Documents

Wildfires are a recurrent and increasing threat in mainland Portugal, where over 4.5 million hectares of forests and scrublands have burned over the last 38 years. These fire-affected landscapes have suffered an intensification of soil erosion processes, which can negatively affect soil carbon storage, reduce fertility and forest productivity, and can become a source of pollutants. The main objective of the present study is to produce a post-fire soil erosion risk map for the forest and shrubland areas in mainland Portugal and assess its reliability. To this end, the semi-empirical Morgan–Morgan–Finney erosion model was used to assess the potential post-fire soil erosion according to distinct burn severity and climate scenarios, and the accuracy of the predictions was verified by an uncertainty analysis and validated against independent field datasets. The proposed approach successfully allowed mapping post-fire soil erosion in Portugal and identified the areas with higher post-fire erosion risk for past and future climate extremes. The outcomes of this study comprise a set of tools to help forest managers in their decision-making for post-fire emergency stabilization, ensuring the adequate selection of areas for mitigation to minimize the economic and environmental losses caused by fire-enhanced soil erosion.

10.1038/s41598-022-07066-x

The forecasting of crop yield is one of the most critical research areas in crop science, which allows for the development of decision support systems, optimization of nitrogen fertilization, and food safety. Many tested modeling approaches can be differentiated according to the models and data used. The models used are traditional crop models that require data that are often difficult to measure. New modeling approaches based on artificial intelligence algorithms have proven to be of high performance, flexible, and can be tested based on available data. In this study, four independent field experiments conducted on Triticum turgidum subsp. durum Desf. in central–southern Italy were used to train a set of machine learning (ML) algorithms to predict the yield using 16 variables: fertilization, nitrogen management, pedoclimatic, and remote sensing data. Four ML algorithms were calibrated and validated over two independent sites, and a linear regression model was used as a control. The calibrated models can predict the grain yield in the two regions by using ancillary data, topsoil physical and chemical properties, multispectral drone imagery, climatic data, and nitrogen fertilizer applied at the site. Among the four ML algorithms, stochastic gradient boosting (root-mean-square error  = 0.58 t ha−1) outperformed others during calibration and transferability. Nitrogen application rate, seasonal precipitation, and temperature are the most important features for predicting wheat yield.

10.1002/agj2.21279

The increase in soil salinity due to human-induced processes poses a severe threat to agriculture on a regional and global scale. Soil salinization caused by natural and anthropogenic factors is a vital environmental hazard, specifically in semi-arid and arid regions of the world. The detection and monitoring of salinity are critical to the sustainability of soil management. The current study compared the performance of machine learning models to produce spatial maps of electrical conductivity (EC) (as a proxy for salinity) in an alluvial irrigation plain. The current study area is located in the Isparta province (100 km2), land cover is mainly irrigated, and the dominant soils are Inceptisols, Mollisols, and Vertisols. Digital soil mapping (DSM) methodology was used, referring to the increase in the digital representation of soil formation factors with today’s technological advances. Plant and soil-based indices produced from the Sentinel 2A satellite image, topographic indices derived from the digital elevation model (DEM), and CORINE land cover classes were used as predictors. The support vector regression (SVR) algorithm revealed the best relationships in the study area. Considering the estimates of different algorithms, according to the FAO salinity classification, a minimum of 12.36% and a maximum of 20.19% of the study area can be classified as slightly saline. The low spatial dependence between model residuals limited the success of hybrid methods. The land irrigated cover played a significant role in predicting the current level of EC.

DOI: 10.3390/land11122148

We present the version 1.2 of the recently published [Nature communications 8, 2013 (2017)] RUSLE-based Global Soil Erosion Modelling platform (GloSEM). Unlike version 1.1, effects of permanent crops, managed pasture and temporary disturbed forest loss are spatially defined based on a probabilistic land by using allocation approach and their implications for soil erosion are assessed in the advanced version 1.2. For 2012, we estimated an annual total soil erosion of 38.9 Pg/a. This constitutes an increase of ca. 8% compared to the previous version (35.9 Pg/a) which is due to an increase of soil erosion mainly related to the new areas classified as managed pasture and to a lesser extent to permanent crop and forest disturbances.

