A Pedotransfer Function to Map Soil Bulk Density from Limited Data

Development of land-use-specific pedotransfer functions for predicting bulk density of acidic topsoil in eastern Himalayas (India)

با برآورد ظرفیت تبادل کاتیونی با استفاده از ویژگی‏های پایه ای و زودیافت خاک می‏توان در وقت و هزینه صرفه‏جویی کرد. هدف این بررسی، مقایسه دو روش رگرسیون درختی و رگرسیون خطی چندگانه در برآورد ظرفیت تبادل کاتیونی (CEC) با استفاده از ویژگی‏هایخاک است. برای این منظور از داده های 106 نمونه خاک UNSODA استفاده شد. جهت برآورد CEC با استفاده از روش‏های رگرسیون درختی و رگرسیون خطی از ویژگی های اجزای بافت خاک،pH، ماده آلی و چگالی ظاهری استفاده شد. کارآیی روش رگرسیون درختی در برابر روش رگرسیون خطی چندگانه در برآورد CEC مقایسه شدند. نتایج نشان داد که در برآورد CEC با استفاده از روش های رگرسیونی تنها ضرایب ماده آلی (183/3) و درصد رس (274/0) که بیشترین همبستگی را با CEC دارند معنی دار شده و مدل رگرسیونی بر اساس این دو پارامتر توسعه یافت. همچنین از میان پارمترهای ورودی در روش رگرسیون درختی تنها پارامترهای ماده آلی و درصد رس در درخت رگرسیون ظاهر شد. روش رگرسیون درختی در دو مجموعه داده صحت سنجی و اعتبار سنجی بر اساس آماره های ارزیابی R2، RMSE، ME و GMER کارآیی بالاتری نسبت به روش های رگرسیونی خطی در برآورد CEC داشت. از میان روش های رگرسیونی خطی، مدل پیشنهادی کارآیی بالاتری نسبت به مدل های بل و ونکولن و بروسما و همکاران داشت.

Water retention and saturated hydraulic conductivity are soil properties that are key determinants in crop growth and hydrological modelling. They are commonly estimated from basic soil characteristics such as bulk density, organic carbon content and texture by means of pedotransfer functions (PTFs). In order to assess and compare the inherent performance and the functional applicability in the Zambezi River Basin (ZRB) of the widely used Saxton & Rawls PTFs and a set of newly developed PTFs, we compiled measurements of water retention at pF0.0, 1.0, 2.0, 2.8, 3.4 and 4.2 and of saturated hydraulic conductivity (Ksat) on 631 soil samples throughout the ZRB. A total of 329 of the samples were related to 55 soil profiles available in the Africa Soil Profile database, whereas our own field campaign carried out in a 2,426‐km 2 subbasin of the ZRB provided the remaining 302 samples related to 119 soil profiles. Apart from evaluating the Saxton & Rawls PTFs, we developed multiple linear regression (MLR) PTFs, and PTFs derived by three machine learning (ML) models: artificial neural network (ANN), random forest (RF) and support vector machine (SVM). All PTFs were first evaluated based on a comparison of the estimated and measured property values by means of R 2 , mean absolute error (MAE) and root mean squared error (RMSE). For the ensemble of MLR‐PTF and ML‐PTFs, the R 2 of the six water content variables and the Ksat ranged from 0.55 to 0.85, whereas for the Saxton & Rawls PTFs the range was between 0.10 and 0.50. Secondly, all PTFs were subjected to a functional evaluation using the Food and Agriculture Organization (FAO) AquaCrop crop growth model. Dry season irrigation requirements for maize as computed by AquaCrop with measured versus estimated soil hydraulic properties revealed that ANN‐PTFs provide AquaCrop outputs that come closest to AquaCrop outputs generated with measured soil hydraulic properties. This study shows the importance of performing functional evaluation of pedotransfer functions before their widespread application. Highlights Developed machine learning and multiple linear regression pedotransfer functions (PTFs). The Saxton & Rawls PTFs are not recommended for use in the Zambezi River Basin. PTFs were functionally evaluated through use of estimated soil hydraulic properties in AquaCrop. More accurate PTFs have better functional performance, although differences are small.

