How can we quantify, explain, and apply the uncertainty of complex soil maps predicted with neural networks?

Artificial neural networks (ANNs) have proven to be a useful tool for complex questions that involve large amounts of data. Our use case of predicting soil maps with ANNs is in high demand by government agencies, construction companies, or farmers, given cost and time intensive field work. However, there are two main challenges when applying ANNs. In their most common form, deep learning algorithms do not provide interpretable predictive uncertainty. This means that properties of an ANN such as the certainty and plausibility of the predicted variables, rely on the interpretation by experts rather than being quantified by evaluation metrics validating the ANNs. Further, these algorithms have shown a high confidence in their predictions in areas geographically distant from the training area or areas sparsely covered by training data. To tackle these challenges, we use the Bayesian deep learning approach "last-layer Laplace approximation", which is specifically designed to quantify uncertainty into deep networks, in our explorative study on soil classification. It corrects the overconfident areas without reducing the accuracy of the predictions, giving us a more realistic uncertainty expression of the model's prediction. In our study area in southern Germany, we subdivide the soils into soil regions and as a test case we explicitly exclude two soil regions in the training area but include these regions in the prediction. Our results emphasize the need for uncertainty measurement to obtain more reliable and interpretable results of ANNs, especially for regions far away from the training area. Moreover, the knowledge gained from this research addresses the problem of overconfidence of ANNs and provides valuable information on the predictability of soil types and the identification of knowledge gaps. By analyzing regions where the model has limited data support and, consequently, high uncertainty, stakeholders can recognize the areas that require more data collection efforts.

Microbial impact on initial soil formation in arid and semiarid environments under simulated climate change.

The microbiota is attributed to be important for initial soil formation under extreme climate conditions, but experimental evidence for its relevance is scarce. To fill this gap, we investigated the impact of in situ microbial communities and their interrelationship with biocrust and plants compared to abiotic controls on soil formation in initial arid and semiarid soils. Additionally, we assessed the response of bacterial communities to climate change. Topsoil and subsoil samples from arid and semiarid sites in the Chilean Coastal Cordillera were incubated for 16 weeks under diurnal temperature and moisture variations to simulate humid climate conditions as part of a climate change scenario. Our findings indicate that microorganism-plant interaction intensified aggregate formation and stabilized soil structure, facilitating initial soil formation. Interestingly, microorganisms alone or in conjunction with biocrust showed no discernible patterns compared to abiotic controls, potentially due to water-masking effects. Arid soils displayed reduced bacterial diversity and developed a new community structure dominated by Proteobacteria, Actinobacteriota, and Planctomycetota, while semiarid soils maintained a consistently dominant community of Acidobacteriota and Proteobacteria. This highlighted a sensitive and specialized bacterial community in arid soils, while semiarid soils exhibited a more complex and stable community. We conclude that microorganism-plant interaction has measurable impacts on initial soil formation in arid and semiarid regions on short time scales under climate change. Additionally, we propose that soil and climate legacies are decisive for the present soil microbial community structure and interactions, future soil development, and microbial responses.

Spectral prediction of soil salinity and alkalinity indicators using visible, near-, and mid-infrared spectroscopy.

