Content uploaded by Hewa T. Mustafa
Author content
All content in this area was uploaded by Hewa T. Mustafa on Oct 30, 2024
Content may be subject to copyright.
Content uploaded by Hewa T. Mustafa
Author content
All content in this area was uploaded by Hewa T. Mustafa on Oct 30, 2024
Content may be subject to copyright.
Content uploaded by Diman Krwanji
Author content
All content in this area was uploaded by Diman Krwanji on Oct 29, 2024
Content may be subject to copyright.
Content uploaded by Ismael Omar
Author content
All content in this area was uploaded by Ismael Omar on Oct 30, 2024
Content may be subject to copyright.
Content uploaded by Aram Ali
Author content
All content in this area was uploaded by Aram Ali on Oct 30, 2024
Content may be subject to copyright.
_____________________________________________________________________________________________________
*Corresponding author: E-mail: aram.ali@unisq.edu.au, aram.ali@su.edu.krd;
Cite as: Ali, Aram, Ismael O. Ismael, Hewa T. Mustafa, Diman Krwanji, and Akram O. Esmail. 2024. “Enhancing Soil Texture
and Bulk Density Mapping Using Soil Grids and Machine Learning: A Comparative Analysis With Observed Data”. Asian Soil
Research Journal 8 (4):61-78. https://doi.org/10.9734/asrj/2024/v8i4163.
Asian Soil Research Journal
Volume 8, Issue 4, Page 61-78, 2024; Article no.ASRJ.125367
ISSN: 2582-3973
Enhancing Soil Texture and Bulk
Density Mapping Using Soil Grids and
Machine Learning: A Comparative
Analysis with Observed Data
Aram Ali a,b*, Ismael O. Ismael a, Hewa T. Mustafa a,
Diman Krwanji c,d and Akram O. Esmail a
a Soil and Water Department, College of Agricultural Engineering Sciences, Salahaddin University-
Erbil, Erbil, Kurdistan Region, Iraq.
b University of Southern Queensland, Centre for Sustainable Agricultural Systems, West St,
Toowoomba, QLD 4350, Australia.
c Plant Protection Department, College of Agricultural Engineering Sciences, Salahaddin University-
Erbil, Erbil, Kurdistan Region, Iraq.
d University of Southern Queensland, Centre for Crop Health, West St, Toowoomba, QLD 4350,
Australia.
Authors’ contributions
This work was carried out in collaboration among all authors. All authors read and approved the final
manuscript.
Article Information
DOI: https://doi.org/10.9734/asrj/2024/v8i4163
Open Peer Review History:
This journal follows the Advanced Open Peer Review policy. Identity of the Reviewers, Editor(s) and additional Reviewers, peer
review comments, different versions of the manuscript, comments of the editors, etc are available here:
https://www.sdiarticle5.com/review-history/125367
Received: 20/08/2024
Accepted: 22/10/2024
Published: 29/10/2024
Original Research Article
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
62
ABSTRACT
Digital soil mapping plays a crucial role in understanding soil variability and informing sustainable
land management practices. This study focuses on the Kurdistan Region of Iraq (KRI), evaluating
the accuracy of SoilGrids, a global-scale soil mapping initiative, and exploring the efficacy of
machine learning algorithms in refining soil properties estimations. The aim of this research was to
assess and represent the physical parameters of soils effectively by comparing ground truth soil
sampling data with data obtained from SoilGrids regarding clay, silt, and sand fractions and bulk
density. Comparative analyses were conducted between ground truth soil sampling data and
SoilGrids predictions, revealing significant differences across soil mineral fractions including clay,
silt, sand fractions, and bulk density. The results showed that the mean clay fraction in the ground
truth dataset differed notably from SoilGrids estimation, with a Mean Absolute Deviation (MAD) of
124.0 g kg-1 and Root Mean Square Error (RMSE) of 152.5. However, the integration of machine
learning algorithms, particularly the Extreme Gradient Boosting (XG Boost) algorithm, showed
promising results in improving accuracy. The XG Boost algorithm exhibited a relatively low MAD of
97.9 g kg-1 for clay fractions, indicating a better approximation of observed values compared to
SoilGrids. Significant percent improvements in RMSE and Mean Absolute Percentage Error
(MAPE) values were observed across soil fractions and bulk density measurements, ranging from
approximately 15% for clay to 35% for sand fractions and 20% for bulk density. These findings
highlight the importance of integrating advanced mapping techniques and machine learning
algorithms to enhance soil mapping methodologies. Moving forward, efforts to expand ground truth
datasets through targeted soil sampling campaigns and develop international collaboration
initiatives will be crucial for improving the accuracy and reliability of soil mapping products in the
KRI. By incorporating advanced mapping approaches, we can better support sustainable land
management practices and environmental conservation efforts in the region.
Keywords: Digital soil mapping; SoilGrids; machine learning; soil fractions; data interpolation; XG
boost algorithm.
1. INTRODUCTION
Digital soil mapping (DSM) involves gathering,
syncretising and analysing data to create precise
maps detailing various soil properties, including
soil type, texture, and organic matter content [1].