Saturated hydraulic conductivity (K sat) is an important soil parameter that governs water movement through horizons, pedons, and soil landscapes. K sat is infamous for its spatial and temporal variability, which contributes to the difficulty and considerable expense in measuring or otherwise quantifying it. Consequently, predictive methods such as pedotransfer functions (PTFs) that use physical soil properties, such as texture and bulk density, have been developed to derive K sat values. Soil texture and structure are key factors influencing K sat because of their direct relationship to pore size distribution. Quantitatively defining the combined effects of texture and structure on pore size distribution in a PTF is a difficult task. The objectives of this research were to: (i) estimate K sat based on pore characteristics derived from soil thin sections via image analysis; and (ii) compare the resultant values with field-measured K sat and with K sat estimated by a PTF using soil texture and bulk density parameters. We digitally scanned 39 thin sections from 11 pedons of soils derived from loess over till and/or over weathered sandstone. Soil voids were classified based on their size and shape. K sat was measured in the field using a Compact Constant-head Permeameter (Amoozemeter) and estimated using a Rosetta PTF. Simple and multiple linear regression (MLR) analyses were used to relate pore indexes and soil physical properties with measured and estimated K sat. The mean measured K sat was 0.74 cm h−1, whereas the PTF-estimated K sat from Rosetta and MLR were 0.36 cm h−1 and 0.49 cm h−1, respectively. The addition of pore characteristics into the model improved K sat predictions compared to predictions using Rosetta alone. The estimated K sat based on the model with added pore characteristics was better correlated with field-measured K sat (r = 0.82) than that based on Rosetta (r = 0.62). The addition of pore characteristics can improve K sat predictions. However, thin section void analysis from additional parent materials is needed.

A pedotransfer function for estimating bulk density of forest soil in Japan affected by volcanic ash

Comparison of estimated soil bulk density using proximal soil sensing and pedotransfer functions

Prediction of soil bulk density in agricultural soils using mid-infrared spectroscopy

Remotely sensed data from both in situ and satellite platforms in visible, near-infrared, and shortwave infrared (VNIR–SWIR, 400–2500 nm) regions have been widely used to characterize and model soil properties in a direct, cost-effective, and rapid manner at different scales. In this study, we assess the performance of machine-learning algorithms including random forest (RF), extreme gradient boosting machines (XGBoost), and support vector machines (SVM) to model salt marsh soil bulk density using multispectral remote-sensing data from the Landsat-7 Enhanced Thematic Mapper Plus (ETM+) platform. To our knowledge, use of remote-sensing data for estimating salt marsh soil bulk density at the vegetation rooting zone has not been investigated before. Our study reveals that blue (band 1; 450–520 nm) and NIR (band 4; 770–900 nm) bands of Landsat-7 ETM+ ranked as the most important spectral features for bulk density prediction by XGBoost and RF, respectively. According to XGBoost, band 1 and band 4 had relative importance of around 41% and 39%, respectively. We tested two soil bulk density classes in order to differentiate salt marshes in terms of their capability to support vegetation that grows in either low (0.032 to 0.752 g/cm3) or high (0.752 g/cm3 to 1.893 g/cm3) bulk density areas. XGBoost produced a higher classification accuracy (88%) compared to RF (87%) and SVM (86%), although discrepancies in accuracy between these models were small (<2%). XGBoost correctly classified 178 out of 186 soil samples labeled as low bulk density and 37 out of 62 soil samples labeled as high bulk density. We conclude that remote-sensing-based machine-learning models can be a valuable tool for ecologists and engineers to map the soil bulk density in wetlands to select suitable sites for effective restoration and successful re-establishment practices.

