Soil-landscape Model Research Articles

Digital soil mapping of soil burn severity

AbstractFire alters soil hydrologic properties leading to increased risk of catastrophic debris flows and post‐fire flooding. As a result, US federal agencies map soil burn severity (SBS) via direct soil observation and adjustment of rasters of burned area reflectance. We developed a unique application of digital soil mapping (DSM) to map SBS in the Creek Fire which burned 154,000 ha in the Sierra Nevada. We utilized 169 ground‐based observations of SBS in combination with raster proxies of soil forming factors, pre‐fire fuel conditions, and fire effects to vegetation to build a digital soil mapping model of soil burn severity (DSMSBS) using a random forest algorithm and compared the DSMSBS map to the established SBS map. The DSMSBS model had a cross‐validation accuracy of 48%. The established technique had 46% agreement between field observations and pixels. However, since the established technique is manual, it could not be compared to the DSMSBS model via cross‐validation. We produced SBS class uncertainty maps, which showed high prediction probabilities around field observations, and low probabilities away from field observations. SBS prediction probabilities could aid post‐fire assessment teams with sample prioritization. We report 107 km2 more area classified as high and moderate SBS compared to the established technique. We conclude that blending soil forming factors based mapping and vegetation burn severity mapping can improve SBS mapping. This represents a shift in SBS mapping away from validating remotely sensed reflectance imagery and toward a quantitative soil landscape model, which incorporates both fire and soils information to directly predict SBS.

Influence of sample size, model selection, and land use on prediction accuracy of soil properties

Digital soil mapping (DSM) uses models that integrate field and laboratory data with environmental factors to predict soils and soil properties. The accuracy of predictions depends on the models used, the data collected, and the environmental factors. This study assesses the influence of sampling density and distribution, covariates, and modeling approach on the prediction accuracy of soil organic matter (SOM) and cation exchange capacity (CEC) at three sites in Indiana (ACRE; DPAC; SEPAC) with different management intensity and sampling designs. Ordinary Kriging (OK) and three machine learning models Cubist (CB), Random Forest (RF), and Regression Kriging (RK) were used. The Coefficient of Determination (R2), Root Mean Square Error (RMSE), Mean Square Error (MSE), concordance coefficient (pc), and bias were used for the accuracy assessment.The accuracy of the predictions was influenced by the site, sample density, model type, soil property, and their interactions. Sites were the single largest source of significant variation followed by sampling density and model type for both SOM and CEC. ACRE, with multiple fields and complex management practices, had a higher average RMSE and wider range of RMSE for SOM compared to SEPAC and DPAC with uniform management. At ACRE the RMSE for SOM decreased from 2.75 to 0.85 and from 17.38 to 3.61 for CEC with increasing number of samples from 36 (6 points/ha) to 66 (12points/ha), but did not change with further increases up to 146 samples. At SEPAC and DPAC the RMSE decreased only slightly at sampling densities above 5 points/ha and 1–2 points/ha, respectively (68 and 43 samples, respectively). Based on cross validation, all models performed poorly for SOM with R2 varying from 0.13 to 0.38, while for CEC the model performance varied widely from 0.11 to 0.64. The accuracy predictions for CEC were higher compared to SOM at all sites. Overall, RF performed better while OK performed the worst for both SOM and CEC. The mean R2 values across all sites were 0.35 (SOM) and 0.51 (CEC) for RF and 0.19 (SOM) and 0.17 (CEC) for OK. At ACRE, OK performed worse for both SOM and CEC with only slight differences among the other models, while at SEPAC and DPAC there were only slight differences among all models. Spatial predictions for CB, RF and RK were more detailed and conformed to soil landscape models compared to OK. The spatial differences between sampling densities for predicted SOM and CEC were greater in lower elevation areas compared to higher elevation areas. The results from this study demonstrate that the selection of modeling approach is site-specific, and depends on sampling density, soil properties and their interactions.

