is a foundational result toward achieving the vision of a G-Local Digital Twin for rural landscape management and ecological restoration. As for that, the GIS-based testing platform provided access to evidence-based insights, empowering local stakeholders to act with evidence-based strategies. Future developments foresee the LoLUC-REarth parameterization to up-scale from individual plots to infra-regional territories (e.g., the AlVeLal community in Spain), thereby supporting adaptive, regenerative strategies and climate resilience. Its scalable and replicable approach will also contribute to developing shared guidelines for farmers and land managers to use in infra-regional regeneration experimental areas.
Il degrado del suolo e la desertificazione minacciano sicurezza alimentare, biodiversità e mitigazione climatica, soprattutto dove la gestione del territorio è insostenibile ed il cambiamento climatico ne amplifica gli effetti. Conseguire la Land Degradation Neutrality richiede non solo obiettivi di recupero degli ecosistemi, ma anche sistemi di monitoraggio capaci di coglierne i feedback dinamici, multiscalari e multitemporali fra suolo, acqua e vegetazione. A questo obiettivo la ricerca risponde presentando un flusso metodologico che passa dalle valutazioni globali del degrado del suolo verso un sistema locale, sensibile al contesto, orientato al monitoraggio della rigenerazione nei paesaggi mediterranei. L’approccio poggia su due Baseline. Baseline 1: un Geospatial Data Management System (GDMS) rielaborato dal metodo MEDALUS, che archivia indicatori ed integra dati sul campo raccolti con Arena ancorati ad un sistema di strutture Survey basate su griglie da 30 × 30 m - 10 × 10 m ed Aree di Validazione. Baseline 2: un workflow Earth Observation (EO) legato al quadro LULUCF; script rGEE estraggono serie NDVI/EVI da Landsat 5/7/8 e Sentinel 2, rilevano cambi d’uso del suolo e costruiscono modelli fenologici a scala del pixel associato ad ogni elemento del Survey. L’integrazione dei due pilastri supera i limiti delle classi LULUCF semi-automatizzate, producendo una Local Land Use Classification (LoLUC) che fonde fenologia satellitare e misure in situ (i.e., sostanza organica, pH, infiltrazione, qualità della vegetazione e tasso di sopravvivenza). Il sistema LoLUC REarth intercetta più picchi vegetazionali stagionali nello stesso pixel, quantifica i benefici di rotazioni perenni, leguminose su sodo, e correla i trend dei pixel con variabili pedologiche e di management (i.e. e tecniche sperimentali di piantumazione agroforestale e sistemi nature-based di raccolta idrica). Questa visione dinamica sostituisce la classificazione rigida, evidenziando fasi precoci di crescita, traiettorie di resilienza e l’efficacia reale delle pratiche di agricoltura rigenerativa. Il flusso dati Glocal armonizzato e l’integrazione fotogrammetrica pongono la base per un futuro Digital Twin: una piattaforma in cui indicatori GDMS ed EO alimenteranno poligoni fenologici automatizzati, una pesatura unificata degli indicatori e un monitoraggio scalabile a reti regionali. Il workflow EO-GDMS di LoLUC REarth dimostra che l’integrazione dei dataset e favorisce il coinvolgimento di pedologi, agronomi, comunità locali e policy maker, garantendo rigore scientifico e rilevanza sociale, preparandosi alla sua potenziale scalabilità territoriale.
Shifting from regional land degradation assessments to glocal regenerative agriculture monitoring: an EO-GDMS workflow within the LULUCF framework
GABRIELE, MARZIA
2024/2025
Abstract
Land degradation and desertification threaten ecosystems, biodiversity, and human well-being, especially where anthropogenic pressures and climate change converge, often driven by unsustainable land use and environmental mismanagement. Given its cascading effects on soil fertility, water resources, and vegetation - essential ecosystem services for human livelihoods, including socioeconomic impact, addressing desertification is one of the main critical issues by reversing the Land Degradation phenomena through regeneration actions to help achieve Land Restoration and reach the Land Degradation Neutrality (LDN) Global objectives. Regenerative actions are placed within a complex context made by interconnected phenomenon requiring a dynamic, multiscale and multi-temporal approach to capture constantly evolving processes that derive from trials-and-errors and progressive ecosystem adaptation. To govern this complexity, the primary research objective of this thesis has been to develop a methodological workflow to shift from global Land Degradation Assessments toward G-local Land Restoration Monitoring, to support the effectiveness of landscape regeneration strategies at a local level, taking in account the specificity of interrelated variables, to improve MultiActor decision making and guide regional up-scalability. Land Use Change is becoming increasingly crucial because it contributes to decarbonization policies under the LUC/LULUCF (Land Use Change/Land Use, Land Use Change, and Forestry) framework, which uses Earth Observation data to monitor global trends such as deforestation rates, carbon stock changes, and greenhouse gas emissions. The thesis aims to complement LULUCF global monitoring by implementing a finer-scale monitoring approach - farm by farm, hectare by hectare - to support regenerative actions. The expected return is at a Global scale, as IPCC guidelines (2006, 2013, 2019; AR6 WGIII 2022) indicate measurable ecosystems benefits deriving from correctly implementing regenerative practices - such as minimal tillage or perennial crop succession - impacting soil