Saturated hydraulic conductivity (Ks), among other soil hydraulic properties, is important and necessary in water and mass transport models and irrigation and drainage studies. Although this property can be measured directly, its measurement is difficult and very variable in space and time. Thus pedotransfer functions (PTFs) provide an alternative way to predict the Ks from easily available soil data. This study was done to predict the Ks in Khuzestan province, southwest Iran. Three Intelligence models including (radial basis function neural networks (RBFNN), multi layer perceptron neural networks (MLPNN)), adaptive neuro-fuzzy inference system (ANFIS) and multiple-linear regression (MLR) to predict the Ks were used. Input variable included sand, silt, and clay percents and bulk density. The total of 175 soil samples was divided into two groups as 130 for the training and 45 for the testing of PTFs. The results indicated that ANFIS and RBFNN are effective methods for Ks prediction and have better accuracy compared with the MLPNN and MLR models. The correlation between predicted and measured Ks values using ANFIS was better than artificial neural network (ANN). Mean square error values for ANFIS, ANN, and MLR were 0.005, 0.02, and 0.17, respectively, which shows that ANFIS model is a powerful tool and has better performance than ANN and MLR in prediction of Ks.

Bulk density (BD) is a key soil property for sustainable land management and C stock assessment. However, data on BD are often missing in various soil survey reports in Central Africa because its measurement is laborious, time consuming at large scales, and may pose expense and facilities challenges, particularly in resource-poor countries. Therefore, there was a need for developing pedotransfer functions (PTFs) to predict BD for soils of Central Africa as an alternative solution. To do so, two approaches, namely multiple linear regression (MLR) and a pattern-recognition approach (k-nearest neighbor [k-NN]) were tested to predict BD for 196 soils of the Lower Congo. Based on their needs and familiarity with one or both of the two approaches, potential users interested in predicting the BD of strongly weathered soils of Lower Congo in particular or of Central Africa in general can immediately use the proposed equations or later apply the k-NN algorithm. Using an independent data set of low-activity clay soils from different tropical countries, nine individual PTFs and three ensemble PTFs were tested for their predictive capability. The results showed that despite the use of various modeling approaches—MLR, k-NN, or ensembles—BD remains a soil property that is difficult to predict with a satisfactory level of reliability because it can have substantial short-term variability in response to natural as well as humaninduced causes. We achieved a minimum root mean square difference value of 0.179 Mg m−3.

Evaluation of machine learning algorithms in the prediction of hydraulic conductivity and soil moisture at the Brazilian Savannah

&lt;p&gt;&lt;strong&gt;Background:&lt;/strong&gt; Soil hydraulic parameters in non-saturated conditions are crucial for explaining soil water dynamics in the field. It is therefore necessary to understand the link between soil water potential and hydraulic conductivity in the soil in order to estimate plant available water and hence simulate its movement within the soils. However, measurement of such hydraulic properties in the field and laboratory is somehow difficult, laborious and costly. &lt;strong&gt;Objective&lt;/strong&gt;: To determine soil hydraulic properties of Ferro-chromic Luvisols in Katumani using the RETC code based on pedo-transfer functions; % sand, silt, clay and soil bulk density. &lt;strong&gt;Methodology:&lt;/strong&gt; Undisturbed soil samples were collected from a profile pit at 0-15, 16-30, 31-45 and 46-60 cm depths using core rings for bulk densities and texture determination. Soil water retention curves and saturated hydraulic conductivities (Ksat) were estimated for all the samples using standard suction apparatus and the constant head method, respectively. The air entry suction (α) and pore size distribution (n) were generated using the RETC code. &lt;strong&gt;Results:&lt;/strong&gt; The permanent wilting point and field capacity were at 0.081, 0.102, 0.107 and 0.121 and 0.188, 0.225, 0.241, 0.262 m&lt;sup&gt;3&lt;/sup&gt;m&lt;sup&gt;-3 &lt;/sup&gt;H&lt;sub&gt;2&lt;/sub&gt;O, whilst its soil water diffusivity ranged from a low of 6.39, 6.94, 9.03 to a high of 12.5 cm&lt;sup&gt;2&lt;/sup&gt;min&lt;sup&gt;-1&lt;/sup&gt; in the 0-15, 16-30, 31-45 and 46-60 cm depth, respectively. Ksat values from RETC code ranged from 29 - 48 cm day&lt;sup&gt;-1&lt;/sup&gt;, while the total and readily available water within the soil profile were 330.4 and 214.7 mm H&lt;sub&gt;2&lt;/sub&gt;O, respectively. &lt;strong&gt;Implication:&lt;/strong&gt; The air entry value (α) and pore size distribution (n) implied an almost even distribution from the top and subsequent horizons. The soils ‘field capacity was achieved at pF 2.0 whilst PWP was arbitrary indicated at pF 4.2 reducing the time it takes to calculate irrigation cycles based on the amount of water available to the crops. &lt;strong&gt;Conclusion:&lt;/strong&gt; The data indicates that pedo-transfer functions; especially high bulk densities negatively impact on soil hydraulics conductivity.&lt;/p&gt;