Disentangling Jenny’s equation by machine learning

The so-called soil-landscape model is the central paradigm which relates soil types to their forming factors through the visionary Jenny’s equation. This is a formal mathematical expression that would permit to infer which soil should be found in a specific geographical location if the involved relationship was sufficiently known. Unfortunately, Jenny’s is only a conceptual expression, where the intervening variables are of qualitative nature, not being then possible to work it out with standard mathematical tools. In this work, we take a first step to unlock this expression, showing how Machine Learning can be used to predictably relate soil types and environmental factors. Our method outperforms other conventional statistical analyses that can be carried out on the same forming factors defined by measurable environmental variables.

How Much Carbon Do the Soils of Papua New Guinea Forests Contain?

Preliminary assessment of carbon in forest soils of Papua New Guinea (PNG) was done for 53 sites. Simple soil-landscape model was constructed to explain how soil carbon relates to landscape position and landform processes. Despite limitations of the soil survey, sufficient data have been obtained to indicate that the amount of carbon in forest soils surveyed so far is related primarily to their carbon-complexing capacity (CCC) which, in turn, depends on their geological parent material and parent rock. Soils were grouped into six categories based on their carbon content. The six categories ranged from very low 0–50 t/ha to extremely high 601–1,000 t/ha. Amalgamating the areas of mapped geological units on the basis of their likely CCC and their median carbon content has enabled the total carbon content of all forest soils in PNG to be estimated as 7,727 × 106 t. This figure is approximate and will be modified as more results are obtained and more sophisticated GIS-based landscape analysis is undertaken. The total carbon in forest biomass, as measured in this National Inventory and estimated for roots, is 4,006 × 106 t. Thus, the total carbon in PNG’s forests at present is estimated to be 11,733 × 106 t, with 66% stored in the soil and 34% in the biomass. The high proportion of ecosystem carbon held in the soil emphasises the importance of implementing soil conservation measures to maintain the total carbon resource in PNG’s forests.

Soil NIR-spectroscopy and object-based landsurface segmentation for fluvial terrace level differentiation

Establishing continuity of fluvial terrace remnants in eroded landscapes is often limited to resource intensive field interpretations of in-situ stratigraphic and physiographic features. Soil NIR spectra are an integrative property of soil that have been successfully combined with digital terrain data for soil-landscape modelling purposes. This study assesses the viability of soil NIR spectroscopy as a rapid and cost-effective way to differentiate between various fluvial terrace levels. First, the correlation between the NIR spectra of 88 soil samples and the height above river channel (HARC) of various fluvial terrace levels were investigated using pre-processed spectra, principle component analysis (PCA) and partial least square regression (PLSR). A strong correlation (R2 = 0.78, RPD = 2.11) was achieved using NIR-spectra located between 7500 and 5446 cm−1, which remained strong (R2 = 0.55, RPD = 1.49) after cross-validation. Next, the Scikit-learn random forest (RF) machine learning library was used to classify two sets of fluvial terrace level classes. The classification of two and three fluvial terrace classes produced, respective, validation scores of 0.73 and 0.76 when using the entire unprocessed spectral dataset as input. Interpretation of the subsequent feature importances indicated that spectral wavelength bands associated with the absorption characteristics of smectite, kaolinite, carbonates and talc were most important. Final validation scores of 0.74 (two fluvial terrace classes) and 0.76 (three fluvial terrace classes) were achieved by isolating specific wavelength bands associated with smectite (5334 cm−1 and 5269 cm−1) and kaolinite (3664 cm−1 and 3699 cm−1) fractions. This study demonstrates the clear potential of NIR spectroscopy as a data-driven alternative to in-situ stratigraphic and physiographic differentiation of fluvial terrace levels.