Drought and the impacts of climate change have led to an escalation in soil salinity and alkalinity across various regions worldwide, including Iran. The Chahardowli Plain in western Iran, in particular, has witnessed a significant intensification of this phenomenon over the past decade. Consequently, modeling of soil attributes that serve as indicators of soil salinity and alkalinity became a priority in this region. To date, only a limited number of studies have been conducted to assess indicators of salinity and alkalinity through spectrometry across diverse spectral ranges. The spectral ranges encompassing mid-infrared (mid-IR), visible, and near-infrared (vis-NIR) spectroscopy were employed to estimate soil properties including sodium adsorption ratio (SAR), exchangeable sodium ratio (ESR), exchangeable sodium percentage (ESP), pH, and electrical conductivity (EC). Five distinct models were employed: Partial Least Squares Regression (PLSR), bootstrapping aggregation PLSR (BgPLSR), Memory-Based Learning (MBL), Random Forest (RF), and Cubist. The calibration and assessment of model performance were carried out using several key metrics including Ratio of Performance to Deviation (RPD) and the coefficient of determination (R2). Analysis of the outcomes indicates that the accuracy and precision of the mid-IR spectra surpassed that of vis-NIR spectra, except for pH, which exhibited a superior RPD compared to other properties. Notably, in the prediction of pH utilizing vis-NIR reflectance spectra, the BgPLSR model exhibited the highest accuracy and precision, boasting an RPD value of 2.56. In the domain of EC prediction, the PLSR model yielded an RPD of 2.64. For SAR, the MBL model achieved an RPD of 2.70, while ESR prediction benefited from the MBL model with an impressive RPD of 4.36. Likewise, the MBL model demonstrated remarkable precision and accuracy in ESP prediction, garnering an RPD of 4.41. The MBL model's efficacy in forecasting with limited datasets was notably pronounced among the models considered. This study underscores the valuable role of spectral predictions in facilitating the work of soil surveyors in gauging salinity and alkalinity indicators. It is recommended that the integration of spectrometry-based salinity and alkalinity predictions be incorporated into forthcoming soil mapping endeavors within semi-arid and arid regions.

Carbon-biodiversity relationships in a highly diverse subtropical forest.

Carbon-focused climate mitigation strategies are becoming increasingly important in forests. However, with ongoing biodiversity declines we require better knowledge of how much such strategies account for biodiversity. We particularly lack information across multiple trophic levels and on established forests, where the interplay between carbon stocks, stand age, and tree diversity might influence carbon-biodiversity relationships. Using a large dataset (>4600 heterotrophic species of 23 taxonomic groups) from secondary, subtropical forests, we tested how multitrophic diversity and diversity within trophic groups relate to aboveground, belowground, and total carbon stocks at different levels of tree species richness and stand age. Our study revealed that aboveground carbon, the key component of climate-based management, was largely unrelated to multitrophic diversity. By contrast, total carbon stocks-that is, including belowground carbon-emerged as a significant predictor of multitrophic diversity. Relationships were nonlinear and strongest for lower trophic levels, but nonsignificant for higher trophic level diversity. Tree species richness and stand age moderated these relationships, suggesting long-term regeneration of forests may be particularly effective in reconciling carbon and biodiversity targets. Our findings highlight that biodiversity benefits of climate-oriented management need to be evaluated carefully, and only maximizing aboveground carbon may fail to account for biodiversity conservation requirements.

Contextual spatial modelling in the horizontal and vertical domains.

Multi-scale contextual modelling is an important toolset for environmental mapping. It accounts for spatial dependence by using covariates on multiple spatial scales and incorporates spatial context and structural dependence to environmental properties into machine learning models. For spatial soil modelling, three relevant scales or ranges of scale exist: quasi-local soil formation processes that are independent of the spatial context, short-range catenary processes, and long-range processes related to climate and large-scale terrain settings. Recent studies investigated the spatial dependence of topsoil properties only. We hypothesize that soil properties within a soil profile were formed due to specific interactions between different features and scales of the spatial context, and that there are depth gradients in spatial and structural dependencies. The results showed that for topsoil, features at small to intermediate scales do not increase model accuracy, whereas large scales increase model accuracy. In contrast, subsoil models benefit from all scales-small, intermediate, and large. Based on the differences in relevance, we conclude that the relevant ranges of scales do not only differ in the horizontal domain, but also in the vertical domain across the soil profile. This clearly demonstrates the impact of contextual spatial modelling on 3D soil mapping.

Impact of Climate and Slope Aspects on the Composition of Soil Bacterial Communities Involved in Pedogenetic Processes along the Chilean Coastal Cordillera.