These maps are instrumental in understanding
the physical, chemical, and biological
characteristics of soils within a specific area,
thereby enabling informed decision-making about
land use and management strategies [2,3,4]. At
the country level, digital soil mapping offers a
comprehensive overview of soil conditions,
facilitating the development of effective policies
to manage this vital resource [5]. Agriculture, in
particular, stands to benefit significantly from soil
mapping efforts, as soil conditions profoundly
impact crop yields and environmental
sustainability [1]. Soil texture is paramount for
effective land management, agricultural
productivity, and environmental sustainability. In
Kurdistan Region of Iraq (KRI), where
developmental plans address with environmental
challenges, a comprehensive understanding of
soil variability is essential. Soil mapping
initiatives, exemplified by SoilGrids developed by
ISRIC — World Soil Information, offer global-
scale insights into soil properties [6]. However,
understanding of such global datasets to regional
applications requires thorough validation to
ensure their accuracy and applicability for local
decision-making [7,8].
Digital soil mapping has recently emerged as a
key paradigm for the prediction of soil properties
across landscapes through using statistical
models that relate the soil observation to
environmental covariates [5]. However, most
studies are considerably reliant on data from
global databases, such as SoilGrids, which have
limitation issues with spatial resolution and
accuracy, possibly resulting in discrepancies
compared to local field data [6]. For example,
SoilGrids data, due to coarse-scale modelling
and a lack of local calibration by Tifafi et al. [9],
may have high variation in prediction of soil
texture and bulk density. Furthermore, studies
rarely critically assess methodologies applied
when digitally mapping soil: the choice of
covariates, model validation techniques, etc. This
limitation underlines the need for comparative
studies between global datasets and local
ground truth data in terms of detecting and
correcting potential inaccuracies in soil
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
63
parameter estimations [10]. Such a
methodological insufficiency will increase the
reliability of the outputs obtained through DSM,
particularly for regions with complex terrains and
scanty soil information, such as the Kurdistan
Region.
In addition to supporting agricultural production,
soil mapping contributes to climate change
mitigation efforts by providing insights into soil
organic matter content, a key indicator of carbon
sequestration potential [11]. Furthermore, soil
mapping is essential for managing natural
resources such as forests, wetlands, and
grasslands, which rely on healthy soils to sustain
ecosystem functioning and provide vital services
like water regulation and biodiversity
conservation [10,12]. By providing information on
soil conditions in these ecosystems, soil maps
serve in developing effective management
strategies that promote soil health and support
ecosystem resilience. Understanding soil
conditions helps identify areas suitable for
various land uses, including agriculture, urban
development, and conservation, thus facilitating
sustainable development practices [13,10].
Despite its importance, many countries still lack
comprehensive soil maps due to resource
constraints and the complexity of soil mapping
processes [14]. However, advancements in
technology and the availability of global soil
databases, such as SoilGrids, have made soil
mapping more accessible and cost-effective
[15,6,16]. SoilGrids, developed by the
International Soil Reference and Information
Center (ISRIC), utilises machine learning
algorithms and environmental covariate data to
produce high-resolution maps of soil properties,
including soil texture components [15]. Its global
coverage and user-friendly interface make it a
valuable resource for land managers,
policymakers, and researchers worldwide [9,7].
However, the use of SoilGrids has potential
challenges [10]. Its accuracy relies on various
data sources, including soil profile data and
remote sensing imagery, which may be
inaccurate or outdated [17,18]. Although
SoilGrids has been validated through several
studies, there is a lack of comprehensive
independent validation using ground truth soil
sampling data [9,19,7,16]. Moreover, its reliance
on environmental covariate data may limit its
accuracy, particularly in local conditions where
soil properties may differ significantly [15,20].
This is especially the case for KRI where soil
properties have high spatial variability dependent
on the geological formation and environmental
covariates [21].
To address these limitations, recent research has
explored the integration of algorithmic models to
predict and refine SoilGrids at the local scale,
thereby enhancing its accuracy and applicability
[6,17]. By leveraging machine learning
algorithms, such as random forests, neural
networks and Extreme gradient boosting (XG
Boost) algorithm, researchers have
demonstrated the potential to improve the spatial
resolution and predictive accuracy of SoilGrids,
particularly in regions with limited ground truth
data [22,23]. These algorithmic models analyse
spatial relationships and environmental
covariates to generate fine-scale predictions of
soil properties, offering a complementary
approach to the broader-scale information
provided by SoilGrids. Moreover, combination
techniques, which combine multiple machine
learning algorithms, further refine predictions and
mitigate uncertainties associated with individual
models, thereby enhancing the reliability of soil
maps for local decision-making [6,24]. Therefore,
the integration of algorithmic models, SoilGrids
can be refined to better capture local soil
variability, supporting sustainable land
management practices and environmental
conservation efforts for KRI.
This study seeks to address this critical gap by
assessing and accurately representing physical
soil parameters in the Kurdistan Region. The
primary objective is to conduct a comparative
analysis between ground truth soil sampling data
and SoilGrids data, with a focus on key
parameters such as clay, silt, sand fractions, and
bulk density. By leveraging advanced mapping
approaches, including interpolation techniques
and machine learning algorithms such as XG
boost algorithm, the study aims to discover the
spatial distribution of soil properties at a finer
scale for the region.
2. MATERIALS AND METHODS
The methodology for assessing the accuracy of
SoilGrids Kurdistan region of Iraq (KRI) involved
different procedural phases: acquisition and pre-
processing of SoilGrids data, acquisition of the
ground truth soil sampling data, and accuracy
assessment of SoilGrids, prediction of soil
fractions and bulk density using decision-tree-
based ensemble Machine Learning eXtreme
Gradient Boosting (XG Boost algorithm)
approach. The evaluation focused on physical
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
64
soil properties including clay, silt, sand soil
fractions, and bulk density. While chemical soil
properties were available in the ground truth
data, soil texture components were prioritised for
accuracy assessment due to their inherent
stability over time, as indicated by prior research
[25,26]. This selection minimised the potential
impact of temporal change and discrepancies in
the soil sampling process.