Effects of neighborhood analysis window forms and derivative algorithms on the soil aggregate stability – Landscape modeling

One of the most important factors in soil-landscape analysis can be the operational scale applied for mathematical calculations of terrain attributes. This study aimed to investigate the effects of different calculation patterns on terrain attributes and soil aggregate stability-landscape modeling at the subwatershed scale. For this purpose, four terrain attributes (slope gradient, slope aspect, plan curvature, and profile curvature) were calculated by two derivative algorithms (Evans-Young and Zevenbergen-Thorne) from a one-meter digital elevation model (DEM), using two neighborhood shapes (square and circular) with ten different neighborhood scales (from 5 m to 50 m). The regression tree model was used for modeling, and Kruskal-Wallis test was performed for determining significant differences in the terrain attributes across different calculation patterns. The results showed that the curvature attributes were mostly sensitive to the type of calculation pattern rather than the slope attributes. In calculation patterns, the neighborhood scale was the most important factor, which influenced differences in terrain attributes more than the neighborhood shape or type of derivative algorithm. The terrain attributes, calculated by the Evans-Young algorithm, were more distinct between circular and square neighborhoods, compared to attributes calculated by the Zevenbergen-Thorne algorithm, because of the difference in the polynomial type. Investigation of the relationship between the aggregate stability and terrain attributes indicated the great importance of the neighborhood scale. Differences in the R-squared values of models indicated that the type of derivative algorithms and the neighborhood shape could have remarkable effects on the soil-landscape models. Overall, the present results revealed that differences in calculation patterns had considerable effects on the soil-landscape analysis, although the results might be influenced by changes in the spatial resolution of DEM.

In-situ multi-sensor characterization of soil cores along an erosion-deposition gradient

Soil landscape research is faced with wide-ranging questions of soil erosion, precision farming, and agricultural risk management. Digital Soil Morphometrics is a powerful tool to provide respective answers or recommendations but requires soil data from the pedon-to-field scale with high horizontal and vertical resolutions, including the subsoil. We present an efficient sampling and measurement method for easily obtainable soil driving cores with low-destructive preparation. Elemental contents and soil organic and mineral matter composition were measured rapidly and in large numbers using a multi-sensor approach, i.e., visible and near-infrared (Vis-NIR), diffuse reflectance infrared Fourier transform (DRIFT), and X-ray fluorescence (XRF) spectroscopy. The suitability of the approach with respect to three-dimensional soil landscape models was tested using soils along a slope representing different stages of erosion and deposition in a hummocky landscape under arable land use (Calcaric Regosols, Calcic Luvisols, Luvic Stagnosols, Gleyic-Colluvic Regosols). The combination of soil core sampling, pedological description, and three spectroscopic techniques enabled rapid determination and interpretation of horizontal and vertical spatial distributions of soil organic carbon (SOC), soil organic and mineral matter composition, as well as CaCO3, Fe, and Mn contents. Depth profiles for SOC, CaCO3, and Fe contents were suitable indicators for site-specific degrees of erosion and matter transport processes at the pedon-to-field scale. Fe and Mn profiles helped identifying zones of reductive and oxic domains in subsoils (gleyzation). Further methodical developments should implement plant-availability of nutrients, characterization of Fe-oxides, and calibration of the spectroscopic techniques to field-moist samples.

Digital mapping of a soil profile

SummaryDetailed information on soil profiles is required for site characterization, soil mapping, studies on pedogenesis, soil–landscape modelling and soil classification. In this study, we used digital soil mapping techniques to map the profile wall of an Alfisol (90‐cm depth × 100‐cm width). Geochemical data (sampled at 10 cm × 10 cm) and digital images (resolution 1 cm × 1 cm) were used as input data in models that predicted a range of properties across the soil profile. Fuzzy c‐means clustering was applied to the profile and colour maps, and a confusion index was calculated. The colour coordinates were strongly correlated with SOC (soil organic carbon), sand and silt contents, and weathering indices. Random forest using the RGB model (R2 = 0.57, RMSE (root mean square error) = 2.41) showed less accurate prediction of SOC content than random forest using the CIE L*a*b* (R2 = 0.84, RMSE = 1.57) and HSV, hue, saturation, chroma (R2 = 0.85, RMSE = 1.46) models. Detailed profile maps showed considerable within‐horizon variation and identified animal holes, roots and limestone fragments. Colour and profile maps of soil properties and weathering indices matched field‐delineated soil horizons. Clusters obtained by soil properties and weathering indices represented the field‐delineated horizons and had the largest overall accuracy (P = 0.85 and 0.86, respectively). The confusion index showed that the horizons had wavy and somewhat irregular boundaries. We conclude that colour coordinates are useful for predicting and mapping soil properties in soil profiles. Random forest can be used to produce high‐resolution soil profile maps, and fuzzy c‐means clustering is useful for delineating soil horizons.HighlightsWe investigated the use of digital images in soil horizon delineation and within soil horizon variation We used various colour models to map soil profile properties and define horizons Colour and profile maps of soil properties and weathering indices matched field delineated soil horizons. Random forest can produce high‐resolution soil profile maps, and fuzzy c‐means clustering can delineate soil horizons