Soil bacteria play a fundamental role in pedogenesis. However, knowledge about both the impact of climate and slope aspects on microbial communities and the consequences of these items in pedogenesis is lacking. Therefore, soil-bacterial communities from four sites and two different aspects along the climate gradient of the Chilean Coastal Cordillera were investigated. Using a combination of microbiological and physicochemical methods, soils that developed in arid, semi-arid, mediterranean, and humid climates were analyzed. Proteobacteria, Acidobacteria, Chloroflexi, Verrucomicrobia, and Planctomycetes were found to increase in abundance from arid to humid climates, while Actinobacteria and Gemmatimonadetes decreased along the transect. Bacterial-community structure varied with climate and aspect and was influenced by pH, bulk density, plant-available phosphorus, clay, and total organic-matter content. Higher bacterial specialization was found in arid and humid climates and on the south-facing slope and was likely promoted by stable microclimatic conditions. The presence of specialists was associated with ecosystem-functional traits, which shifted from pioneers that accumulated organic matter in arid climates to organic decomposers in humid climates. These findings provide new perspectives on how climate and slope aspects influence the composition and functional capabilities of bacteria, with most of these capabilities being involved in pedogenetic processes.

Organic and conservation agriculture promote ecosystem multifunctionality.

Ecosystems provide multiple services to humans. However, agricultural systems are usually evaluated on their productivity and economic performance, and a systematic and quantitative assessment of the multifunctionality of agroecosystems including environmental services is missing. Using a long-term farming system experiment, we evaluated and compared the agronomic, economic, and ecological performance of the most widespread arable cropping systems in Europe: organic, conservation, and conventional agriculture. We analyzed 43 agroecosystem properties and determined overall agroecosystem multifunctionality. We show that organic and conservation agriculture promoted ecosystem multifunctionality, especially by enhancing regulating and supporting services, including biodiversity preservation, soil and water quality, and climate mitigation. In contrast, conventional cropping showed reduced multifunctionality but delivered highest yield. Organic production resulted in higher economic performance, thanks to higher product prices and additional support payments. Our results demonstrate that different cropping systems provide opposing services, enforcing the productivity-environmental protection dilemma for agroecosystem functioning.

Iron mineral dissolution releases iron and associated organic carbon during permafrost thaw.

It has been shown that reactive soil minerals, specifically iron(III) (oxyhydr)oxides, can trap organic carbon in soils overlying intact permafrost, and may limit carbon mobilization and degradation as it is observed in other environments. However, the use of iron(III)-bearing minerals as terminal electron acceptors in permafrost environments, and thus their stability and capacity to prevent carbon mobilization during permafrost thaw, is poorly understood. We have followed the dynamic interactions between iron and carbon using a space-for-time approach across a thaw gradient in Abisko (Sweden), where wetlands are expanding rapidly due to permafrost thaw. We show through bulk (selective extractions, EXAFS) and nanoscale analysis (correlative SEM and nanoSIMS) that organic carbon is bound to reactive Fe primarily in the transition between organic and mineral horizons in palsa underlain by intact permafrost (41.8 ± 10.8 mg carbon per g soil, 9.9 to 14.8% of total soil organic carbon). During permafrost thaw, water-logging and O2 limitation lead to reducing conditions and an increase in abundance of Fe(III)-reducing bacteria which favor mineral dissolution and drive mobilization of both iron and carbon along the thaw gradient. By providing a terminal electron acceptor, this rusty carbon sink is effectively destroyed along the thaw gradient and cannot prevent carbon release with thaw.

A tale of scale: Plot but not neighbourhood tree diversity increases leaf litter ant diversity.

Diversity of producers (e.g. plants) usually increases the diversity of associated organisms, but the scale (i.e. the spatial area of plant diversity considered) at which plant diversity acts on other taxa has rarely been studied. Most evidence for cross-taxon diversity relations come from above-ground consumers that directly interact with plants. Experimental tests of plant diversity effects on elusive organisms inhabiting the leaf litter layer, which are important for nutrient cycling and decomposition, are rare. Using a large tree diversity experiment, we tested whether tree diversity at the larger plot (i.e. community) or the smaller neighbourhood scale relates to the abundance, species richness, functional and phylogenetic diversity of leaf litter ants, which are dominant organisms in brown food webs. Contrary to our expectations of scale-independent positive tree diversity effects, ant diversity increased only with plot but not neighbourhood tree diversity. While the exact causal mechanisms are unclear, nest relocation or small-scale competition among ants may explain the stronger tree diversity effects at the plot scale. Our results indicate that even for small and less mobile organisms in the leaf litter, effects of tree diversity are stronger at relatively larger scales. The finding emphasizes the importance of diverse forest stands, in which mixing of tree species is not restricted to small patches, for supporting arthropod diversity in the leaf litter.