quality, vegetation cover, and overall resilience. The successfulness depends on local decisions and actions within a global framework. The Thesis proposes a Shifting from Regional Land Degradation Assessments to Local Regenerative Agriculture Monitoring uniting top-down global objectives with bottom-up local data components (G-Local). To achieve this primary objective, the research is grounded in the Mediterranean Desertification and Land Use (MEDALUS) methodology and Land Use, Land-Use Change, and Forestry (LULUCF) framework and aims to develop a scientifically grounded methodological workflow to transition the global framework for assessing land degradation into a localized, context-sensitive system for monitoring land restoration activities. To achieve this, the study builds upon, and advances, two foundational technical baselines identified: Baseline 1 - the Geospatial Data Management System (GDMS), and Baseline 2 - Earth Observation (EO). The refinement of these baselines streams from local context relevance and analytical precision, identifying cross-relations, lacking points and possible enhancement. By uniting these baselines, the system overcomes the limitations of LULUCF automated classification - particularly in heterogeneous, mixed-use landscapes - by integrating a Local Land Use Classification (LoLUC) through Phenological (EO) and Local-Scale Analysis (GDMS). The overall data stream structures resulted in a Local Land Use Change Regenerating Earth “LoLUC-REarth” an integrated MultiTemporal and MultiDimensional System, to monitor complex micro-scale granular trends and regenerative context-based scenarios. Instead of relying on broad global classifications, the methodology introduces “real” polygons for micro-experimental plots, accurately delineated from on-site surveys collected over three years. By examining fractions of a hectare under differing management regimes - traditional, organic, and regenerative - while systematically integrating ground-truth data, the approach captures a highly granular picture of land-use changes and evolving ecological processes. LoLUC-REarth has been tested across different climatic, morphological, and geographical settings with varying degrees of desertification, ensuring its adaptability and robustness. The study’s first applied demonstration of the system’s potential adaptability takes place in Basilicata (Italy) (first case study), a region characterized by diverse topographical and hydrological conditions that compound risks associated with land degradation. The second case study application tested the methodology in Murcia Region, Spain. Specifically, in Altiplano Camp, part of Ecosystem Regeneration Community Network. The Camp is an experimental area where place-based regeneration actions (crop rotation, perennial rotation) allowed to monitor the results deriving from ecosystem restoration, meanwhile controlling them with on-site soil experts (Regeneration Academy). The third case study is in Nicosia District, Cyprus and represents a riverine infrastructure, the Klirou-Malounta-Akaki Dam. The Klirou case represents a typical case of river watershed renaturation, specifically of buffer zones along river streams, providing an opportunity to test, analyze and monitor the effects of dam construction on spontaneous ecosystems. Collectively, these case studies facilitated the implementation of each specific research enhancement and contributed to progressively structure the LoLUC-REarth workflow, thus enabling the test and refinement of possible interactions between global tools, local EO-based monitoring and Geospatial Data Management System (GDMS). The first research Enhancement (a) derived from the application in Basilicata (Italy) and restructured the Mediterranean desertification and land use system (MEDALUS) GDMS to facilitate transitioning from a desertification sensitivity analysis to a local-scale data-driven regeneration monitoring infrastructure. For this purpose, a geospatial system has been implemented: accordingly, the system’s indicators (Soil Quality, Vegetation Quality, Management Quality, and Climate Quality) were restructured to include micro-level data, and progressively evolved to incorporate remote sensing data, through cloud computing (Google Earth Engine). In this way, the new GDMS indicators shift the attention from land degradation, toward supporting the monitoring of local-scale LUCs deriving from regenerative practices, fostering multidisciplinary cooperation to monitor positive outcomes. To enable re-scaling to local, the research proposes a second Enhancement (b), based on Enhancement (a) – which provided the baseline GDMS infrastructure tailored to Mediterranean regions. The enhancement (b) envisions its direct implementation in operational tools and systems for monitoring Land Restoration: LoLUC-REarth. To do that, LoLUC-REarth evolves the GDMS toward an adaptive system capable of supporting Local-scale data integration and EO data streams, opening to integrate advancements from remote sensing technologies (i.e., Google Earth Engine - GEE), and local-scale survey data. To integrate remote sensing, and field data to track localized regeneration dynamics into the GDMS, the research defines the parameters for a detailed local survey methodology, the “LoLUC-REarth Survey Structure”. The LoLUC-REarth Survey Structure is based on Arena FAO Data collection Platforms principles and introduces a spatial grid system and survey validation areas, to support the integration of local Scale Indices deriving from Field Observations and Laboratory Analyses. R Studio supports Data Management and