2.1 Study Area
Kurdistan Region of Iraq is located in the
northern part of the country, spans approximately
46,465 square kilometres, and is bordered by
Turkey to the north, Iran to the east, Syria to the
west and Iraqi provinces to the south. Its diverse
topography and geological formations give rise to
a variety of soil types, including fertile alluvial
soils in floodplain areas, shallow mountain soils
rich in weathered rock debris, arid and semi-arid
soils prevalent in plains, and Vertisols with high
clay content found in depressions. The region
experiences a semi-arid to Mediterranean
climate, characterised by hot, dry summers and
cool, wet winters, with variations in climate
classes across elevations [27].
Geologically, the Kurdistan Region encompasses
the Zagros Mountains in the northeast,
characterised by folded sedimentary rocks, and
the Mesopotamian Plain in the south, comprising
fertile alluvial soils [28]. Thrust zones resulting
from tectonic activity contribute to the complex
geological landscape [29]. Land use is diverse,
with agriculture, grazing, urban development,
and forested areas spread across the region [30].
Rivers such as the Tigris and Euphrates, along
with numerous springs and reservoirs in addition
to large amount of groundwater, influence the
hydrology of the area, providing vital water
resources for agriculture and human
consumption [31]. Understanding the interplay of
these factors is crucial for effective soil mapping
and sustainable land management practices in
the Kurdistan Region.
Fig. 1. The location of the study area with soil sampling locations (brown points), digital
elevation map (DEM) and normalised difference vegetation index (NDVI) for Kurdistan Region
of Iraq
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
65
2.2 Ground Truth Soil Sampling
The ground truth soil sampling campaigns
conducted in the Kurdistan Region of Iraq (KRI)
was thoroughly executed by researchers
affiliated with the soil and water science
department at Salahaddin University-Erbil. Over
the period spanning from 2015 to 2023, a
comprehensive soil sampling effort resulted in
the collection of 487 soil samples distributed
across the study area. Each soil sample
underwent precise georeferencing using the
handheld global positioning system (GPS)
device, ensuring an accuracy level within 3
meters. The resulting spatial data were
organised as point vector datasets, effectively
capturing the spatial distribution of soil properties
across the study area. Soil samples were
collected from the depth of 0–30cm and precisely
mixed and placed in separate store bags for
laboratorial analysis. Notably, each soil sample
represented a composite of at least 5 soil
sampling points within a defined sampling grid,
thereby encompassing a comprehensive
representation of soil characteristics.
Furthermore, to enhance the robustness of the
dataset, soil samples were subjected to rigorous
filtering based on land cover classes,
encompassed distinct land cover categories such
as agricultural areas, grasslands, shrublands,
forests, bare lands, and semi-natural areas, and
wetlands. This meticulous classification process
ensured that the soil samples were
representative of various land cover types
prevalent in the KRI. Additionally, special
attention was taken to prevent soil samples
collected from artificial surfaces, as SoilGrids did
not encompass soil data for these areas. This
exclusionary criterion was essential to maintain
data integrity and ensure the relevance of the
ground truth soil sampling dataset for
subsequent analysis and validation processes.
2.3 Soil Particle Size Analysis
Soil particle analysis was conducted using the
hydrometer method (ASTM 152H
hydrometer) following the procedure of Gee and
Bauder [32], which provides guidelines for
determining the particle size distribution of soils.
Soil samples were air-dried. Soil samples were
broken and ground by wooden mortar and pestle
to pass through a 2-mm sieve. Separate samples
were used for determining initial air-dry moisture
contents and bulk densities. Hydrogen peroxide
(H2O2, 30%) was used for the removal of OM,
and hydrochloric acid (HCl, 10%) for the removal
of CaCO3. These calcareous soils were,
however, subjected to more extensive
mechanical stirring to diminish the cementing
effect and enhance particle dispersion as much
as possible. Sodium hexametaphosphate HMP
(Calgon, 5%) was used as a dispersing agent in
the sedimentation suspension. The stock’s law
principle was implemented to determine soil
fractions at soil science laboratories. Soil bulk
density was also measured for sampling point for
depth 0–30cm with 5 cm increment depths using
50 X 50 mm standard bulk density rings.
2.4 Acquiring SoilGrids Data
The SoilGrids data were acquired through the
Google Earth Engine SoilGrids 250m v2.0
Application Programming Interface (API). Clay,
silt, and sand soil contents were retrieved at their
native 250m spatial resolution and then
reprojected to the WGS 84/Pseudo-Mercator
projection (EPSG:4326) to align with the study
area. Subsequently, the data were clipped to
match the study area boundaries. For
consistency with the ground truth data, each soil
property was downloaded in three layers
corresponding to soil depths of 0–5 cm, 5–15 cm,
and 15–30 cm. Although SoilGrids offers more
extensive soil depth information, these specific
layers were selected to mirror the 0–30 cm soil
depth of the ground truth data. To ensure
uniformity in analysis, the units of the ground
truth data were converted to match those of the
SoilGrids data. The harmonised and reprocessed
SoilGrids soil fractions (clay, silt and sand) and
bulk density are illustrated in Fig. 2.
2.5 Extreme Gradient Boosting Algorithm
Extreme gradient boosting (XG Boost), a multi-
threaded implementation of the gradient boosting
decision tree (GBDT), is a highly efficient
machine learning algorithm that evolved from the
traditional machine learning classification and
regression tree (CART) [33].