Core Ideas K s was estimated using MLR‐type PTFs, ANN‐type PTFs and state‐space analysis. State‐space modeling was scale‐sensitive in estimating K s in Loess Plateau. Spatial correlations revealed in state‐space analyses were consistent with wavelet coherency. Bulk density, clay content and topography dominated K s spatial distribution. A precise description of saturated hydraulic conductivity ( K s ) and its spatial variability is required for modeling soil and water transport in the vadose zone. Nevertheless, the direct measurement of K s is expensive and laborious especially for large domains crossing hundreds of kilometers. The objective was to estimate K s from easily accessible soil properties and environmental factors using pedotransfer functions (PTFs) and state‐space analysis. Along an 860‐km south–north transect in the Loess Plateau of China, soil cores for K s measurements were collected at depths of 0 to 10, 10 to 20, and 20 to 40 cm at 10‐km intervals from 15 Apr. to 15 May 2013. Multiple linear regression (MLR) and artificial neural network (ANN) were used to derive PTFs for K s estimation. Based on the eight factors of bulk density, soil organic carbon, sand content, clay content, mean annual precipitation and temperature, slope gradient and elevation, the state‐space analysis appeared to outperform the PTFs in calibrating K s over the entire transect. The adjusted coefficients of determination ( R 2 adj ) for the state‐space models were all greater than 0.9, whereas the corresponding R 2 adj were much lower for the MLR‐ and ANN‐type PTFs (ranging from 0.398 to 0.880). However, the state‐space approach is quite scale‐sensitive, and overfitting occurred when it was cross‐validated with a leave‐one‐out procedure. It performed almost perfectly in calibration as implied in the R 2 adj of ∼1 but rather poorly in validation with R 2 adj typically &gt;0.4. The ANN method exhibited the best K s estimations at all depths. Both wavelet coherency and state‐space modeling quantified the spatial correlations of K s with the eight factors investigated and manifested consistent results, that is, bulk density, clay content, and topography were the primary properties controlling K s distribution. These findings are critical for hydrological modeling and irrigation management in the Loess Plateau of China and possibly other arid and semi‐arid regions.

Small-scale spatial variability of hydro-physical properties of natural and degraded peat soils

Context High-mountain páramos provide essential hydrological services across the Andes. However, green onion cultivation in these ecosystems has led to nitrogen (N) losses, compromising water quality. Aims This study aimed to improve understanding of soil N dynamics in páramo agroecosystems by analyzing the vertical and temporal distribution of total and mineral N in páramo soils, and by developing pedotransfer functions (PTFs) to predict these pools. Method From two cultivated plots and one native grassland plot, 288 soil samples were collected over 7 and 4 months, respectively, at three depths. Samples were analyzed for soil N concentrations, mineral N stocks, and 16 edaphic properties. Using random forest and multiple linear regression (MLR), PTFs were developed incorporating edaphic and remote-sensing covariates. Key results Green onion cultivation altered mineral N distribution, with subsurface nitrate-N reaching 114 mg kg−1, compared to 1.76 mg kg−1 in native grassland. Monthly declines in mineral N stocks reached 75.8 kg N ha−1 in cultivated soils, nearly triple the losses in uncultivated soil. The MLR using edaphic properties performed best for total N and nitrate-N prediction (R2 = 0.80 and 0.64, respectively), while ammonium-N was poorly predicted. Conclusions Shifts in N dynamics were consistent with soil acidity neutralization associated with repeated chicken manure applications, fallow history, and soil texture. These conditions favored subsoil mineral N accumulation, increasing leaching risk. The developed PTFs demonstrate potential to inform site-specific assessments of soil N pools in páramos. Implications Future research should include additional N pools, microbial processes, and isotopic tracers to deepen understanding of N dynamics.