Soil Systems for Upscaling Saturated Hydraulic Conductivity for Hydrological Modeling in the Critical Zone

Core Ideas Saturated hydraulic conductivity (Ksat) was measured with different methods. Ksat was upscaled from point to catchment to watershed scales. Upscaled Ksat predicted streamflow at a large watershed without model calibration. A soil system approach was used to successfully upscale Ksat for streamflow predictions Successful hydrological model predictions depend on appropriate framing of scale and the spatial‐temporal accuracy of input parameters describing soil hydraulic properties. Saturated soil hydraulic conductivity (Ksat) is one of the most important properties influencing water movement through soil under saturated conditions. It is also one of the most expensive to measure and is highly variable. The objectives of this research were (i) to assess the ability of Amoozemeters, wells, piezometers, and flumes to accurately represent Ksat at a small catchment scale and (ii) to extrapolate Ksat to a larger watershed based on available soil data and soil landscape models for simulating streamflow using the Distributed Hydrological Soil Vegetation Model. The mean Ksat between Amoozemeters, wells, and flumes varied from 2.4 to 4.9 × 10−7 m s−1, and differences were not significant. Mixed trends in mean Ksat for slope positions and soil series were observed. The strongest significant and consistent trend in mean Ksat was observed for soil depth. The mean Ksat decreased exponentially with depth, from 6.51 × 106 m s−1 for upper horizons to 2.37 × 10−7 m s−1 for bottom horizons. Recognizing the significantly decreasing trend of Ksat with soil depth and the lack of consistent trends between soils and slope positions for small catchments, Ksat values were extrapolated from the small catchments occurring in Dillon Creek to another large watershed (Hall Creek) based on soil similarity and distribution. The Nash–Sutcliffe model overall efficiency of 0.52 indicated a good performance in simulating streamflows without model calibration. Combining Ksat measurement methods in small catchments with an understanding of soil landscapes and soil distribution relationships allowed successful upscaling of localized soil hydraulic properties for streamflow predictions to larger watersheds.

Soil depth spatial prediction by fuzzy soil-landscape model

Soil depth is a soil property that influences land use, land suitability, and earth surface processes. This article presents a simple method for predicting soil depth by constructing a membership function based on fuzzy C-means. This paper incorporates the soil type map, the land use map, and one type of DEM data to construct a soil-landscape model for soil depth prediction. It compares a fuzzy C-means classifier that includes expert judgment with a conditional autoregressive (CAR) model. Prediction efficiency was evaluated in the Three Gorges area of China using the root-mean-square error (RMSE) and the agreement coefficient (AC) of predictions at validation points. The prediction stability of soil depth values from the fuzzy model is close to the regression model; the AC value indicates a better agreement by the fuzzy C-means method (0.428) than when using the regression model (0.420). The purposive sampling approach was provided by our method by the centroid where the fuzzy membership value is above 0.85, which improves the efficiency of the field sampling. The expansibility of our method is limited as the typical centroid sample location is dependent on the study area. The fuzzy membership value must be recalculated to provide a new typical centroid for field sample when enlarging the study area. The results indicate that the soil-landscape model constructed by the fuzzy membership value with fuzzy C-means method and the conventional soil map provides better quality soil depth spatial information on soil depth.

