T.G. Orton

Evaluation of modelling approaches for predicting the spatial distribution of soil organic carbon stocks at the national scale

Abstract Soil organic carbon (SOC) plays a major role in the global carbon budget. It can act as a source or a sink of atmospheric carbon, thereby possibly influencing the course of climate change. Improving the tools that model the spatial distributions of SOC stocks at national scales is a priority, both for monitoring changes in SOC and as an input for global carbon cycles studies. In this paper, we compare and evaluate two recent and promising modelling approaches. First, we considered several increasingly complex boosted regression trees (BRT), a convenient and efficient multiple regression model from the statistical learning field. Further, we considered a robust geostatistical approach coupled to the BRT models. Testing the different approaches was performed on the dataset from the French Soil Monitoring Network, with a consistent cross-validation procedure. We showed that when a limited number of predictors were included in the BRT model, the standalone BRT predictions were significantly improved by robust geostatistical modelling of the residuals. However, when data for several SOC drivers were included, the standalone BRT model predictions were not significantly improved by geostatistical modelling. Therefore, in this latter situation, the BRT predictions might be considered adequate without the need for geostatistical modelling, provided that i) care is exercised in model fitting and validating, and ii) the dataset does not allow for modelling of local spatial autocorrelations, as is the case for many national systematic sampling schemes.

Generic Issues on Broad-Scale Soil Monitoring Schemes： A Review

Numerous scientific challenges arise when designing a soil monitoring network (SMN), especially when assessing large areas and several properties that are driven by numerous controlling factors of various origins and scales. Different broad approaches to the establishment of SMNs are distinguished. It is essential to establish an adequate sampling protocol that can be applied rigorously at each sampling location and time. We make recommendations regarding the within-site sampling of soil. Different statistical methods should be associated with the different types of sampling design. We review new statistical methods that account for different sources of uncertainty. Except for those parameters for which a consensus exists, the question of testing method harmonisation remains a very difficult issue. The establishment of benchmark sites devoted to harmonisation and inter-calibration is advocated as a technical solution. However, to our present knowledge, no study has addressed crucial scientific issues such as how many calibration sites are necessary and how to locate them.

Estimation of Soil Carbon Input in France: An Inverse Modelling Approach

Development of a quantitative understanding of soil organic carbon (SOC) dynamics is vital for management of soil to sequester carbon (C) and maintain fertility, thereby contributing to food security and climate change mitigation. There are well-established process-based models that can be used to simulate SOC stock evolution; however, there are few plant residue C input values and those that exist represent a limited range of environments. This limitation in a fundamental model component (i.e., C input) constrains the reliability of current SOC stock simulations. This study aimed to estimate crop-specific and environment-specific plant-derived soil C input values for agricultural sites in France based on data from 700 sites selected from a recently established French soil monitoring network (the RMQS database). Measured SOC stock values from this large scale soil database were used to constrain an inverse RothC modelling approach to derive estimated C input values consistent with the stocks. This approach allowed us to estimate significant crop-specific C input values (P -1 year -1 (silage corn) to 5.15 ± 0.12 t C ha -1 year -1 (grassland/pasture). Furthermore, the incorporation of climate variables improved the predictions. C input of 4 crop types could be predicted as a function of temperature and 8 as a function of precipitation. This study offered an approach to meet the urgent need for crop-specific and environment-specific C input values in order to improve the reliability of SOC stock prediction.

First evidence of large-scale PAH trends in French soils

Polycyclic aromatic hydrocarbons (PAHs) are widespread organic pollutants. Soils are a reservoir of PAHs because some soil constituents favour PAH accumulation. Therefore, soil is a key indicator of the degree of contamination. So far, studies mapping soil PAH levels over large territories are very rare. Here, we report the first nation-wide maps of soil PAHs in France. Results were obtained within the French National Soil Monitoring Network, which is the first European network monitoring systematically soil PAHs. We used advanced geostatistics to map PAH distribution over the whole French territory. Our results show clear trends of PAH levels at the nation scale. For instance, the highest PAH levels are found in Northern and Eastern France. This high contamination is explained by the intense industrial activity of these regions during the last century. High levels of PAH are also found near some coastlines. This observation could be explained by long-range atmospheric transportation. In addition, we found that light PAHs are rarely found in French topsoils.

Spatial distribution of Lindane concentration in topsoil across France.

