Geostatistical based Digital Soil Mapping

Geostatistical based Digital soil mapping presentation for the Ph.D. Research work Doctoral seminar given by Research Scholar Dhiraj M Kadam from VNMKV Parbhani

#case study #digital soil mapping #doctoral seminar #dsm #geostatistical based dsm #geostatistics #interpolation methods #kriging technique #pdf #pedometrics #ph.d. #ppt #research #soil science #statistical analysis

1. Geostatistical Based Digital Soil Mapping
Research Guide & Seminar In charge
Dr. P. H. Vaidya
Head & Professor
Department of Soil Science and Agricultural Chemistry
College of Agriculture, VNMKV, Parbhani
Presented by
Mr. Kadam Dhiraj Madhav
(Reg. No. 2022A/29P)
Ph.D. III
rd
Semester

Course No : SOILS 692
(Doctoral Seminar 1)

2. Framework
Introductory Geostatistics : Roles
Introductory Digital soil Mapping
Tradiotional vs DSM Mapping
Framework of Geostatistical based DSM
Semivariogram
Interpolation
Interpolation methods : Deterministic and Geostatistical methods : IDW, Kriging,
Terminologies related to geostatistics etc.
Case Study 1
Case Study 2
Case Study 3
Case Study 4
Case Study 5
Case Study 6
Case Study 7
Case Study 8
Conclusion

5. Geostatistical based digital soil mapping is a sophisticated methodology that harnesses the power of
spatial statistics to generate high-resolution soil maps. By integrating field observations, laboratory analyses, and
environmental covariates, this approach offers a quantitative framework for understanding and predicting soil
spatial variability. Core to this approach is geostatistics, which models spatial dependence through techniques
such as kriging and co-kriging. These methods enable the interpolation of soil properties at unsampled locations
by considering the spatial correlation between data points and their relationship with environmental factors.
Variogram analysis, a fundamental component, characterizes the spatial structure of soil attributes, informing the
selection of appropriate interpolation techniques.
Digital soil maps derived from geostatistical modeling provide valuable information for a wide range of
applications, including precision agriculture, environmental monitoring, and land use planning. By delineating soil
patterns, quantifying soil heterogeneity, and assessing soil quality, these maps support informed decision-making
and sustainable resource management.
However, the accuracy and reliability of geostatistical soil mapping are contingent upon several factors.
The quality and density of input data, the appropriateness of geostatistical models, and the incorporation of
uncertainty assessments are critical for generating robust and defensible soil maps.
Recent advancements in remote sensing, machine learning, and computational power have expanded
the capabilities of geostatistical soil mapping. Integration of these technologies holds promise for improving
prediction accuracy, enhancing spatial resolution, and providing more comprehensive soil information. As a
result, geostatistical based digital soil mapping is poised to become an increasingly indispensable tool for
characterizing and managing soil resources in the face of global challenges.

6. Geostatistical analysis plays a crucial role in digital soil mapping by:

1. Predicting soil properties: Geostatistical techniques like kriging and inverse distance weighting
(IDW) help predict soil properties at unsampled locations, creating continuous maps.

2. Quantifying uncertainty: Geostatistics provides uncertainty estimates for soil property predictions,
essential for understanding map reliability.

3. Identifying spatial patterns: Geostatistical analysis reveals spatial autocorrelation and patterns in
soil properties, informing mapping and sampling strategies.

4. Optimizing sampling design: Geostatistics helps design efficient sampling schemes, minimizing the
number of samples needed to achieve accurate maps.

5. Integrating multiple data sources: Geostatistical methods combine diverse data types (e.g., soil
surveys, remote sensing, laboratory analysis) to produce comprehensive soil maps.

6. Accounting for spatial variability: Geostatistics considers the spatial heterogeneity of soil
properties, ensuring maps reflect local conditions.

7. Enhancing map accuracy: By incorporating geostatistical analysis, digital soil maps become more
accurate and reliable, supporting informed decision-making. As like Variogram Analysis is The process
of modeling the spatial structure of a variable through the calculation and analysis of the variogram,
which describes the spatial correlation as a function of distance.