Analytics and moreover, allows to align EO metrics with on-field measurement. In this way the Survey Structure itself can support to integrate r and Google Earth Engine (rGEE). The applicative use case to structure the Survey and populate the GDMS was the living laboratory at the Altiplano Camp. In this setting, field validation data - collected by soil experts, agronomists, and local stakeholders - were combined with LULUCF EO-derived metrics to populate and refine the local-scale MEDALUS-based GDMS. This unified workflow enables point-level analyses at sub-hectare scales, an essential capability when small plot variations can vastly influence soil quality and vegetation outcomes. At the heart of Enhancement (b) lies the EO segment with phenological modeling of ecological succession to monitor phenological trends (greening, browning and no trend). Using rGEE (R-Google Earth Engine) scripts, Landsat 5, 7, 8 and Copernicus Sentinel-2 imagery are processed for phenological analyses. The phenological modelling involves large temporal resolution time-series analysis of key Vegetation Indictors such as the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) to detect start-of-season (SOS) and end-of-season (EOS) and identify transitions in vegetative cycles (ΔT) to guide on-site interventions. The system tracks multiple vegetative peaks within the same year, capturing complex coexisting plant communities. For instance, a single parcel might show overlapping signals from perennial crop succession, no-till legumes that enhance soil carbon retention, improved evaporative activities, and Mediterranean hedges designed to preserve soil moisture. By streamline workflow interconnections field validation data - combined with EO-derived metrics - are fed into the LoLUC-REarth GDMS and enable point-level sub-hectare analyses using precisely delineated micro-experimental plots. Although the thesis primarily focuses on agricultural settings, the application to contexts dominated by man-made interventions and large-scale infrastructures, such as the Klirou Dam large- highlights the potential framework’s adaptability in scenarios where spontaneous revegetation of external taxa occurs under changed hydrological patterns. Results proved that LoLUC-REarth can handle multiple forms of revegetation, suggesting the methodology’s future potential in managing watershed forestry planning and other non-cultivated domains. The research builds upon the insights and frameworks established in the initial phases and designs a system that can be scaled, adapted, and iteratively improved over time. As for that, The Research enhancement (c) proposes an Iterative refinement plan which identifies two main on-field data integration: on one side, photogrammetric surveys (UAV), which have been addressed to generate 3D landscape textured models to refine the delineation of experimental plots, detect subtle topographical features affecting runoff and infiltration; on the other side, Leaf Area Index ground measurements which are going to future iteratively refine the vegetation trends to discern thresholds of species composition. The results evidenced how LoLUC-REarth workflow allows to analyze the EO in relation to the local scale indicator, and vice versa, thus providing for each micro experimental plot the quantitative outcomes deriving from soil conservation, crop rotation efficacy, and vegetation recovery. The EO phenological model detected multiple vegetation trends and pixel-based ecological succession per year, reflecting mixed species growth peaks, deriving from implementing practices like perennial crops, no-till legumes, and soil moisture-preserving hedges. Moreover, by actively involving local stakeholders in data-gathering, the system captured knowledge about traditional, organic, and regenerative cultivation methods. This inclusive process enabled evidence-based interventions, guiding decisions about land management improvements while also deepening societal engagement. The envisioned LoLUC-REarth System creates a unified and adaptive framework for managing regenerative land restoration, aligning large-scale geospatial insights with site-specific, context-aware implementation. The system supports continuous, multi-scale tracking of land-use regeneration dynamics and restoration efforts, effectively bridging the gap between remote sensing analytics and field realities. The research stands ready for further refinements and applications, transforming traditional desertification risk assessments into an active, data-driven system, envisioning a digital twin infrastructure. The creation of a LoLUC-REarth GIS-based testing platform is a foundational result toward achieving the vision of a G-Local Digital Twin for rural landscape management and ecological restoration. As for that, the GIS-based testing platform provided access to evidence-based insights, empowering local stakeholders to act with evidence-based strategies. Future developments foresee the LoLUC-REarth parameterization to up-scale from individual plots to infra-regional territories (e.g., the AlVeLal community in Spain), thereby supporting adaptive, regenerative strategies and climate resilience. Its scalable and replicable approach will also contribute to developing shared guidelines for farmers and land managers to use in infra-regional regeneration experimental areas.| File | Dimensione | Formato | |
|---|---|---|---|
|
05_05_2025_GABRIELE_MARZIA_PhD_Thesis.pdf
non accessibile
Descrizione: PhD Thesis Complete Manuscript
Dimensione
427.76 MB
Formato
Adobe PDF
|
427.76 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/237837