The relevant range of scales for multi-scale contextual spatial modelling.

Spatial autocorrelation in the residuals of spatial environmental models can be due to missing covariate information. In many cases, this spatial autocorrelation can be accounted for by using covariates from multiple scales. Here, we propose a data-driven, objective and systematic method for deriving the relevant range of scales, with distinct upper and lower scale limits, for spatial modelling with machine learning and evaluated its effect on modelling accuracy. We also tested an approach that uses the variogram to see whether such an effective scale space can be approximated a priori and at smaller computational cost. Results showed that modelling with an effective scale space can improve spatial modelling with machine learning and that there is a strong correlation between properties of the variogram and the relevant range of scales. Hence, the variogram of a soil property can be used for a priori approximations of the effective scale space for contextual spatial modelling and is therefore an important analytical tool not only in geostatistics, but also for analyzing structural dependencies in contextual spatial modelling.

Comparison of catchment scale 3D and 2.5D modelling of soil organic carbon stocks in Jiangxi Province, PR China.

As limited resources, soils are the largest terrestrial sinks of organic carbon. In this respect, 3D modelling of soil organic carbon (SOC) offers substantial improvements in the understanding and assessment of the spatial distribution of SOC stocks. Previous three-dimensional SOC modelling approaches usually averaged each depth increment for multi-layer two-dimensional predictions. Therefore, these models are limited in their vertical resolution and thus in the interpretability of the soil as a volume as well as in the accuracy of the SOC stock predictions. So far, only few approaches used spatially modelled depth functions for SOC predictions. This study implemented and evaluated an approach that compared polynomial, logarithmic and exponential depth functions using non-linear machine learning techniques, i.e. multivariate adaptive regression splines, random forests and support vector machines to quantify SOC stocks spatially and depth-related in the context of biodiversity and ecosystem functioning research. The legacy datasets used for modelling include profile data for SOC and bulk density (BD), sampled at five depth increments (0-5, 5-10, 10-20, 20-30, 30-50 cm). The samples were taken in an experimental forest in the Chinese subtropics as part of the biodiversity and ecosystem functioning (BEF) China experiment. Here we compared the depth functions by means of the results of the different machine learning approaches obtained based on multi-layer 2D models as well as 3D models. The main findings were (i) that 3rd degree polynomials provided the best results for SOC and BD (R2 = 0.99 and R2 = 0.98; RMSE = 0.36% and 0.07 g cm-3). However, they did not adequately describe the general asymptotic trend of SOC and BD. In this respect the exponential (SOC: R2 = 0.94; RMSE = 0.56%) and logarithmic (BD: R2 = 84; RMSE = 0.21 g cm-3) functions provided more reliable estimates. (ii) random forests with the exponential function for SOC correlated better with the corresponding 2.5D predictions (R2: 0.96 to 0.75), compared to the 3rd degree polynomials (R2: 0.89 to 0.15) which support vector machines fitted best. We recommend not to use polynomial functions with sparsely sampled profiles, as they have many turning points and tend to overfit the data on a given profile. This may limit the spatial prediction capacities. Instead, less adaptive functions with a higher degree of generalisation such as exponential and logarithmic functions should be used to spatially map sparse vertical soil profile datasets. We conclude that spatial prediction of SOC using exponential depth functions, in conjunction with random forests is well suited for 3D SOC stock modelling, and provides much finer vertical resolutions compared to 2.5D approaches.

The strength of soil-plant interactions under forest is related to a Critical Soil Depth.

Soil properties and terrain attributes are of great interest to explain and model plant productivity and community assembly (hereafter P&CA). Many studies only sample surface soils, and may therefore miss important variation of deeper soil levels. We aimed to identify a critical soil depth in which the relationships between soil properties and P&CA were strongest due to an ideal interplay among soil properties and terrain attributes. On 27 plots in a subtropical Chinese forest varying in tree and herb layer species richness and tree productivity, 29 soil properties in six depth columns and four terrain attributes were analyzed. Soil properties varied with soil depth as did their interrelationships. Non-linearity of soil properties led to critical soil depths in which different P&CA characteristics were explained best (using coefficients of determination). The strongest relationship of soil properties and terrain attributes to most of P&CA characteristics (adj. R2 ~ 0.7) was encountered using a soil column of 0-16 cm. Thus, depending on the biological signal one is interested in, soil depth sampling has to be adapted. Considering P&CA in subtropical broad-leaved secondary forests, we recommend sampling one bulk sample of a column from 0 cm down to a critical soil depth of 16 cm.

