Data Package Metadata   View Summary

Due to current climate change patterns, terrestrial ecosystems are being seriously affected with respect to their biodiversity, structure, and function. Specifically, the Mediterranean region is one of the areas most sensitive to climate change effects. Given the recent environmental policies derived from these serious threats caused by global climate change, practical measures to decrease net CO2 emissions must be established. Carbon sequestration is a major measure to reduce atmospheric CO2 concentrations within the short and medium term, in which terrestrial ecosystems play an essential role as carbon sinks. Quantification and monitoring of organic carbon reservoirs is needed to inform environmental management, adapt local policies and assess potential impacts.

The main aim of this thesis is to provide the methodological strategy for the quantification of organic carbon reservoirs in terrestrial ecosystems, based on standardized techniques at different spatial and management scales, in addition to establishing dynamic models to predict results under different management scenarios. This thesis is focused on two critical aspects of carbon stock modeling at the national level: (1) estimation of the carbon in aboveground biomass, and (2) quantification of the soil carbon storage, as well as its potential sequestration under different land use management scenarios; using spatially explicit models in both cases.

After the general introduction and methods (chapter 1 and 2), chapter 3 integrates two complementary remote sensing technologies to detail information about the spatial distribution of carbon stored in biomass. The multitemporal and global distribution of moderate resolution indexes (MODIS) are balanced with the temporal limitation of the high-precision airborne laser scanning (ALS) data. As a case study, this methodology was applied in a Mediterranean semiarid region in the southeastern Iberian Peninsula (specifically the region of Murcia). The results shows a robust performance in modeling of ALS data calibrated with plot-level ground-based measures, and bio-geophysical spectral variables (8 different indexes derived from MODIS), confirming its applicability at coarser resolutions.

Chapter 4 improves different digital soil mapping (DSM) techniques to develop a local soil organic carbon (SOC) map and test it against estimates derived from available regional-to-global carbon products. The aim of this chapter is to define a high-resolution SOC map framework, analyzing diverse aspects. These aspects are refered to different carbon variables (SOC concentration -g/kg-, and SOC stock -tC/ha-) using different spatial interpolation methods (linear model, quantile regression forest –QRF-; random forest and support vector machine) at three spatial resolutions (100m, 250m, 1000m). Considering again the ‘Region de Murcia’ as a case study, the results show an optimal framework based on local soil data, environmental covariates (including single and/or multitemporal remote sensing indices), DMS modeling and spatially explicit uncertainty quantification. Specifically, the QRF approach parameterized with SOC concentration data at 100 m spatial resolution confirms the best data-model agreement and the best balance for accuracy, external validation, and interpretability of results. This study provides a better understanding of SOC storage across complex soil-forming environments with limited soil samples.

In view of the results of chapter 4, chapter 5 maps the SOC spatial distribution at the national level at 90 m and the associated spatially explicit uncertainties. Modeling of the legacy soil database (8,361 profile samples) and a selection of environmental data-driven covariates was based on three supervised learning approaches: quantile regression forest, ensemble machine learning and auto-machine learning. The maps are estimated at 0-30 cm, 30-100cm and the effective soil depth. For the final SOC spatial distribution maps, a novel methodology is used. It is based on a combination of different predictive ensemble models where each pixel is assigned the prediction from the most accurate model, i.e. lowest uncertainty. The mean value of the SOC concentration map is 15.7 g/kg, storing approximately 25% in subsoils (>30cm). The total SOC stock at its effective depth is 3.8 Pg C, of which 2.82 Pg C are stored in the upper 30 cm (74% of the total).

Chapter 6 predicts SOC potential sequestration map to detect land uses, sites and regions with greater potential to absorb SOC stocks for peninsular Spain under different management scenarios. In this study, the RothC model at the national level (1 km grids) is used for the projection of the 2020-2040 period. The results shows that the SOC sequestration in peninsular Spain, supposing the current environmental conditions remain constant for the next 20 years, will decrease at a rate of 430 Gg C/yr approximately. However, a potential sequestration of 1,980 Gg C/yr can be expected approximately under the adoption of sustainable management practices that increase the carbon input rate into soils by at least 5% in the next 20 years.

In summary, the advances provided by this thesis contribute to improving the quality and precision of current methodologies available at the national level for the carbon in terrestrial ecosystem, with novel information on its carbon storage, and serve as a reference for climate change mitigation strategies and policies.

The data does not require regular maintenance.