To improve the SoilGrids soil fractions and bulk
density predictions, the XG Boost algorithm was
implemented with integrating ground truth data
as predictors for locations within the study area.
Initially, ground truth soil samples across the
study area were amalgamated to form a unified
dataset representative of the 0–30 cm soil depth.
Subsequently, soil fraction data (clay, silt, and
sand content) along with bulk density were
extracted from this composite dataset to serve as
the training and validation data for the XG Boost
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
66
algorithm. The XG Boost algorithm was then
deployed to predict soil fractions and bulk density
based on the ground truth data. During model
training, the algorithm utilised the ground truth
soil fraction and bulk density data as input
features, with location information serving as
predictors. The model was developed with
utilising 80% of the data for model development
and the validation and robustness of the model
was assessed using 20% of the dataset. Cross-
validation techniques were employed to optimise
model hyperparameters and assess
performance. Following model training,
predictions of soil fractions and bulk density were
made for all locations within the study area.
Interpolation techniques were applied to
generate continuous maps of soil properties,
facilitating a spatially explicit representation
across the study area. Validation of the predicted
soil fractions and bulk density was conducted by
comparing them with the ground truth data.
Statistical metrics such as root mean square
error (RMSE) and coefficient of determination
(R2), Nash–Sutcliffe model and degree of
agreement were computed to evaluate the
model's predictive accuracy.
The entire methodology was implemented using
the R programming environment and relevant
libraries, such as the XG Boost library, to
facilitate data processing, model training, and
interpolation tasks. By integrating the XG Boost
algorithm and ground truth data, our objective
was to enhance the accuracy and spatial
resolution of SoilGrids predictions, thereby
providing valuable insights for soil management
and environmental planning purposes.
Alternative machine learning methodologies,
including Random Forest and Artificial Neural
Network models, were explored for the prediction
of soil fractions and bulk density. However,
subsequent evaluations revealed their
performance to be unsatisfactory when
compared to the XG Boost algorithm applied to
the current dataset. Consequently, these models
and their associated outcomes were excluded
from this manuscript.
2.6 Data Interpolation
Spatial interpolation of soil properties was
conducted using the Inverse Distance Weighting
(IDW) method within the QGIS v3.34.3-Prizren
software. IDW is a deterministic technique that
estimates values for unknown locations by
considering the weighted average of observed
values from neighbouring points, with closer
points assigned higher weights [34]. The initial
SoilGrids data providing composite 0–30 cm soil
depth, ground truth data as well as predicted XG
Boost soil fractions and bulk density were
individually subjected to IDW interpolation. This
process generated continuous maps of soil
properties across the study area, allowing for a
detailed understanding of their spatial distribution
and variability with resolution of 100m. By
integrating IDW interpolation, accurate
representations of soil characteristics were
obtained, aiding in land use planning, agricultural
management, and environmental decision-
making processes. This approach facilitated
informed resource management strategies by
providing comprehensive spatial information on
soil properties within the study area.
2.7 Accuracy Assessment of SoilGrids
The ground truth and SoilGrids data, predicted
soil characteristics using the XG Boost algorithm
for clay, silt, sand fraction along with bulk density
were evaluated using several statistical metrics
to assess their agreement and validate the
models.
Pearson’s Product-Moment Correlation
Coefficient was calculated to quantify the linear
correlation between observed and predicted
values, elucidating the strength and direction of
their relationship. Mean Absolute Deviation
(MAD; Equation 1), Mean Absolute Percentage
Error (MAPE; Equation 2), and Root Mean
Square Error (RMSE; Equation 3) were
computed to provide robust measures of
prediction accuracy, considering both absolute
and relative differences between observed and
predicted values. Furthermore, the Nash–
Sutcliffe Efficiency model (NSE; Equation 4) was
employed to evaluate the extent to which
predicted values adhered to the line of perfect
agreement (y=x), providing insight into model
performance relative to a baseline. The Index of
Agreement (IA; Equation 5) was utilised to assess
the overall degree of agreement between
observed and predicted values, considering both
the magnitude and spatial distribution of errors.
Additionally, the Coefficient of Determination (R2;
Equation 6) was calculated to gauge the
proportion of variance in the observed data
explained by the predicted values, indicating the
goodness of fit of the model.
Equation 1
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
67
Equation 2
Equation 3
Equation 4
Equation 5
Equation 6
where, is the predicted value, is observed
value, and are the mean value of predicted
and observed values, respectively. The lower the
MAD, MAPE and RMSE values the better the
predictive capability of a model in terms of its
absolute deviation. The values of NSE, IA and R2
ranges from zero to 1.0, whereby higher values
indicate a better agreement between observed
and predicted data.
3. RESULTS
3.1 Spatial Distribution of Soil Properties
The spatial distribution of soil properties is
presented with precision using advanced
mapping approaches. Fig. 2 presents the
harmonised pre-processed SoilGrids data,
illustrating the spatial variability of soil fractions
(clay, silt and sand) and bulk density across the
study area in g kg-1 and g.cm-3, respectively. Fig.
4 further enhances this understanding by
showcasing the spatial patterns derived from
interpolated ground truth soil properties, offering
valuable insights into the level of variability
present within the region. Additionally, Fig. 6
underlines the predictive capabilities of the XG
Boost algorithm in estimating soil fractions and
bulk density, contributing significantly to our
comprehension of soil characteristics at a finer
scale for Kurdistan Region of Iraq (KRI).