8. The General Principle-DSM
Jenny (1941) proposed that soil development is a function of climate, organisms,
topography, parent material and time; this hypothesis is the basic assumption of
DSM.
McBratney et al., (2003) introduced further variables space (spatial position) and
soil information derived from other investigations, as soil can be predicted from
its own properties in the so-called ―SCORPAN model. The SCORPAN approach is
expressed as by the equation:

17. Framework of Geostatistical DSM

Validation and Accuracy Assessment
Validating the accuracy and reliability of geostatistical digital soil maps is a crucial step in the mapping
process. This phase ensures that the final product meets the intended goals and provides trustworthy
information for decision-making.

Accuracy assessment
typically involves comparing the predicted soil properties from the mapping exercise against independent
field observations or laboratory measurements not used in the original model development.

Cross-validation techniques,
such as leave-one-out or k-fold cross-validation, are used to evaluate the model's performance and identify
potential biases or areas of high uncertainty.

Statistical metrics, including root mean square error (RMSE), coefficient of determination (R²), and mean
absolute error (MAE), are calculated to quantify the goodness of fit between predicted and observed
values.

Visual diagnostic plots,
such as scatter plots and residual plots, help identify spatial patterns in the model's performance and guide
further improvements to the mapping approach. The validation process not only assesses the accuracy of
the final soil maps but also provides valuable feedback for refining the geostatistical modeling techniques,
data collection strategies, and preprocessing methods. This iterative approach ensures the continuous
improvement of the digital soil mapping framework.

19. Semivariogram

The semivariogram, also referred to as a variogram or semivariance function, is a fundamental tool used in geostatistics To
analyze spatial variability and quantify spatial autocorrelation within a dataset. This statistical measure illustrates how data points vary
concerning their spatial separation or lag distance. Essentially, the semivariogram reveals how the similarity of data values changes with
distance.
To calculate the semivariogram, one employs semivariance, which is half the average squared difference between data points
within a given lag distance. By doing so, it quantifies the level of similarity or dissimilarity between data points at a specific distance apart.
When the lag distance is small, the semivariance tends to be low since nearby points exhibit higher similarity. However, as
the lag distance increases, the semivariance may increase up to a certain point, representing the spatial autocorrelation range or
"nugget." Beyond the nugget, the semivariance may reach a plateau, indicating that the spatial dependence has reached its maximum.

The semivariogram typically includes:
1. Nugget effect (small-scale variability, random spatially independent noise )
2. Sill (maximum variability)
3. Range (distance beyond which data points are no longer correlated, after the slope)
4. Spatial autocorrelation (degree of correlation between data points where it forms a slope)

By examining the semivariogram, we can:
- Identify the spatial relationships in the data
- Choose the appropriate kriging technique
- Select the optimal parameters for kriging

20. Key Differences

- Shape and Slope:
- Spherical: Has a linear increase followed by a sharp leveling off at the range.
- Exponential: Shows a rapid increase initially that gradually slows down, approaching
the sill asymptotically.
- Gaussian: Exhibits a slow, smooth increase toward the sill, indicating stronger spatial
continuity.

- Range Interpretation:
- Spherical: Has a clear range after which the correlation is negligible.
- Exponential and Gaussian: Use an effective range concept due to their asymptotic
approach to the sill.

- Spatial Continuity Representation:
- Spherical: Indicates moderate spatial continuity.
- Exponential: Suitable for data with moderate to long-range spatial dependence.
- Gaussian: Best for data with strong spatial continuity and long-range dependence.

25. *Structural Factors : Mean, trend, covariates etc. *Random Factors : Variability, Noise, Residuals, uncertainty

29. The deterministic way

A deterministic method is an algorithm or
approach that consistently generates the
same output for a given input, regardless
of the number of times it is executed. It
operates without any randomness or
uncertainty, resulting in a completely
predictable and consistent outcome.
Their key characteristics include
repeatability, predictability, the elimination
of uncertainty, and overall consistency. By
offering reproducibility and stability,
deterministic methods ensure reliable and
accurate results in various applications,
ranging from simulations to critical
decision making processes.
The geostatistical way

Geostatistical methods comprise a set of
statistical techniques specifically designed to
analyze and model spatially correlated data.
Geostatistics takes into account the spatial
dependence or autocorrelation that may
exist in the data, enabling more accurate
predictions and interpolation of values at
unsampled locations. The primary objectives
of geostatistical methods are to create spatial
maps, identify spatial patterns, estimate
values at unsampled locations, and quantify
uncertainty in Predictions.

