Assessing countrywide soil organic carbon stock using hybrid machine learning modelling and legacy soil data in Cameroon

Silatsa, Francis B. T.; Yemefack, Martin; Tabi, Fritz Oben; Heuvelink, G.B.M.; Leenaars, J.G.B.

doi:10.1016/j.geoderma.2020.114260

Cited by 47 publications

(23 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several remote sensing techniques that varying depending on their spatial, spectral, temporal and radiometric resolution and the platforms that are mounted on (spaceborne platforms, airborne platforms and unmanned aerial systems) can help to quantify the soil carbon sequestration [ 34 ]. Gholizadeh et al [ 110 ] found that Red and NIR bands, also spectral indices BI, SAVI among others provide the strongest correlations with SOC. Which agrees with the results of this study, where the BI was found as SOC predictor, so too the TOA Brightness Temperature.…”

Section: Discussionmentioning

confidence: 99%

Multi-predictor mapping of soil organic carbon in the alpine tundra: a case study for the central Ecuadorian páramo

Izurieta

Márquez

García

et al. 2021

Carbon Balance Manage

View full text Add to dashboard Cite

Background Soil organic carbon (SOC) affects essential biological, biochemical, and physical soil functions such as nutrient cycling, water retention, water distribution, and soil structure stability. The Andean páramo known as such a high carbon and water storage capacity ecosystem is a complex, heterogeneous and remote ecosystem complicating field studies to collect SOC data. Here, we propose a multi-predictor remote quantification of SOC using Random Forest Regression to map SOC stock in the herbaceous páramo of the Chimborazo province, Ecuador. Results Spectral indices derived from the Landsat-8 (L8) sensors, OLI and TIRS, topographic, geological, soil taxonomy and climate variables were used in combination with 500 in situ SOC sampling data for training and calibrating a suitable predictive SOC model. The final predictive model selected uses nine predictors with a RMSE of 1.72% and a R2 of 0.82 for SOC expressed in weight %, a RMSE of 25.8 Mg/ha and a R2 of 0.77 for the model in units of Mg/ha. Satellite-derived indices such as VARIG, SLP, NDVI, NDWI, SAVI, EVI2, WDRVI, NDSI, NDMI, NBR and NBR2 were not found to be strong SOC predictors. Relevant predictors instead were in order of importance: geological unit, soil taxonomy, precipitation, elevation, orientation, slope length and steepness (LS Factor), Bare Soil Index (BI), average annual temperature and TOA Brightness Temperature. Conclusions Variables such as the BI index derived from satellite images and the LS factor from the DEM increase the SOC mapping accuracy. The mapping results show that over 57% of the study area contains high concentrations of SOC, between 150 and 205 Mg/ha, positioning the herbaceous páramo as an ecosystem of global importance. The results obtained with this study can be used to extent the SOC mapping in the whole herbaceous ecosystem of Ecuador offering an efficient and accurate methodology without the need for intensive in situ sampling.

show abstract

Section: Discussionmentioning

confidence: 99%

Multi-predictor mapping of soil organic carbon in the alpine tundra: a case study for the central Ecuadorian páramo

Izurieta

Márquez

García

et al. 2021

Carbon Balance Manage

View full text Add to dashboard Cite

show abstract

“…Interest in the use of agricultural offsets as a way to mitigate climate change has hastened efforts to develop national and regional soil organic carbon (SOC) inventory (1)(2)(3)(4) that can incentivize carbon sequestration and greenhouse gas reduction (5) and inform end-users such as stakeholders, professional organizations, and policymakers (6). Such efforts often utilize digital soil mapping (DSM) and statistical modeling approaches along with the ever-improving observations and inventories of soil and environmental covariates (7)(8)(9)(10).…”

Section: Introductionmentioning

confidence: 99%

“…ssoil texture(19), soil type(14), soil order or suborder(6), bare soil reflectance(5), soil available water capacity (4), soil erosion rate (4), soil alkalinity (3), mean soil particle size (3), soil depth (3), gravel content (3), soil group or subgroup (3), soil pH (3), soil carbonate contents or index (3), cation exchange capacity (2), soil mapping unit (2), K content (2), P content (2), N content (2), soil moisture (2), rate of river network development and persistence (2), electrical conductivity (1), sum of exchange cations (1), drainage class (1), inherit fertility rating (1), soil intensity index (1), Ca content (1), Mg content (1), Na content (1), soil structure (1) c precipitation (67), temperature (66), potential evapotranspiration (13), solar radiation (9), relative humidity (4), vapor pressure deficit (4), water or moisture regime (3), direct or diffuse insolation (2), aridity index (1), duration of sunshine (1), ecological region (1), Emberger index (1), hydro-thermal coefficient (1), Martonne index (1), geographical region (1) o land use, land cover, or vegetation form and cover (54), normalized difference vegetation index (35), sensor-based surface vegetation reflectance (17), enhanced or soil-adjusted vegetation index (10), net primary production (…”

mentioning

confidence: 99%

Digital Mapping of Agricultural Soil Organic Carbon Using Soil Forming Factors: A Review of Current Efforts at the Regional and National Scales

Xia

McSweeney²,

Wander³

2022

Front. Soil Sci.