DOI: 10.1007/978-981-16-7916-2_1

The European Commission's Thematic Strategy for Soil Protection (COM(2012)46) identified soil erosion as an important threat to European Union's (EU) soil resources. Gully erosion is an important but hitherto poorly understood component of this threat. Here we present the results of an unprecedented attempt to monitor the occurrence of gully erosion across the EU and UK. We integrate a soil erosion module into the 2018 LUCAS Topsoil Survey, which was conducted to monitor the soil health status across the EU and to support actions to prevent soil degradation. We discuss and explore opportunities to further improve this method. The 2018 LUCAS Topsoil Survey consisted of soil sampling (0–20 cm depth) and erosion observations conducted in ca. 10% (n = 24,759) of the 238,077 Land Use/Cover Area frame Survey (LUCAS) 2018 in-field survey sites. Gully erosion channels were detected for ca. 1% (211 sites) of the visited LUCAS Topsoil sites. Commission (false positives, 2.5%) and omission errors (false negatives, 5.6%) were found to be low and at a level that could not compromise the representativeness of the gully erosion survey. Overall, the findings indicate that the tested 2018 LUCAS Topsoil in-field gully erosion monitoring system is effective for detecting the incidence of gully erosion. The morphogenesis of the mapped gullies suggests that the approach is an effective tool to map permanent gullies, whereas it appears less effective to detect short-lived forms like ephemeral gullies. Spatial patterns emerging from the LUCAS Topsoil field observations provide new insights on typical gully formation sites across the EU and UK. This can help to design further targeted research activities. An extension of this approach to all LUCAS sites of 2022 would significantly enhance our understanding of the geographical distribution of gully erosion processes across the EU. Repeated every three years, LUCAS soil erosion surveys would contribute to assess the state of gully erosion in the EU over time. It will also enable monitoring and eventually predicting the dynamics of gully erosion. Data collected were part of the publicly available Gully Erosion LUCAS visual assessment (GE-LUCAS v1.0) inventory.

10.1016/j.iswcr.2021.09.002

10.1016/j.catena.2021.105957

Promoting sustainable soil management is a possible option for achieving net-zero greenhouse gas emissions in the future. Several efforts in this area exist, and the application of spatially explicit models to anticipate the effect of possible actions on soils at a regional scale is widespread. Currently, models can simulate the impacts of changes on land cover, land management, and the climate on the soil carbon stocks. However, existing modeling tools do not incorporate the lateral transport and deposition of soil material, carbon, and nutrients caused by soil erosion. The absence of these fluxes may lead to an oversimplified representation of the processes, which hinders, for example, a further understanding of how erosion has been affecting the soil carbon pools and nutrients through time. The sediment transport during deposition and the sediment loss to rivers create dependence among the simulation units, forming a cumulative effect through the territory. If, on the one hand, such a characteristic implies that calculations must be made for large geographic areas corresponding to hydrological units, on the other hand, it also can make models computationally expensive, given that erosion and redeposition processes must be modeled at high resolution and over long timescales. In this sense, the present work has a three-fold objective. First, we provide the development details to represent in matrix form a spatially explicit process-based model coupling sediment, carbon, and erosion, transport, and deposition (ETD) processes of soil material in hillslopes and valley bottoms (i.e., the CE-DYNAM model). Second, we illustrate how the model can be calibrated and validated for Europe, where high-resolution datasets of the factors affecting erosion are available. Third, we presented the results for a depositional site, which is highly affected by incoming lateral fluxes from upstream lands. Our results showed that the benefits brought by the matrix approach to CE-DYNAM enabled the before-precluded possibility of applying it on a continental scale. The calibration and validation procedures indicated (i) a close match between the erosion rates calculated and previous works in the literature at local and national scales, (ii) the physical consistency of the parameters obtained from the data, and (iii) the capacity of the model in predicting sediment discharge to rivers in locations observed and unobserved during its calibration (model efficiency (ME) =0.603, R2=0.666; and ME =0.152, R2=0.438, respectively). The prediction of the carbon dynamics on a depositional site illustrated the model's ability to simulate the nonlinear impact of ETD fluxes on the carbon cycle. We expect that our work advances ETD models' description and facilitates their reproduction and incorporation in land surface models such as ORCHIDEE. We also hope that the patterns obtained in this work can guide future ETD models at a European scale.

10.5194/gmd-15-7835-2022

Soil erosion is time compressed into a number of episodic erosive rainfall events with an associated potential to detach and transport soil particles (rainfall erosivity), each possessing unique spatial and temporal characteristics. Rainfall erosivity events in Europe follow extreme value distributions in which a limited number of rainstorms dominate the long-term budget of available erosive energy. To combat soil erosion in Europe in a targeted manor, timely erosion mitigation measures should derive from dynamic model simulations that incorporate spatially and temporally distributed estimations of rainfall erosivity. Rain gauge measurements from singular points are typically used to quantify rainfall erosivity, however the spatial uniqueness of rainfall presents a key limitation to dynamically model rainfall across broad spatial scales with a limited number of point measurements. Discretised gridded precipitation datasets with a widespread (e.g. continental) spatial coverage potentially offer an opportunity to adequately replicate the dynamics of rainfall erosivity events, however their performance remains poorly tested in the pan-European context.

This study builds upon the comprehensive Rainfall Erosivity Database at European Scale (REDES) archive of over 300,000 events from 1181 gauge stations to develop a two-step modelling process: 1) firstly, optimal monthly models were fitted and evaluated between gauge-recorded rainfall depth and rainfall erosivity (EI30) across European climatic regions to develop a European-scale parameter surface, 2) secondly, three datasets (EMO-5 (6-hr), E-OBS (24-hr), UERRA MESCAN-SURFEX (24hr)) were directly evaluated via a grid-to-point analysis based on their ability to simulate the station-specific event rainfall erosivity timeseries at a random selection of 32 locations. EMO-5 (Nash-Sutcliffe model efficiency mean = 0.24) outperformed other tested gridded datasets, showing the capability to adequately replicate the event number, timing, and their average magnitude. A higher model performance in Northern compared with Southern European climatic regions, in which characteristically higher and spatially-complex event rainfall erosivity magnitudes are found, was symptomatic of a poor ability of grid-based simulations to replicate the magnitudes of events in the outer extents of the frequency-magnitude spectrum. The absence of a clear global systematic predictive bias amongst simulated locations suggests the need for future upscaling of this analysis to the entire European REDES dataset to fully understand and correct for the method-derived bias in a climate region-specific way.