32. Inverse distance weighting (IDW)
Inverse distance weighting (IDW) is a widely used
interpolation technique in spatial analysis and
geostatistics for estimating values at unsampled
locations based on nearby sampled data points. The
fundamental assumption of IDW is that values at
unsampled locations are influenced more by the
values of nearby points than those farther away. To
achieve this, the method employs a power parameter
"p" that controls the influence of nearby points on the
estimation. Typically, "p" is set between 1 and 3, with
lower values giving more weight to points closer to the
target location and higher values providing more equal
weight to all points.

It does have some limitations.
One such limitation is its sensitivity to the choice of the
power parameter, which can affect the interpolation
results significantly. Additionally, IDW tends to produce
"bull'seye“ artifacts around data points, particularly
when the data is sparse or unevenly distributed. As a
result, IDW is often utilized for basic interpolation tasks
and serves as a baseline for more sophisticated
interpolation methods in GIS and spatial data analysis.

34. Splines

Concept:
- Splines are a deterministic interpolation method that fits smooth surfaces (or curves in one dimension) through the
known data points. The goal is to create a smooth function that minimizes the overall bending (curvature) of the surface.

Key Features:
- Smoothness: Splines are designed to create smooth surfaces with minimal curvature, which makes them suitable for
applications requiring smooth transitions between data points.
- Types of Splines: Includes cubic splines, thin-plate splines, and B-splines, each with different mathematical formulations
and properties.

Disadvantages:
-Can produce unrealistic results when data points are irregularly spaced or when there is significant noise in the
data.
- Does not inherently provide measures of prediction uncertainty.

36. Kriging
Kriging is a powerful geostatistical interpolation method that delivers the best
linear unbiased estimate (BLUE) of a variable at unsampled locations . It incorporates
both spatial correlation and uncertainty in the data, making it a robust and reliable
interpolation technique. The fundamental principle underlying is to minimize prediction
error by assigning appropriate weights to neighboring data points based on their spatial
distance and correlation. The Kriging method assumes that the spatial correlation in the
data can be modeled using a variogram (or semivariogram). This variogram describes how
the variance of the variable changes with the distance between data points. By using the
variogram model, Kriging can provide a continuous and spatially smooth surface,
allowing for accurate estimation at unsampled locations.
Co-Kriging
Co-Kriging is an extension of the traditional Kriging method, Co-Kriging becomes
valuable in situations where two or more variables exhibit spatial relationships, and it
can provide more precise predictions compared to using Kriging independently,
especially when data for one variable is sparse or missing. Both Kriging and Co-Kriging are
powerful tools for spatial interpolation, enabling the estimation of values at unsampled
locations while taking into account spatial correlation and uncertainty. The choice
between Kriging and Co-Kriging depends on the characteristics of the data and the
presence of multiple correlated variables. When multiple correlated variables are
available
The Heart of Geostatistics…..

37. Empirical Bayesian kriging (EBK) or Universal Kriging
Empirical Bayesian kriging (EBK) is an advanced geostatistical interpolation
method that combines the principles of kriging and Bayesian statistics to estimate values
at unsampled locations. In traditional kriging, the variogram model, which measures
spatial correlation, is assumed to be known or directly estimated from the data. However,
in empirical Bayesian kriging, the variogram model parameters are treated as random
variables and estimated using additional data called the "drift" or "external drift" data.
Drifts : RS data, DEM, Climate data etc.

Simple Kriging
The estimation of simple kriging (SK) is based on a slightly modified where μ is
a known stationary mean. The parameter μ is assumed constant over the whole domain
and calculated as the average of the data (Wackernagel 2003). SK is used to estimate
residuals from this reference value μ given a priority and is therefore sometimes referred
to as “kriging with known mean” (Wackernagel 2003).