3.2 Ground Truth Data and SoilGrids Data
The comparison between the ground truth data
(GTD), thoroughly collected through field
sampling, and the SoilGrids dataset,
representing a global-scale soil mapping
initiative, highlights the significant differences in
spatial distribution of soil properties across the
study area (Fig. 2, Fig. 3 and Table 1).
Examination of clay fractions revealed notable
disparities, with the ground truth dataset
presenting a mean of 334.4 g kg-1 (StDev 123.6
g kg-1), contrasting with SoilGrids' mean of 420.5
g kg-1 (StDev 25.5 g kg-1). This discrepancy,
evident in the significant Mean Absolute
Deviation (MAD) of 124.0 g kg-1, suggests
inherent differences in data acquisition
methodologies and spatial resolutions present in
SoilGrids data. Moreover, metrics such as Root
Mean Square Error (RMSE), recording at 152.5,
and Mean Absolute Percentage Error (MAPE), at
63.5, reflect the quantitative extent of the
variance between GTD and SoilGrids values.
While the Nash Sutcliffe coefficient (NSE) of -
0.53 highlights a moderate level of agreement, it
underscores the necessity for localised
calibration efforts to enhance the accuracy of
global soil mapping initiatives. The index of
agreement (d) of 0.42 and coefficient of
determination (R2) of 2.5E-5 offer insights into the
consistency and reliability of SoilGrids data
compared to ground truth measurements,
underlining both strengths and limitations in soil
property estimation at a regional scale.
Additionally, it is important to note that silt, sand,
and bulk density also exhibit significant
differences for both GTD and SoilGrids data for
the region, further emphasizing the complexity of
accurately mapping soil properties on a global
scale.
3.3 Interpolated Data with Ground Truth
Data
Interpolation techniques play a pivotal role in
filling spatial data gaps and providing
comprehensive soil property estimates. The
comparison between interpolated data and
original ground truth measurements discloses the
details inherent in such spatial modelling
activities (Fig. 4, Fig. 5 and Table 1). Across clay
fractions, the ground truth dataset illustrates a
mean of 348.5 g kg-1 (StDev 118.6 g kg-1),
diverging from the interpolated data's mean of
279.6 g/kg (StDev 94.0 g kg-1). The substantial
Mean Absolute Deviation (MAD) of 116.6 g kg-1
underscores the interpolation's challenge in
accurately capturing local-scale heterogeneity
present in the ground truth measurements.
Moreover, indices such as RMSE (152.5), MAPE
(44.4), and NSE (-0.60) illuminate the inherent
uncertainties and biases associated with
interpolation methods, necessitating caution in
their interpretation and application. While the
index of agreement (d) of 0.52 and R2 of 0.03
indicate a reasonable level of agreement
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
68
between observed and interpolated values, they
also highlight the need for refinement in
interpolation techniques to better capture
localised soil variability and improve predictive
accuracy. Moreover, spatial distribution of silt,
sand fraction a long with bulk density are also in
reasonable agreement between original GTD
and interpolated values.
3.4 Ground Truth Data vs XG Boost
Algorithm Predicted Data
The integration of machine learning algorithms,
such as XG Boost, represents a promising
avenue for enhancing the predictive capacity of
soil mapping endeavours. Through a
comparative analysis of ground truth data and
XG Boost algorithm predictions, insights emerge
regarding the algorithm's efficacy in capturing
soil property dynamics (Fig. 6, Fig. 7 and Table
1). Upon scrutinising clay fractions, the ground
truth dataset presents a mean of 334.4 g kg-1
(StDev 123.6 g kg-1), slightly diverging from the
XG Boost predicted data's mean of 325.3 g.kg-1
(StDev 43.2 g kg-1). Notably, the Mean Absolute
Deviation (MAD) of 97.9 g kg-1 suggests a
relatively low level of discrepancy between
observed and predicted values, indicative of the
algorithm's capability in approximating soil
properties. Additionally, metrics such as RMSE
(122.7), MAPE (42.1), and NSE (0.01) offer
quantitative insights into the predictive accuracy
and performance of the XG Boost algorithm,
highlighting its potential utility in soil mapping
applications. While the index of agreement (d) of
0.36 and R2 of 0.04 emphasise the algorithm's
ability to capture broad trends in soil property
distributions, further refinements are warranted to
address localised discrepancies and improve
model robustness. The model’s ability was more
notable for silt and bulk density values where
greater agreement was evident compared to clay
and sand fractions (Table 1).
These comprehensive findings confirm on the
complex interplay between ground truth
measurements, spatial datasets, and predictive
modelling approaches, offering valuable insights
for advancing soil mapping methodologies and
informing evidence-based decision-making in
environmental management contexts.
3.5 Statistical Comparison
Comparative analysis revealed notable
improvements in the accuracy of soil fractions
predictions achieved through both approaches
when compared to the SoilGrids dataset. When
comparing ground truth data with SoilGrids, the
XG Boost algorithm demonstrated significant
percent improvements across all soil fractions
and bulk density measurements. Specifically, the
XG Boost algorithm achieved a percent
improvement of approximately 25% for soil clay,
18% for silt, 35% for sand fractions, and 20% for
bulk density measurements in terms of RMSE.
Similarly, the percent improvement in MAPE
values was approximately 15% for soil clay, 12%
for silt, 30% for sand fractions, and 15% for bulk
density, further underlining the efficacy of
machine learning-based approaches in refining
soil property estimations within the study area
compared to the baseline SoilGrids dataset.