Lindane [γ-hexachlorocyclohexane (γ-HCH)] is an organochlorine pesticide with toxic effects on humans. It is bioaccumulative and can remain in soils for long periods, and although its use for crop spraying was banned in France in 1998, it is possible that residues from before this time remain in the soil. The RMQS soil monitoring network consists of soil samples from 2200 sites on a 16 km regular grid across France, collected between 2002 and 2009. We use 726 measurements of the Lindane concentration in these samples to (i) investigate the main explanatory factors for its spatial distribution across France, and (ii) map this distribution. Geostatistics provides an appropriate framework to analyze our spatial dataset, though two issues regarding the data are worth special consideration: first, the harmonization of two subsets of the data (which were analyzed using different measurement processes), and second, the large proportion of data from one of these subsets that fell below a limit of quantification. We deal with these issues using recent methodological developments in geostatistics. Results demonstrate the importance of land use and rainfall for explaining part of the variability of Lindane across France: land use due to the past direct input of Lindane on cropland and its subsequent persistence in the soil, and rainfall due to the re-deposition of volatilized Lindane. Maps show the concentrations to be generally largest in the north and northwest of France, areas of more intensive agricultural land. We also compare levels to some contamination thresholds taken from the literature, and present maps showing the probability of Lindane concentrations exceeding these thresholds across France. These maps could be used as guidelines for deciding which areas require further sampling before some possible remediation strategy could be applied.

Impacts of fire on soil organic carbon stocks in a grazed semi-arid tropical Australian savanna: accounting for landscape variability

Fire and grazing are commonplace in Australian tropical savannas and the effects of these management practices on soil organic carbon stocks (SOC) is not well understood. A long-term (20 years) experiment studying the effects of fire on a grazed semi-arid tropical savanna was used to increase this understanding. Treatments, including frequency of fire (every 2, 4 and 6 years), season of fire [early (June) vs late (October) dry season] and unburnt control plots, were imposed on Vertosol grassland and Calcarosol woodland sites, which were grazed. Additionally long-term enclosures [unburnt (except the Calcarosol in 2001) and ungrazed since 1973] on each soil type adjacent to each site were sampled, although not included in statistical analyses. SOC stocks were measured to a soil depth of 0.3 m using a wet oxidation method (to avoid interference by carbonates) and compared on an equivalent soil mass basis. Significant treatment differences in SOC stocks were tested for, while accounting for spatial background variation within each site. SOC stocks (0–0.3 m soil depth) ranged between 10.1 and 28.9 t ha–1 (Vertosol site) and 20.7 and 54.9 t ha–1 (Calcarosol site). There were no consistent effects of frequency or season of fire on SOC stocks, possibly reflecting the limited statistical power of the study and inherent spatial variability observed. Differences in the response to frequency and season of fire observed between these soils may have been due to differences in clay type, plant species composition and/or preferential grazing activity associated with fire management. There may also have been differences in C input between treatments and sites due to differences in the herbage mass and post-fire grazing activity on both sites and changed pasture composition, higher herbage fuel load, and a reduction in woody cover on the Vertosol site. This study demonstrated the importance of accounting for background spatial variability and treatment replication (in the absence of baseline values) when assessing SOC stocks in relation to management practices. Given the absence of baseline SOC values and the potentially long period required to obtain changes in SOC in rangelands, modelling of turnover of SOC in relation to background spatial variability would enable management scenarios to be considered in relation to landscape variation that may be unrelated to management. These considerations are important for reducing uncertainty in C-flux accounting and to provide accurate and cost-effective methods for land managers considering participation in the C economy.

On Soil Carbon Monitoring Networks

The design of a Soil Monitoring Network (SMN) poses numerous scientific challenges, especially for the assessment of national or continental areas. The task is particularly challenging because soil carbon content and stocks are driven by controlling factors of disparate origins and scales. Various approaches to the establishment of SMNs are reviewed here. Frameworks for soil monitoring exist in numerous countries, especially in Europe. Although some countries work using standard monitoring methodologies and coverage, there is considerable variation in approaches to the monitoring of soil carbon even within a country. In addition to achieving harmonization, there are generic issues which must be addressed when SMNs are established and operated: the SMN should be effective for different soils, and it must enable the detection of change in soil carbon at relevant spatial and temporal scales with adequate precision and statistical power. We present examples which address these issues and summarize previous reviews on this topic. It is essential to establish an adequate sampling protocol which can be applied at each sampling location and time. The design must address the questions that the user of data has and provide information with accuracy and precision at the spatial and temporal scales that match the users’ needs. Furthermore, the design must match the methods of analysis so that statistical assumptions can be justified. At the global scale, the question of harmonizing sampling and analytical methods is difficult. Here, we propose the establishment of benchmark sites devoted to harmonization and inter-calibration. We present a case study from France which addresses scientific issues such as how many calibration sites are necessary and how they should be selected.