38. Regression Kriging (RK)

The Regression kriging method combines regression and kriging by treating these as two
separate, consecutive steps. At first part linear regression is applied to data. Regression coefficients and
residuals are obtained. Regression model is obtained by multiplying the regression coefficients by each
grid value of data which the predictions will be calculated on that surface. Next, a kriging step is done in
which the regression residual is no longer treated as uncorrelated but allowed to be spatially correlated.
When regression residuals are spatially correlated, the map may become more accurate by interpolating
these residuals and adding them to the predicted values from the regression model. At this step, simple
kriging is applied to the residuals (i.e., the differences between the observations and the predicted values
with linear regression). Simple kriging is used instead of ordinary kriging because it can be assumed that the
residual has a known mean (namely zero). Finally, the kriged residual is added to the regression model
result. This method can thus be seen as an extension of Multiple Linear Regression (MLR) because by
adding residual kriging to regression one has the ability to include additional information and gain more
accurate predictions. This constitutes the Regression kriging (RK) The RK prediction formula is given by
Equation :

40. From the Window :

43. Cross Validation or Error estimation

To identify the most suitable model for a particular soil property, the selection process involves minimizing
the error and maximizing the model's efficiency known as the error calculation or cross-validation. The correctness of
the spatial model is checked with the error percentages. The "error percentage" provides a quantitative representation
of the difference between predicted and observed values at unsampled locations. This broadly incorporates mean
absolute error (MAE), root mean square Error (RMSE), and mean squared prediction error (MSPE).

RMSE (Root Mean Square Error) is a measure of the average magnitude of the errors in a model's predictions. A smaller
RMSE value indicates better model performance i.e. RMSE values close to 0 indicate excellent model performance, with
predictions very close to the actual values.

R2 (Coefficient of Determination) is a measure of how well a model explains the variability in the data. R2 values range
from 0 to 1, and the closer to 1, the better the model fits the data.


In cross-validation, the available data is split into two parts:

1.Training set / Calibration Set : Used to build the model and understand relation
2.Validation set (or testing set): Used to evaluate the model's accuracy performance

Common cross-validation techniques in geostatistics include:
1. K-fold cross-validation (e.g., 5-fold or 10-fold)
2. Leave-one-out cross-validation (each sample is used once as the validation set)
3. Block cross-validation (used for spatial data, dividing the study area into blocks)

where n is the number of samples, P is the predicted value, and O is the observed value.

44. 2. Data exploration tools
These are the tools used to explore or understand the dataset in detail. On the basis of their utility and properties
they can be further subdivided as histogram, normal Q-Q plot, voronoi maps, trend analysis, semivariogram, general
Q-Q plot and cross variance cloud etc.
Histogram
A histogram is a graphical representation of a
dataset's distribution, offering a visual means
to comprehend the underlying frequency or
probability distribution of numerical data. The
dataset is divided into intervals or bins, and
the height of each bar in the histogram
corresponds to the frequency or count of
observations falling within that bin.
Histograms prove invaluable in identifying
patterns, understanding central tendencies
and data spread, detecting outliers, and
visualizing the overall shape of the
distribution.
A normal quantile-quantile (Q-Q)
plot
A quantile-quantile plot, is a graphical
tool used to whether a dataset
adheres to a normal distribution. It is
achieved by comparing the quantiles
of the dataset against the quantiles of
a theoretical normal distribution.
When the points on the Q-Q plot
closely align along a straight line, it
indicates that the data is
approximately normally distributed.,
Q-Q plots provide a powerful tool for
evaluating the conformity of a dataset
to a normal distribution and play a
vital role in statistical analyses.
The N:S (Nugget : Sill) ratio

provides insight into the degree of spatial dependence of a
soil parameter. Different N:S value ranges, such as <0.25,
0.25-0.75, and >0.75, indicate strong, moderate, and weak
spatial variability or dependance of a particular soil
parameter, respectively. Estimating variogram parameters
(sill, range, and nugget effect) involves fitting various
theoretical models to the experimental variogram.

46. Mapping Soil Properties at a Regional Scale: Assessing
Deterministic vs. Geostatistical Interpolation Methods at
Different Soil Depths
Case Study 1
Location: Autonomous Region of
Extremadura, located in the southwest of
Spain
L.C. Joaquín Francisco et al., (2022) Sustainability, 14,
10049 PN: 01-20

47. Autonomous Region of Extremadura, located in the southwest of Spain

49. Inverse Distance Weighting (IDW), Completely Regularized Spline (CRS), Spline With Tension (SWT), Multiquadric (M-Q)
and Inverse Multiquadric (IM-Q).Local Polynomial Interpolation (LPI), Ordinary Kriging (OK), Simple Kriging (SK),
Empirical Bayesian Kriging (EBK)