Soil Lead Spatial Prediction Using a Fuzzy Soil-landscape Model

Soil lead pollution does great harm to the environment. The relation between soil lead and environmental covariates, however, is complicated. A fuzzy logic approach with expert knowledge has proven to be successful in mapping spatial variation of soil lead. The reasons are the limited availability of expert knowledge and the ignorance of soil type and land use of the non-soil area when predicting soil lead in current fuzzy clustering methodology. This paper incorporates the soil type map, the land use map and the DEM data to construct a soil\'96\'6 landscape model for soil lead prediction. It compares a fuzzy C-means classifier that includes expert judgment with conditional regressive model. Prediction efficiency was evaluated in the Three Gorges area of China using the root mean square error (RMSE) and the agreement coefficient (AC) of predictions at validation points. The result indicates that the soil-landscape model constructed by the fuzzy membership functions with fuzzy c-means method and the conventional soil map were able to produce good quality soil lead spatial information.

Digital soil mapping based on wavelet decomposed components of environmental covariates

Multi-scale soil variations are increasingly employed to improve the accuracy for digital soil mapping (DSM). In this study, we attempted to explore a methodology of wavelet analysis on this topic. The terrain attributes of a study area were decomposed using the wavelet analysis, and the resulted components were applied to map soil organic carbon (SOC) content, pH and clay content using multiple linear regression (MLR) and regression kriging (RK). The results showed that the wavelet components strengthened soil-landscape relationships in terms of correlation coefficients, enhanced soil-landscape modelling in terms of MLR modelling coefficients of determination (R2). Compared with several standard DSM approaches, i.e., ordinary kriging (OK), MLR and RK with the original terrain attributes, the use of wavelet components improved the prediction accuracy at some scales, but not all the scales. Most of the improvements were at the slight to moderate levels, e.g., 3.66–14.24% increases in the accuracy based on mean error, mean absolute error, root mean square error and R2. Maps made with wavelet components were relatively smooth and sometimes contained hotspots due to characteristics of wavelet components, which differed a lot from those made by the standard DSM methods. The potential benefits of using wavelet components as predictors in DSM may be further revealed in the future when more predictor selection approaches and mapping methods are considered.

Tree-based techniques to predict soil units

Quantitative soil–landscape models offer a method for conducting soil surveys that use statistical tools to predict natural patterns in the occurrence of particular map units across a landscape. The aim of the present study was to predict soil units in a watershed with wide variation in landscape conditions. The approach relied on a modelling of soil-forming factors in order to understand the variability of the landscape components in the region. Models were generated for landscape attributes related to pedogenesis, specifically elevation, slope, curvature, compound topographic index, Euclidean distance from stream networks, landforms map, clay minerals index, iron oxide index and normalised difference vegetation index, along with an existing geology map. The soil classification was adapted from the World Reference Base System for Soil Resources, and the predominant soil taxonomic orders observed were Ferrasols, Acrisols, Gleysols, Cambisols, Fluvisols and Regosols. The algorithms used to predict the soil units were based on decision tree (DT) and random forest (RF) methods. The criteria used to evaluate the models’ performance were statistical indices, coherence between predicted units and the legacy map, as well as accuracy checks based on control samples. The best performing model was found to be the RF algorithm, with resulting statistical indices considered excellent (overall=0.966, kappa=0.962). The accuracy of the map as determined by control points was 67.89%, with a kappa value of 61.39%.

Modelling long-term soil organic carbon dynamics under the impact of land cover change and soil redistribution