Analyzing the spatial distribution of PCB concentrations in soils using below-quantification limit data

Polychlorinated biphenyls (PCBs) are highly toxic environmental pollutants that can accumulate in soils. We consider the problem of explaining and mapping the spatial distribution of PCBs using a spatial data set of 105 PCB-187 measurements from a region in the north of France. A large proportion of our data (35%) fell below a quantification limit (QL), meaning that their concentrations could not be determined to a sufficient degree of precision. Where a measurement fell below this QL, the inequality information was all that we were presented with. In this work, we demonstrate a full geostatistical analysis-bringing together the various components, including model selection, cross-validation, and mapping-using censored data to represent the uncertainty that results from below-QL observations. We implement a Monte Carlo maximum likelihood approach to estimate the geostatistical model parameters. To select the best set of explanatory variables for explaining and mapping the spatial distribution of PCB-187 concentrations, we apply the Akaike Information Criterion (AIC). The AIC provides a trade-off between the goodness-of-fit of a model and its complexity (i.e., the number of covariates). We then use the best set of explanatory variables to help interpolate the measurements via a Bayesian approach, and produce maps of the predictions. We calculate predictions of the probability of exceeding a concentration threshold, above which the land could be considered as contaminated. The work demonstrates some differences between approaches based on censored data and on imputed data (in which the below-QL data are replaced by a value of half of the QL). Cross-validation results demonstrate better predictions based on the censored data approach, and we should therefore have confidence in the information provided by predictions from this method.

The effects of glyphosate, glufosinate, paraquat and paraquat-diquat on soil microbial activity and bacterial, archaeal and nematode diversity

In this study, we investigated the effects of one-off applications of glyphosate, glufosinate, paraquat, and paraquat-diquat on soil microbial diversity and function. All herbicides were added to soil as pure compounds at recommended dose and were incubated under laboratory conditions for 60 days. High-throughput phylogenetic marker gene sequencing revealed that none of the herbicides significantly influenced the richness, evenness and composition of bacterial and archaeal communities. Likewise, the diversity, composition and size of nematode communities were not significantly influenced by any of the herbicides. From a functional perspective, herbicides did not significantly affect fluorescein diacetate hydrolysis (FDA) and beta-glucosidase activities. Furthermore, the ability of soil organisms to utilise 15 substrates was generally unaffected by herbicide application. The only exception to this was a temporary impairment in the ability of soil organisms to utilise three organic acids and an amino acid. Given the global and frequent use of these herbicides, it is important that future studies evaluate their potential impacts on microbial communities in a wider-range of soils and environmental conditions.

Effects of land-use change and management on soil carbon and nitrogen in the Brigalow Belt, Australia: II. Statistical models to unravel the climate-soil-management interaction

The impact of grazing on soil carbon (C) and nitrogen (N) cycles is complex, and across a large area it can be difficult to uncover the magnitude of the effects. Here, we have linked two common approaches to statistical modelling – regression trees and linear mixed models – in a novel way to explore various aspects of soil C and N dynamics for a large, semiarid bioregion where land use is dominated by grazing. The resulting models, which we term RT-LMM, have the pleasing visual appeal of regression trees, and they account for spatial autocorrelation as per a linear mixed model. Our RT-LMM were developed from explanatory variables that related information on climate, soil and past land management. Response variables of interest were: stocks of soil total organic carbon (TOC), soil total nitrogen (TN), and particulate organic C (POC); the ratio of TOC stock to TN stock; and the relative abundance of stable isotopes δ13C and δ15N in the soil. Each variable was sampled at the depth interval 0–0.3 m. The interactions of land use with, in particular, air temperature and soil phosphorus were strong, but three principal management-related effects emerged: (i) the use of fire to clear native vegetation reduced stocks of TOC and TN, and the TOC : TN ratio, by 25%, 19% and 9%, respectively, suggesting that TOC is more sensitive to fire than TN; (ii) conversion of native vegetation to pasture enriched soil with δ13C by 1.7 ‰; subsequent regrowth of the native vegetation among the pasture restored δ13C to its original level but there was no corresponding change in TOC stock; and, (iii) the time elapsed since clearing reduced POC stocks and the TOC : TN ratio.