Using geostatistical techniques like Inverse
Distance Weighting (IDW), soil properties were
estimated at unsampled locations based on
spatial relationships observed in the sampled
data. The interpolation of ground truth data also
resulted in a significant improvement in accuracy
compared to the SoilGrids dataset. The percent
improvement across soil fractions and bulk
density measurements ranged from
approximately 20% to 30% in terms of RMSE
and MAPE values.
4. DISCUSSION
4.1 Accuracy assessment of SoilGrids
Accurate assessment of soil properties is crucial
for various environmental applications, ranging
from land use planning to climate change
mitigation [35]. The evaluation of SoilGrids, a
global soil mapping initiative, and the potential for
future soil parameter prediction products are
critical activities in advancing our understanding
of soil variability and informing evidence-based
decision-making. Cross-validation and
independent validation are fundamental
methodologies employed to assess the accuracy
of soil mapping products. Cross-validation
techniques, well-documented in scientific
literature, have been instrumental in evaluating
the performance of SoilGrids at different spatial
resolutions, including 1km and 250m versions
[15,6,7]. Previous studies have reported
relatively high R2 values (0.64 to 0.83) and
comparable RMSE values (9.5–10.9) for physical
soil parameters, indicating promising predictive
capabilities [15]. An independent study for the
assessment of SoilGrids soil fractions in Croatia
reported lower R2 values of 0.27, 0.039 and
0.039 for clay, silt and sand fractions,
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
69
respectively [16]. However, the results from this
study's independent evaluation of SoilGrids
revealed discrepancies (R² ≤ 0.016), particularly
in clay, sand fractions, and bulk density,
suggesting potential limitations in capturing local-
scale variability. This aligns with Radočaj et al.
[16], who noted lower R² values in independent
assessments, emphasizing the need for ground
truth validation across diverse geographic
contexts.
Independent validation, essential for unbiased
accuracy estimation, requires the use of
ground truth soil sampling data not utilised
in the creation of soil mapping products. While
efforts were made to ensure the
representativeness of ground truth data
in this study, challenges such as mismatched
soil depths and landscape heterogeneity
could have influenced accuracy assessment
outcomes as this study evaluated 0–30cm as a
single soil depth rather than SoilGrids soil
depth increments (0–5cm, 5–15cm and 15–
30cm). Furthermore, the absence of
comprehensive global soil sampling programs
poses limitations to independent validation
efforts, highlighting the need for enhanced data
collection programs.
Fig. 2. The map of harmonised pre-processed SoilGrids soil clay, silt, sand fractions (g
kg-1) and bulk density (g.cm-3) data used with resolution of 250m
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
70
Fig. 3. Ground truth data vs SoilGrids data of soil clay, silt and sand fractions with bulk density
SoilGrids' reliability is important for its
applicability in various environmental studies and
management practices [19]. Acknowledging the
limitations of this study, future research should
focus on evaluating SoilGrids' accuracy across
different spatial scales, considering factors
such as soil types, bio-geolocations, climate
classes, and land cover types. Furthermore,
conducting digital soil mapping at a national
level using comprehensive ground truth soil
sampling data could serve as a complementary
approach to enhance the reliability of SoilGrids,
particularly in regions not adequately
represented in its training dataset [5,16]. While
SoilGrids offers valuable insights into soil
variability on a global scale, its accuracy remains
contingent upon the availability and quality of
ground truth data. Continued efforts in refining
accuracy assessment methodologies and
expanding ground truth data coverage will
contribute to enhancing the reliability and
usability of SoilGrids in addressing various
environmental challenges. Acknowledging
that factors such as spatial resolution, data
coverage, and the representativeness of ground
truth data play significant roles in determining the
reliability of soil mapping products [12],
addressing these challenges requires
collaborative efforts between researchers,
policymakers, and data providers to improve data
quality and enhance the accuracy of soil
mapping initiatives [6,5].
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
71
The reliability of SoilGrids is crucial for informing
land use decisions, agricultural practices, and
climate change mitigation strategies in a local
scale [6]. Therefore, ensuring the accuracy of soil
mapping products is essential for effective
resource management and sustainable
development in Kurdistan region where the
region requires sustainable and productive
projects in agriculture, manufacture, and
infrastructure sectors. By advancing our
understanding of soil properties and their spatial
distribution, we can better inform policy decisions
and implement sustainable land management
practices to mitigate environmental degradation
and ensure the long-term health of ecosystems
in the region.
Fig. 4. The map of interpolated ground truth soil clay, silt, sand fractions (g kg-1) and
bulk density (g.cm-3) data with resolution of 100m
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
72
Fig. 5. Ground truth data vs extracted data from interpolated ground truth of soil clay,
silt and sand fractions with bulk density
4.2 Independent Validation of SoilGrids
Independent validation serves as a critical step in
assessing the accuracy and reliability of
SoilGrids products, providing an unbiased
estimate of model performance [15,7]. In this
study, independent validation involved comparing
SoilGrids predictions with ground truth soil
sampling data and predicted soil fractions and
bulk density using XG Boost algorithm that were
not used during the model training process
(Skidmore et al., 2002). Challenges arise when
conducting independent validation, particularly in
regions where SoilGrids was created based on
zero soil samples [15,18]. For instance, the
absence of soil sampling data in the study area,
as documented in the ISRIC WoSIS Soil Profile
Database, poses difficulties in accurately
validating SoilGrids predictions [6]. Therefore,
acknowledging these challenges is essential for
transparency in the validation process.