During the last millennia, anthropogenic land cover change has caused a significant release of carbon from soils while also accelerating rates of soil erosion by one to two orders of magnitude. However, mechanistic frameworks that explicitly link land cover change, erosion and soil organic carbon (SOC) cycling and, in addition, can be applied to longer timescales are currently lacking. This study presents a 4D soil-landscape model coupling sediment fluxes and carbon dynamics in response to continuous land cover change over a period of 1000years. We applied the model to a well-studied catchment of 591km2 in Central Belgium. The model evaluation showed that the simulated magnitude and spatial patterns of soil redistribution and SOC stocks are in good agreement with field observations. At the catchment scale, land cover change over the last 1000years decreased the SOC stock of the top 1m of soil by 33.1tC·ha−1 (32% loss compared to the initial SOC stock) while in contrast, erosion resulted in an uptake of 31.9tC·ha−1 from the atmosphere (29% gain compared to the initial SOC stock). As a result, the combined effect is a small source of c. 1.2tC·ha−1 over the last 1000years. While the SOC released from the soil to the atmosphere quickly responded to land conversion, the erosion induced sink operated at a much lower intensity but with a longer duration. These transient simulations show that the soil carbon budget is highly variable in both space and time.

Soil – landscape relationships in the Empordà basin (Catalonia, NE Iberian Peninsula)

Soils developed in representative landforms, which were previously mapped at a detailed scale in the Empordà Basin, were selected to characterize their main pedogenetic processes and to improve the soil maps through a better understanding of the soil – landscape relationships. This basin is a relatively large region (1,300 km<sup>2</sup>) in Northeastern Catalonia, where Neogene and Quaternary sediments outcrop. They are alluvial and delta fan deposits that mainly reflect a continental environment. Besides varying degrees of soil rubefaction, we can identify calcium carbonate redistribution, clay illuviation and sodication as the main soil forming processes, together with abrupt textural changes, vertic and redoximorphic features. These processes and features are expressed under different morphologies in the area, depending not only on parent material, landform and age, but also on human action, which allows us to refine the conceptual soil-landscape model. Calcium carbonate redistribution is a key process reflecting both changing general environmental conditions and local chemical soil conditions. The actual soil characteristics and the soil forming processes allow us to propose that (i) aeolian dust inputs in these soils have been low to moderate throughout, and that (ii) the rainfall pattern in the last part of the Holocene was able to remove these dust inputs, but unable to leach carbonates from medium textured, moderately calcareous soils in the area when they are some kilometres from the sea and not directly affected by the dune system.

Estimation of Soil Texture at a Regional Scale Using Local Soil-Landscape Models

Estimation of Soil Texture at a Regional Scale Using Local Soil-Landscape Models

Mapping the soils of an Argentine Pampas region using structural equation modelling

Current digital soil mapping (DSM) methods have limitations. For instance, it is difficult to predict a large number of soil properties simultaneously, while preserving the relationships between them. Another problem is that prevalent prediction models use pedological knowledge in a very crude way only. To tackle these problems, we investigated the use of structural equation modelling (SEM). SEM has its roots in the social sciences and is recently also being used in other scientific disciplines, such as ecology. SEM integrates empirical information with mechanistic knowledge by deriving the model equations from known causal relationships, while estimating the model parameters using the available data. It distinguishes between endogenous and exogenous variables, where, in our application, the first are soil properties and the latter are external soil forming factors (i.e. climate, relief, organisms). We introduce SEM theory and present a case study in which we applied SEM to a 22,900km2 region in the Argentinian Pampas to map seven key soil properties. In this case study, we started with identifying the main soil forming processes in the study area and assigned for each process the main soil properties affected. Based on this analysis we defined a conceptual soil-landscape model, which was subsequently converted to a SEM graphical model. Finally, we derived the SEM equations and implemented these in the statistical software R using the latent variable analysis (lavaan) package. The model was calibrated using a soil dataset of 320 soil profile data and 12 environmental covariate layers. The outcomes of the model were maps of seven soil properties and a SEM graph that shows the strength of the relationships. Although the accuracy of the maps, based on cross-validation and independent validation, was poor, this paper demonstrates that SEM can be used to explicitly include pedological knowledge in prediction of soil properties and modelling of their interrelationships. It bridges the gap between empirical and mechanistic methods for soil-landscape modelling, and is a tool that can help produce pedologically sound soil maps.