55. Conclusion :

In this study, nine interpolation methods were used to predict 12 soil variables
that were measured at three different soil depth intervals. Statistics such as mean error,
coefficient of determination and mean square error were used to evaluate the accuracy of
the methods. Although a general preference to use geostatistical methods is observed in
general, we conclude that deterministic methods provide better results than geostatistical
ones. Our results show that geostatistical methods were more accurate in 19 of the 36 case
studies. However, the observed difference between the interpolation techniques is
negligible in some cases, allowing different ones to be used interchangeably. In this regard,
our results indicate that the accuracy of the methods varies depending on the case study.
Results also varied, in general, when the different depths are considered, identifying
deterministic methods as more accurate for the topsoil and geostatistical ones for the
deeper layer. Therefore, we also conclude the necessity to use a variety of soil mapping
methods and techniques to achieve the best results.

56. Characterizing spatial variability of soil properties in salt
affected coastal India using geostatistics and kriging
Case Study 2
Location: : five
coastal blocks Bhograi, Baliapal,
Balwswar, Remuna, and
Bahanaga in Balasore and Basudebpur,
Chandbali blocks in
Bhadrak districts of Odisha, Tripathi R., et al. (2015) Saudi Society for Geosciences
A total of 132 soil samples were collected.

58. It was observed that ordinary kriging
could successfully interpolate pH,
ECe, and SOC values as evident from
the values of ME, RMSE, ASE, MSE,
and RMMSE, whereas available soil
N, P, and K were specified with
maximum error indicators. The
model with the lowest RMMSE was
chosen as optimal for each soil
property

60. A large range indicates that the observed values of a soil variable are influenced by other values for this variable over greater
distances than soil variables, which have smaller ranges (Lopez-Granados et al. 2002). A range of 2220.10 m for soil organic
carbon (least range among six soil properties) indicates that these variable values influenced neighboring values over smaller
distances compared to other soil variables.

The nugget effect is related to spatial variability over distances shorter than the lowest separation distance between
measurements (Cemek et al. 2007). Meanwhile, the large nugget effect means that an additional sampling of these properties
at smaller distances and in larger numbers might be needed to detect spatial dependence, and a greater sampling density will
result in a more accurate map.

62. Spatial map developed by kriging for a
a soil pH,
b soil electrical conductivity (ECe),
c soil organic carbon (SOC),
d available soil nitrogen,
e available soil phosphorus, and
f available soil potassium

63. Conclusion

Semivariogram models were fitted for six soil properties, i.e., soil pH, soil
electrical conductivity, soil organic carbon, available soil nitrogen, available soil
phosphorus, and available soil potassium, and the best variogram model for each soil
property was identified using cross-validation approach. Cross validation of variogram
models through ordinary kriging showed that spatial prediction of soil properties is
better than assuming the mean of the observed values at any unsampled location.
Finally, maps for above six properties were developed using best fit semivariogram
models and ordinary kriging. This study confirms that a geostatistical approach can be
applied for assessing high-risk areas of soil salinity and nutrient deficiency which
requires some immediate remedial treatment and an effective management plan
providing an informed decision support to help farmers. Traditional techniques do not
provide any measurement of the reliability of the estimates; thus, no risk assessment
can be made. Critical concentrations of electrical conductivity or other soil properties
may be estimated and farmers may be advised to act, and an appropriate and
effective soil and crop management plan can be developed based on these estimates.

64. Prediction of Soil Depth in Karnataka Using Digital Soil
Mapping Approach
Case Study 3
Location: : Karnataka
Dharumarajan, S., et al. (2020) Journal of the Indian
Society of Remote Sensing

65. Sampling Methodology

Normally the soil profiles were opened up to 200 cm or to the depth limited by rock
or hard substratum to measure the soil depth. The major limitation in organizing the soil profile
database is unavailability of exact geolocations for the samples collected under SRM project.
The coordinates were assigned for those profiles based on district name and village name and
soil morphological properties (e.g. soil colour) mentioned in the profile sheet and also by
matching with satellite imagery. A total of 5174 soil profiles were used for this study after
removing the outliers.