View full text Add to dashboard Cite

To explore how well large spatial scale digital soil mapping can contribute to efforts to monitor soil organic carbon (SOC) stocks and changes, we reviewed regional and national studies quantifying SOC within lands dominated by agriculture using SCORPAN approaches that rely on soil (S), climate (C), organisms (O), relief (R), parent material (P), age (A), and space (N) covariates representing soil forming factors. After identifying 79 regional (> 10,000 km2) and national studies that attempted to estimate SOC, we evaluated model performances with reference to soil sampling depth, number of predictors, grid-distance, and spatial extent. SCORPAN covariates were then investigated in terms of their frequency of use and data sources. Lastly, we used 67 studies encompassing a variety of spatial scales to determine which covariates most influenced SOC in agricultural lands using a subjective ranking system. Topography (used in 94% of the cases), climate (87%), and organisms (86%) covariates that were the most frequently used SCORPAN predictors, aligned with the factors (precipitation, temperature, elevation, slope, vegetation indices, and land use) currently identified to be most influential for model estimate at the large spatial extent. Models generally succeeded in estimating SOC with fits represented by R2 with a median value of 0.47 but, performance varied widely (R2 between 0.02 and 0.86) among studies. Predictive success declined significantly with increased soil sampling depth (p < 0.001) and spatial extent (p < 0.001) due to increased variability. While studies have extensively drawn on large-scale surveys and remote sensing databases to estimate environmental covariates, the absence of soils data needed to understand the influence of management or temporal change limits our ability to make useful inferences about changes in SOC stocks at this scale. This review suggests digital soil mapping efforts can be improved through greater use of data representing soil type and parent material and consideration of spatio-temporal dynamics of SOC occurring within different depths and land use or management systems.

show abstract

“…ML's overall purpose is to discover patterns in data that inform how problems that are not visible are addressed [30,31]. Numerous studies integrated a ML technique to spatial interpolation models such as Artificial Neural Network (ANN) for solar radiation estimation [32]; deep learning for seismic intensity [33]; ensemble ANN for atmospheric studies [34]; decision tree (DT) approach for land cover data and sodium absorption [35,36]; support vector machine (SVM) for basin precipitation [37]; long short-term memory (LSTM) neural network for PM2.5 [38]; extremely randomized trees for meteorological drought forecasting [39]; support vector regression (SVR) and correlation-based feature selection (CFS) for vehicular emissions prediction [40]; stochastic gradient boosting, cubist, random forest (RF), and model averaged neural networks for temperature maps [41]; random forest for solar radiation observation [42]; ensemble prediction approach for lake acidity prediction [43]; RF and generalized boosted regression (GBR) for soil organic carbon [44]; Non-linear AutoRegressive eXogenous (NARX) model for groundwater (GW) level prediction [45][46][47]; dynamic and long-term prediction of toxic HM [48,49]; and water quality prediction [50]. The use of ML integrated with spatial interpolation technique qualifies as an innovative superior substitute for traditional data application approaches due to its capability to distinguish non-linear associations among numerous constraints as opposed to other techniques that assume these connections are linear [51].…”

Section: Introductionmentioning

confidence: 99%

A Hybrid Neural Network–Particle Swarm Optimization Informed Spatial Interpolation Technique for Groundwater Quality Mapping in a Small Island Province of the Philippines

Jesus

Senoro

Cruz

et al. 2021

Toxics

View full text Add to dashboard Cite

Water quality monitoring demands the use of spatial interpolation techniques due to on-ground challenges. The implementation of various spatial interpolation methods results in significant variations from the true spatial distribution of water quality in a specific location. The aim of this research is to improve mapping prediction capabilities of spatial interpolation algorithms by using a neural network with the particle swarm optimization (NN-PSO) technique. Hybrid interpolation approaches were evaluated and compared by cross-validation using mean absolute error (MAE) and Pearson's correlation coefficient (R). The governing interpolation techniques for the physicochemical parameters of groundwater (GW) and heavy metal concentrations were the geostatistical approaches combined with NN-PSO. The best methods for physicochemical characteristics and heavy metal concentrations were observed to have the least MAE and R values, ranging from 1.7 to 4.3 times and 1.2 to 5.6 times higher than the interpolation technique without the NN-PSO for the dry and wet season, respectively. The hybrid interpolation methods exhibit an improved performance as compared to the non-hybrid methods. The application of NN-PSO technique to spatial interpolation methods was found to be a promising approach for improving the accuracy of spatial maps for GW quality.

show abstract

Assessing countrywide soil organic carbon stock using hybrid machine learning modelling and legacy soil data in Cameroon

Cited by 47 publications

References 69 publications

Multi-predictor mapping of soil organic carbon in the alpine tundra: a case study for the central Ecuadorian páramo

Multi-predictor mapping of soil organic carbon in the alpine tundra: a case study for the central Ecuadorian páramo

Digital Mapping of Agricultural Soil Organic Carbon Using Soil Forming Factors: A Review of Current Efforts at the Regional and National Scales

A Hybrid Neural Network–Particle Swarm Optimization Informed Spatial Interpolation Technique for Groundwater Quality Mapping in a Small Island Province of the Philippines

Contact Info

Product

Resources

About