Arctic soil development on a series of marine terraces on central Spitsbergen, Svalbard: a combined geochronology, fieldwork and modelling approach

Abstract. Soils in Arctic regions currently enjoy attention because of their sensitivity to climate change. It is therefore important to understand the natural processes and rates of development of these soils. Specifically, there is a need to quantify the rates and interactions between various landscape- and soil-forming processes. Soil chronosequences are ideal natural experiments for this purpose. In this contribution, we combine field observations, luminescence dating and soil–landscape modelling to improve and test our understanding of Arctic soil formation. The field site is a Holocene chronosequence of gravelly raised marine terraces in central Spitsbergen. Field observations show that soil–landscape development is mainly driven by weathering, silt translocation, aeolian deposition and rill erosion. Spatial soil variation is mainly caused by soil age, morphological position within a terrace and depth under the surface. Luminescence dating confirmed existing radiocarbon dating of the terraces, which are between ∼ 1.5 and ∼ 13.3 ka old. The soil–landscape evolution model LORICA was used to test our hypothesis that the field-observed processes indeed dominate soil–landscape development. Model results additionally indicated the importance of aeolian deposition as a source of fine material in the subsoil for both sheltered and vegetated trough positions and barren ridge positions. Simulated overland erosion was negligible. Consequently, an un-simulated process must be responsible for creating the observed erosion rills. Dissolution and physical weathering both play a major role. However, using present-day soil observations, the relative contribution of physical and chemical weathering could not be disentangled. Discrepancies between field and model results indicate that soil formation is non-linear and driven by spatially and temporally varying boundary conditions which were not included in the model. To conclude, Arctic soil and landscape development appears to be more complex and less straightforward than could be reasoned from field observations.

COMPARISON OF THREE MODELS FOR PREDICTING THE SPATIAL DISTRIBUTION OF SOIL ORGANIC CARBON IN BOALEMO REGENCY, SULAWESI

<p><em>Information on the spatial distribution of soil organic carbon content is required for sustainable land management. But, creating this map is time consuming and costly. Digital soil mapping methodology make use legacy soil data to create provisional soil organic carbon map. This map helps soil surveyors in allocating next soil observation. This study aimed: (i) to develop predictive statistical soil organic carbon models for Sulawesi, and (ii) to evaluate the best model between the three obtained models. Boalemo Regeny in Gorontalo Province (Sulawesi) was selected as studying area due to abundant legacy soil data. The study covered dataset preparation, model development, and model comparison. Dataset of soil organic carbon at 6 different depths as target was established from 176 soil profiles and 7 terrain parameters were selected as predictors. Soil-landscape models for each soil depth were created using regression tree, conditional inference tree, and multiple linear regression technique. Result showed that model performance differed among 3 modelling techniques and soil depths. The tree models were better than the multiple linear regression model as they have the lowest RMSE index. The best model in the mountanious area seems to be the regression tree model, whereas in the plains it may be the conditional inference tree. In creating provisional map, several model should be developed and the median of predicted value is used as provisional map.</em></p><p><em> </em></p><p><em>Keywords: Digital soil mapping, multiple linear regression, regression tree, soil-landscape model, soil organic carbon map</em></p>

Predicting the Distribution of Naturally Occurring Phosphatic Soils across a Countywide Landscape, Florida, USA

We used a GIS approach to test a conceptual soil-landscape model which integrates topographic, geologic and soil information for predicting the distribution of phosphatic soils. The hypothesis was that phosphatic soils occur on erosional landscapes where phosphatic geologic materials are exposed at or near land surface. Areas of high (scarp), medium (plain), and low (plateau) likelihood of such exposure were delineated based on topographic and geologic information. An initial predictive phosphatic soils map was created and validated by field sampling. Samples were taken at two depths: upper (0–25 cm) and lower (100–125 cm); and the total phosphorus (TP) was determined. Mean TP concentration for the lower depth was 6195 mg kg−1 for the scarp, 2485 mg kg−1 for the plain, and 193 mg kg−1 for the plateau. The initial map was revised to show that the probability of finding a phosphatic soil was highest for the scarp region (73%), medium for the plain region (33%), and lowest for the plateau region (7%).