66. Different covariates used for prediction of soil depth
Statistical results of soil depth of Karnataka

67. Concordance correlation coefficient (CCC) :
Value close to 1 indicate perfect agreement
between predicted and observed values

Percentage data point within prediction
interval (PICP) :
value close to 100% show
Reliability of prediction interval

68. Statistical results of predicted soil depth of Karnataka
Importance of variables for prediction of soil depth in Karnataka

69. Conclusion :

We developed a baseline data for soil depth at 250 m resolution over Karnataka using
digital soil mapping approach. The RK algorithm performed better than QRF for
prediction of soil depth with R2 and RMSE of 30% and 34 cm, respectively. The error in
locating coordinates for decade-old datasets might also be the reason for lower accuracy
rate. Nevertheless, the predicted soil depth map can be used for preparation of proper
land use planning and other hydrological studies. The data augmentation and removal
of incorrect profile points helps in improving the model accuracy. The spatial
resolution of the maps acts as a baseline data and helps in monitoring the soil health
in the future.

70. Digital soil mapping in the Bara district of
Nepal using kriging tool in ArcGIS
Case Study 4
Location: : Bara district in Nepal Panday, D., et al.(2018), PLoS ONE 13(10): e0206350.

72. SD = standard deviation CV = coefficient of variation, Min = minimum, Max = maximum, skew = skewness. Skew (O) and Kurtosis (O) =
skewness and kurtosis obtained from original data. Skew (T) and Kurtosis (T) = skewness and kurtosis obtained from log transformed
data. Similar units for Mean, SD, Minimum, Maximum, Skew and Kurtosis, but % for CV.

80. According to ESRI, for the theory of random patterns, when the p-value is very small (in this study, p < 0.05) and the z-score is either very high or very
low (1.96 < z and z < −1.96), the spatial pattern is not likely to reflect a random form of distribution. A positive Moran’s I index value indicates the
neighboring values are similar, suggesting spatial dependency. A negative Moran’s I index value indicates the neighboring values are dissimilar,
suggesting inverse spatial dependence. A Moran’s I index value of zero implies a lack of spatial pattern .

81. Conclusion :

The descriptive statistics showed that most of the measured soil chemical variables
were skewed and non-normally distributed and the available K2O data were highly variable (5
to 696 kg ha-1). Geostatistical interpolation identified that exponential, spherical, or Gaussian
models provided the best fit to the semivariograms, depending on the soil chemical variable
and, in general, showed weak or moderate spatial dependency for all of the variables. The
kriging maps of soil chemical properties were found effective in explaining the distribution of
soil properties in non-sampled locations based on sampled data. These maps aid farmers in to
making efficient management decisions based on their proper understanding of the
conditions of existing farm soils. These results show geostatistical analysis using kriging is an
effective prediction tool for exploring the spatial variability of soil nutrients, and we recommend
this tool for future soil sampling campaigns in Nepal.

82. Evaluating the Spatial Variability of Soil Physicochemical
Characteristics in an Indian Lesser-Himalayan Region
Case Study 5
Location: : Baramulla district, situated
within the esteemed union territory of
Jammu and Kashmir

Javed et al., Biological Forum – An International
Journal 15(11): 298-305(2023)

84. The spatial dependencies exhibited notable variability, with a range spanning from 1036.7 metres for available nitrogen to
14973 metres for available phosphorus (P2O5). This suggests that the ideal sampling interval can differ significantly
depending on the specific soil properties being assessed. The determination of range values allows for the assessment of the
correlation between various sampling locations, as well as the identification of the maximum distance of spatial dependence
between them. The obtained outcomes can be utilised to formulate suggestions concerning optimal agricultural practices
and the development of soil-plant interaction models for forthcoming research endeavours.

88. Conclusion :

The geostatistical interpolation technique effectively determined that the
exponential, spherical, and Gaussian models exhibited optimal conformity with the
semivariograms, contingent upon the specific soil chemical variable. In a broader sense, these
models demonstrated a relatively weak to moderate degree of spatial dependency across all
variables. The utilisation of kriging maps for soil chemical properties has proven to be highly
effective in elucidating the spatial distribution patterns of soil properties in areas where no
samples were taken, solely relying on the available sampled data. The assessment of spatial
heterogeneity in soil physical and chemical attributes is an essential step in implementing
targeted soil and crop management strategies. The soil property maps, along with their
corresponding spatial structures, have successfully delineated the priority management zones
that should be addressed in the future to enhance soil quality. These maps can also be utilised
to develop more effective sampling designs for making informed management decisions.