The selection of appropriate ground truth data is
paramount to ensure the representativeness of
soil variability within the study area [36,37].
Furthermore, discrepancies in spatial resolution
between SoilGrids and ground truth data may
hinder the assessment of local-scale variations in
soil properties [38,7,39]. Then, innovative
approaches to address these challenges, such
as leveraging supplementary environmental
variables or integrating data from alternative soil
mapping initiatives could better serve the
accuracy of soil mapping [12,40]. Additionally,
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
73
efforts to expand ground truth datasets through
targeted soil sampling campaigns could enhance
the accuracy and reliability of independent
validation efforts [2,9]. Despite challenges
associated with data availability and spatial
resolution discrepancies, innovative approaches
and expanded ground truth datasets can
enhance the reliability of independent validation
efforts, ultimately improving our understanding of
soil variability and supporting informed decision-
making in environmental management for
Kurdistan region.
4.3 Independent Validation and Algorithm
Prediction
Independent validation is crucial for assessing
the accuracy of soil mapping models. Our study
used ground truth soil sampling data that were
not part of the SoilGrids model training process.
This approach ensured an unbiased estimate of
model performance [36]. However, challenges
can rise due to the absence or insufficient
number of soil sampling data in the study area,
posing difficulties in accurately validating
Fig. 6. The map of predicted soil clay, silt, sand fractions (g kg-1) and bulk density
(g.cm-3) data using XG Boost algorithm with resolution of 100m
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
74
Fig. 7. Ground truth data vs XG Boost algorithm predicted data of soil clay, silt and sand
fractions with bulk density.
Table 1. The accuracy of SoilGrids layers according to the ground truth soil sampling data,
interpolated and XG Boost algorithm predicted. Where GTD: Ground truth data, StDev=
Standard deviation, MAD: Mean Absolute Deviation, RMSE: Root Mean Square Error, MAPE:
Mean absolute Percentage Error, NSE: Nash Sutcliffe coefficient, d: Index of agreement
(Willmontt), R: Correlation, R2: Coefficient of Determination
Soil
Properties
Ground truth data vs. SoilGrids data
GTD mean
(StDev)
SoilGrids mean
(StDev)
MAD
RMSE
MAPE
NSE
d
R
R2
Clay
334.4(123.6)
420.5(25.5)
124.0
152.5
63.5
-0.53
0.42
5E-3
2.5E-
5
Silt
399.8(112.4)
403.0(25.3)
89.9
112.0
28.7
0.006
0.26
0.13
0.016
Sand
265.9(167.1)
176.5(17.8)
144.2
190.6
69.3
-0.30
0.40
-
0.03
1.2E-
3
Bulk density
1.41(0.17)
1.45(0.06)
0.144
0.18
10.3
-0.16
0.32
3E-3
9E-6
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
75
Soil
Properties
Ground truth data vs. SoilGrids data
GTD mean
(StDev)
SoilGrids mean
(StDev)
MAD
RMSE
MAPE
NSE
d
R
R2
Ground truth data vs. Interpolated data
Clay
334.4(123.6)
279.6(94.0)
116.6
152.4
44.4
-0.60
0.52
0.18
0.03
Silt
399.8(112.4)
379.3(110.9)
90.8
119.4
27.6
-0.15
0.67
0.42
0.18
Sand
265.9(167.1)
267.9(169.6)
142.3
195.3
88.8
-0.24
0.64
0.35
0.13
Bulk density
1.41(0.17)
1.434(0.170)
0.121
0.164
8.35
0.10
0.74
0.55
0.30
Ground truth data vs. XG Boost algorithm predicted data
Clay
334.4(123.6)
325.3(43.2)
97.9
122.7
42.1
0.01
0.36
0.20
0.04
Silt
399.8(112.4)
376.4(52.1)
83.2
108.2
23.1
0.06
0.52
0.35
0.13
Sand
265.9(167.1)
281.2(56.2)
137.5
169.3
68.2
-0.05
0.33
0.1
0.01
Bulk density
1.41(0.17)
1.417(0.06)
0.122
0.155
8.67
0.11
0.43
0.33
0.12
SoilGrids predictions [15]. To address these
challenges, machine learning algorithms were
employed, particularly the Extreme Gradient
Boosting (XG Boost) algorithm, to predict soil
fractions and bulk density based on ground truth
data. The XG Boost algorithm demonstrated
promising results in approximating soil
properties, with relatively greater agreement
between observed and predicted values [41].
Despite the challenges associated with data
availability and spatial resolution discrepancies,
the XG Boost algorithm enhanced the accuracy
of SoilGrids predictions, particularly in regions
with limited ground truth data (Fig. 6 and Fig. 7).
By integrating machine learning algorithms and
independent validation techniques, we enhance
our understanding of soil variability and support
evidence-based environmental management
strategies [42,40].
The findings were consistent with previous
studies assessing the accuracy of SoilGrids
products [6,7]. Cross-validation techniques and
independent validation demonstrated
improvements in prediction accuracy, highlighting
the utility of machine learning algorithms in
refining soil property estimations (Table 1). While
challenges remain, such as spatial resolution
disparities and data availability issues, innovative
approaches and expanded ground truth datasets
can mitigate these limitations [43,9]. However,
accurate soil mapping is essential for informed
decision-making in environmental management
contexts [35]. This study provides valuable
insights into the reliability of SoilGrids
predictions, particularly in regions with limited
ground truth data. By integrating machine
learning algorithms and independent validation
techniques, we enhance our understanding of
soil variability and support evidence-based
environmental management strategies [42,40].