Impacts of species richness on productivity in a large-scale subtropical forest experiment.

Biodiversity experiments have shown that species loss reduces ecosystem functioning in grassland. To test whether this result can be extrapolated to forests, the main contributors to terrestrial primary productivity, requires large-scale experiments. We manipulated tree species richness by planting more than 150,000 trees in plots with 1 to 16 species. Simulating multiple extinction scenarios, we found that richness strongly increased stand-level productivity. After 8 years, 16-species mixtures had accumulated over twice the amount of carbon found in average monocultures and similar amounts as those of two commercial monocultures. Species richness effects were strongly associated with functional and phylogenetic diversity. A shrub addition treatment reduced tree productivity, but this reduction was smaller at high shrub species richness. Our results encourage multispecies afforestation strategies to restore biodiversity and mitigate climate change.

Experimental Evidence of Functional Group-Dependent Effects of Tree Diversity on Soil Fungi in Subtropical Forests.

Deconvoluting the relative contributions made by specific biotic and abiotic drivers to soil fungal community compositions facilitates predictions about the functional responses of ecosystems to environmental changes, such as losses of plant diversity, but it is hindered by the complex interactions involved. Experimental assembly of tree species allows separation of the respective effects of plant community composition (biotic components) and soil properties (abiotic components), enabling much greater statistical power than can be achieved in observational studies. We therefore analyzed these contributions by assessing, via pyrotag sequencing of the internal transcribed spacer (ITS2) rDNA region, fungal communities in young subtropical forest plots included in a large experiment on the effects of tree species richness. Spatial variables and soil properties were the main drivers of soil fungal alpha and beta-diversity, implying strong early-stage environmental filtering and dispersal limitation. Tree related variables, such as tree community composition, significantly affected arbuscular mycorrhizal and pathogen fungal community structure, while differences in tree host species and host abundance affected ectomycorrhizal fungal community composition. At this early stage of the experiment, only a limited amount of carbon inputs (rhizodeposits and leaf litter) was being provided to the ecosystem due to the size of the tree saplings, and persisting legacy effects were observed. We thus expect to find increasing tree related effects on fungal community composition as forest development proceeds.

Tree species richness increases ecosystem carbon storage in subtropical forests.

Forest ecosystems are an integral component of the global carbon cycle as they take up and release large amounts of C over short time periods (C flux) or accumulate it over longer time periods (C stock). However, there remains uncertainty about whether and in which direction C fluxes and in particular C stocks may differ between forests of high versus low species richness. Based on a comprehensive dataset derived from field-based measurements, we tested the effect of species richness (3-20 tree species) and stand age (22-116 years) on six compartments of above- and below-ground C stocks and four components of C fluxes in subtropical forests in southeast China. Across forest stands, total C stock was 149 ± 12 Mg ha-1 with richness explaining 28.5% and age explaining 29.4% of variation in this measure. Species-rich stands had higher C stocks and fluxes than stands with low richness; and, in addition, old stands had higher C stocks than young ones. Overall, for each additional tree species, the total C stock increased by 6.4%. Our results provide comprehensive evidence for diversity-mediated above- and below-ground C sequestration in species-rich subtropical forests in southeast China. Therefore, afforestation policies in this region and elsewhere should consider a change from the current focus on monocultures to multi-species plantations to increase C fixation and thus slow increasing atmospheric CO2 concentrations and global warming.

Biodiversity across trophic levels drives multifunctionality in highly diverse forests.