10.1016/j.catena.2022.106157

The cover management factor (C-factor) calculation requires the assessment of the intra‐annual spatiotemporal variability of biomass cover, owed to the natural growth cycle of vegetation and the impact of agriculture on land cover. However, this is frequently omitted, and the vegetation conditions are approximated by assigning constant values to static classified Land Use/Land Cover (LULC) maps, such as the Coordination of Information on the Environment (CORINE) Land Cover (CLC). Using as test site the Sperchios River catchment, Central Greece, this study introduces a new approach to estimate C-factor in a spatiotemporally exhaustive manner. The goal is to increase estimation accuracy in erosion modelling applications. The C-factor computations are performed on monthly scale, based on LULC maps that portray the basin’s agricultural areas in unprecedented detail. The methodology involves the use of a biophysical index, namely Fraction of Vegetation Cover (Fcover) and empirical literature data on crop types. Fcover was developed from Sentinel-2 (S2) imagery in 10-m analysis. Such analysis (compared to the 300-m one provided by the EU) is a major improvement towards a more precise estimation of C-factor. The study identified the monthly C-factor fluctuation at basin scale, and the most susceptible months seasons at localities in terms of land cover/soil loss potential. The higher C-factor values were acquired in October and the lower in May. Mean annual (numerical) C-factor complies with the value of July. All monthly values are significantly higher – almost double – than the mean annual stationary one. The revealed patterns would not have been detected in a lower temporal (e.g., annual) resolution without the incorporation of vegetation density seasonality. The study shows high reproducibility and upscaling potential, as the utilized datasets are available in all European Union (EU) Member States, having similar structure, thus they can be harmonized towards a unified continental approach.

10.1016/j.catena.2022.106149

Since soil biodiversity sustains above-ground life, the European Union (EU) has recently announced its new Soil Strategy to better protect soil ecosystems as part of the Biodiversity Strategy for 2030. Also, the EU's Farm to Fork Strategy and the Zero Pollution Action Plan aim for soil protection. However, the status of soil biodiversity protection has not been comprehensively assessed. Therefore, we explored regulatory, incentive-based and knowledge-based instruments and strategic policy documents at the EU and national levels to determine whether they adequately protect soil biodiversity. Our review of 507 literature references concluded that only eight EU member states explicitly address threats to soil biodiversity in 14 regulatory instruments while 13 countries mainly focus on implicit threats to soil biodiversity, whereas six countries do not consider soil biodiversity. At the EU level, current directives and regulations only tackle individual threats to soil biodiversity. An EU-wide, legally binding protection could ensure a standardised minimum level of soil biodiversity protection while preventing surging costs of not acting. The EU Soil Health Law foreseen for 2023 could couple land management practices beneficial for soil biodiversity with incentive-based instruments. Simultaneously, models should be designed to predict soil biodiversity, considering soil biodiversity's spatial and temporal heterogeneity.

10.1016/j.biocon.2022.109475

This study addressed a critical resource—soil—through the prism of processing big data at the continental scale. Rapid progress in technology and remote sensing has majorly improved data processing on extensive spatial and temporal scales. Here, the manuscript presents the results of a systematic effort to geo-process and analyze soil-relevant data. In addition, the main highlights include the difficulties associated with using data infrastructures, managing big geospatial data, decentralizing operations through remote access, mass processing, and automating the data-processing workflow using advanced programming languages. Challenges to this study included the reproducibility of the results, their presentation in a communicative way, and the harmonization of complex heterogeneous data in space and time based on high standards of accuracy. Accuracy was especially important as the results needed to be identical at all spatial scales (from point counts to aggregated countrywide data). The geospatial modeling of soil requires analysis at multiple spatial scales, from the pixel level, through multiple territorial units (national or regional), and river catchments, to the global scale. Advanced mapping methods (e.g., zonal statistics, map algebra, choropleth maps, and proportional symbols) were used to convey comprehensive and substantial information that would be of use to policymakers. More specifically, a variety of cartographic practices were employed, including vector and raster visualization and hexagon grid maps at the global or European scale and in several cartographic projections. The information was rendered in both grid format and as aggregated statistics per polygon (zonal statistics), combined with diagrams and an advanced graphical interface. The uncertainty was estimated and the results were validated in order to present the outputs in the most robust way. The study was also interdisciplinary in nature, requiring large-scale datasets to be integrated from different scientific domains, such as soil science, geography, hydrology, chemistry, climate change, and agriculture.

10.3390/land11122287