89. Digital Mapping of Soil Properties in the Western-Facing
Slope of Jabal Al-Arab at Suwaydaa Governorate, Syria
Case Study 6
Location: : western region of Suwaydaa
governorate in southern Syria
Khallouf, A., et al. (2020) Jordan Journal of Earth and
Environmental Sciences. All rights reserved, 11 (3): 193-
201, ISSN 1995-6681

96. Conclusion

The application of geostatistical approach involving descriptive statistics and semivariogram
analysis improved the description of spatial variability for soil chemical properties at 0 to 30 cm deep.
The descriptive statistics showed that most of measured soil variables were skewed and abnormally
distributed, and the available K2O data were highly variable (338 to 595 mg/kg). Geostatistical
interpolation identified that exponential, spherical or Gaussian models provided the best fit to
semivariograms, depending on the soil chemical variable, showing in general strong, moderate or weak
spatial dependency for all variables. Kriging maps of soil variables were found effective in interpreting
the distribution of soil properties in non-sampled locations based on sampled data. These maps help
farmers in making efficient management decisions based on their proper understanding of the
conditions of existing farm soils. These results show that Kriging-geostatistical analysis is an effective
prediction tool for exploring the spatial variability of soil nutrients. Generally speaking, this tool is
recommended for future soil sampling campaigns in Syria.

97. Improving the Prediction Auccuracy of Soil Mapping
through Geostatistics
Case Study 7
Location: : Ismailia province, Egypt
Omran E.E. (2012), International Journal of
Geosciences, 3, 574-590
The study area was covered by 146 sampling sites. The aver-age distance between soil
sampling locations is approximately 50 m.

99. Histograms and QQPlot for soil properties (EC, pH, CaCO3).

103. Conclusion :

This study showed that the UK method is more accurate for predicting the spatial
patterns of soil (EC, pH, CaCO3) properties than the other methods, namely OK, IDW and Splines.
It is also the most accurate method when we avoid the outlier effects in assessing the
performance of all methods. The generally superior performance of UK is due to less prediction
errors. UK obtains the soil property surfaces without oscillation problems (Unnatural fluctuations
give unrealistic prediction) and Bull’s eyes. Therefore, UK can be considered as an accurate
method for interpolating soil (EC, pH, CaCO3) properties. These results may help GIS specialists to
select the best method for the generation of soil properties. A technique should be chosen not
only for its performance on a specific soil type and data density, but also for its applicability to a
wide range of spatial scales.

104. Analysis and prediction of soil properties using local
regression-kriging
Case Study 8
Location: : Ismailia province, Egypt Sun, W. Et al. (2012) Geoderma 171-172, 16–23

109. Conculsion :

1. We have developed a local RK approach which is implemented in an easy to use software
programme.

2. Local RK is an efficient tool for helping to investigate the variances and relationships of the
variables.

3. The performance of local RK in the examples presented in this paper appears superior to others
in general. However the performance seems to be highly dependent on the local environment and
the quality of data. Further investigation and numerous datasets are required.

110. Conclusion

Geostatistics has emerged as an indispensable and versatile tool for understanding spatial relationships
and making accurate predictions in various scientific fields. In the context of soil parameter analysis,
geostatistics provides a powerful means to comprehend the spatial variability of soil properties. By
identifying spatial patterns and quantifying uncertainties, geostatistics facilitates informed decision-
making for sustainable land management and environmental applications. The key to harnessing the
full potential of geostatistics lies in careful data collection, preprocessing, and validation. Ensuring the
accuracy and reliability of results requires diligent attention to data quality and representative sampling
techniques. By capitalizing on the spatial autocorrelation inherent in natural phenomena, geostatistical
methods offer valuable insights that enable effective management of natural resources, environmental
protection, and informed decision-making across a broad spectrum of applications. The integration
of geostatistics into digital soil mapping provides a powerful approach to understanding and
managing soil resources. incorporating auxiliary variables, and providing uncertainty estimates,
geostatistics enhances the accuracy and reliability of soil maps, supporting sustainable land
management and agricultural productivity. As technologies and methodologies advance, the role of
geostatistics in DSM is likely to grow, offering even greater precision and insights into soil variability
and dynamics.