Human-induced biodiversity change impairs ecosystem functions crucial to human well-being. However, the consequences of this change for ecosystem multifunctionality are poorly understood beyond effects of plant species loss, particularly in regions with high biodiversity across trophic levels. Here we adopt a multitrophic perspective to analyze how biodiversity affects multifunctionality in biodiverse subtropical forests. We consider 22 independent measurements of nine ecosystem functions central to energy and nutrient flow across trophic levels. We find that individual functions and multifunctionality are more strongly affected by the diversity of heterotrophs promoting decomposition and nutrient cycling, and by plant functional-trait diversity and composition, than by tree species richness. Moreover, cascading effects of higher trophic-level diversity on functions originating from lower trophic-level processes highlight that multitrophic biodiversity is key to understanding drivers of multifunctionality. A broader perspective on biodiversity-multifunctionality relationships is crucial for sustainable ecosystem management in light of non-random species loss and intensified biotic disturbances under future environmental change.

Predicting reference soil groups using legacy data: A data pruning and Random Forest approach for tropical environment (Dano catchment, Burkina Faso).

Predicting taxonomic classes can be challenging with dataset subject to substantial irregularities due to the involvement of many surveyors. A data pruning approach was used in the present study to reduce such source errors by exploring whether different data pruning methods, which result in different subsets of a major reference soil groups (RSG) - the Plinthosols - would lead to an increase in prediction accuracy of the minor soil groups by using Random Forest (RF). This method was compared to the random oversampling approach. Four datasets were used, including the entire dataset and the pruned dataset, which consisted of 80% and 90% respectively, and standard deviation core range of the Plinthosols data while cutting off all data points belonging to the outer range. The best prediction was achieved when RF was used with recursive feature elimination along with the non-oversampled 90% core range dataset. This model provided a substantial agreement to observation, with a kappa value of 0.57 along with 7% to 35% increase in prediction accuracy for smaller RSG. The reference soil groups in the Dano catchment appeared to be mainly influenced by the wetness index, a proxy for soil moisture distribution.

Archaeopedological analysis of colluvial deposits in favourable and unfavourable areas: reconstruction of land use dynamics in SW Germany.

Colluvial deposits, as the correlate sediments of human-induced soil erosion, depict an excellent archive of land use and landscape history as indicators of human-environment interactions. This study establishes a chronostratigraphy of colluvial deposits and reconstructs past land use dynamics in the Swabian Jura, the Baar and the Black Forest in SW Germany. In the agriculturally favourable Baar area multiple main phases of colluvial deposition, and thus intensified land use, can be identified from the Neolithic to the Modern times. In the unfavourable Swabian Jura increased colluvial deposition began later compared to the more favourable areas in the Baar. The same holds true for the unfavourable areas of the Black Forest, but intensified land use can only be reconstructed for the Middle Ages and Early Modern times instead of for the Bronze and Iron Age as in the Swabian Jura. Land use intensity and settlement dynamics represented by thick, multilayered colluvial deposits increase in the Baar and the Black Forest during the Middle Ages. In between those phases of geomorphodynamic activity and colluviation, stable phases occur, interpreted as phases with sustainable land use or without human presence.

Belowground top-down and aboveground bottom-up effects structure multitrophic community relationships in a biodiverse forest.

Ecosystem functioning and human well-being critically depend on numerous species interactions above- and belowground. However, unraveling the structure of multitrophic interaction webs at the ecosystem level is challenging for biodiverse ecosystems. Attempts to identify major relationships between trophic levels usually rely on simplified proxies, such as species diversity. Here, we propose to consider the full information on species composition across trophic levels, using Procrustes correlation and structural equation models. We show that species composition data of a highly diverse subtropical forest-with 5,716 taxa across 25 trophic groups- reveal strong interrelationships among plants, arthropods, and microorganisms, indicating complex multitrophic interactions. We found substantial support for top-down effects of microorganisms belowground, indicating important feedbacks of microbial symbionts, pathogens, and decomposers on plant communities. In contrast, aboveground pathways were characterized by bottom-up control of plants on arthropods, including many non-trophic links. Additional analyses based on diversity patterns revealed much weaker interrelationships. Our study suggests that multitrophic communities in our forest system are structured via top-down effects of belowground biota on plants, which in turn affect aboveground arthropod communities across trophic levels. Moreover, the study shows that the consequences of species loss will be more complex than indicated by studies based solely on diversity.