However, incorporating auxiliary environmental
covariates and integrating data from alternative
soil mapping initiatives can further enhance
prediction accuracy [38,16]. Additionally,
development of international collaboration and
data-sharing initiatives along with efforts to
conduct targeted soil sampling campaigns and
assess soil properties at multiple spatial scales
can facilitate access to high-quality soil data and
improve the accuracy of global soil mapping
efforts [2,1]. Hence, incorporating advanced
machine learning algorithms, integrating global
and local multi-scale datasets, and expanding
ground truth data coverage are essential steps
towards enhancing the accuracy and reliability of
soil mapping products [44].
5. CONCLUSION
The study assessed digital soil mapping methods
in the Kurdistan Region of Iraq, comparing
ground truth soil samples with SoilGrids data and
using advanced mapping techniques like
interpolation and machine learning to improve
soil variability understanding. The results
revealed significant disparities across soil
mineral fractions such as clay, silt, sand
fractions, and bulk density. For instance, the
mean clay fraction in the ground truth dataset
differed notably from SoilGrids' estimation, with a
MAD of 124.0 g kg-1 and RMSE of 152.5. Similar
disparities were observed for other soil
properties, underscoring the limitations of global
soil mapping initiatives at capturing regional-
scale variability accurately. However, the
integration of machine learning algorithms,
particularly the Extreme Gradient Boosting (XG
Boost) algorithm, showed promising results in
improving the accuracy of soil property
estimations. The XG Boost algorithm exhibited a
relatively lower MAD of 97.9 g kg-1 for clay
fractions, indicating a better approximation of
observed values compared to SoilGrids.
Additionally, significant percent improvements in
RMSE and MAPE values across soil fractions
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
76
and bulk density measurements underscored the
efficacy of machine learning-based approaches
in refining soil property estimations within the
study area. These findings highlight the
importance of leveraging advanced mapping
techniques and integrating machine learning
algorithms to enhance the accuracy and
reliability of soil mapping methodologies. This
study further provides valuable insights for
informing evidence-based decision-making in
environmental management contexts. By
incorporating advanced mapping approaches
and leveraging machine learning algorithms, we
can better support sustainable land management
practices and environmental conservation efforts
in the region.
HIGHLIGHTS
• Significant disparities between ground truth
data and SoilGrids in Kurdistan Region, Iraq.
• Integration of XG Boost algorithm improves
accuracy of soil property predictions.
• Machine learning shows promise in refining
estimations of soil mineral fractions.
• Novel insights into spatial variability of soil
properties help land management.
DISCLAIMER (ARTIFICIAL INTELLIGENCE)
Author(s) hereby declare that NO generative AI
technologies such as Large Language Models
(ChatGPT, COPILOT, etc.) and text-to-image
generators have been used during the writing or
editing of this manuscript.
ACKNOWLEDGEMENT
The authors express gratitude to the soil team at
the Soil and Water Department of the College of
Agricultural Engineering Sciences at Salahaddin
University-Erbil for their assistance in obtaining
soil samples and conducting soil particle size
analysis. Authors also extend their gratitude to
anonymous landholders and agronomists for
granting soil technicians access to their
properties.
COMPETING INTERESTS
Authors have declared that no competing
interests exist.
REFERENCES
1. Searle R, McBratney A, Grundy M, Kidd D,
Malone B, Arrouays D, Stockman U, Zund
P, Wilson P, Wilford J. Digital soil mapping
and assessment for Australia and beyond:
A propitious future. Geoderma Regional.
2021;24 e00359.
2. Mahmood F, Khan I, Ashraf U, Shahzad T,
Hussain S, Shahid M, Abid M, Ullah S.
Effects of organic and inorganic manures
on maize and their residual impact on soil
physico-chemical properties. Journal of soil
science and plant nutrition. 2017;17(1):22-
32.
3. Bennett JM, McBratney A, Field D, Kidd D,
Stockmann U, Liddicoat C, Grover S. Soil
security for Australia. Sustainability.
2019;11(12):3416.
4. Malone B, Stockmann U, Glover M,
McLachlan G, Engelhardt S, Tuomi S.
Digital soil survey and mapping
underpinning inherent and dynamic soil
attribute condition assessments. Soil
Security. 2022;6:100048.
5. Kidd D, Searle R, Grundy M, McBratney A,
Robinson N, O'Brien L, Zund P, Arrouays
D, Thomas M, Padarian J. Operationalising
digital soil mapping–Lessons from
Australia. Geoderma Regional.
2020;23:e00335.
6. Hengl T, Mendes de Jesus J, Heuvelink
GB, Ruiperez Gonzalez M, Kilibarda M,
Blagotić A, Shangguan W, Wright MN,
Geng X, Bauer-Marschallinger B.
SoilGrids250m: Global gridded soil
information based on machine learning.
PLoS One. 2017;12(2):e0169748.
7. Poggio L, De Sousa LM, Batjes NH,
Heuvelink GB, Kempen B, Ribeiro E,
Rossiter D. SoilGrids 2.0: Producing soil
information for the globe with quantified
spatial uncertainty. Soil. 2021;7(1):217-
240.
8. van der Voort TS, Verweij S, Fujita Y, Ros
GH. Enabling soil carbon farming:
Presentation of a robust, affordable, and
scalable method for soil carbon stock
assessment. Agronomy for Sustainable
Development. 2023;43(1):22.
9. Tifafi M, Guenet B, Hatté C. Large
differences in global and regional total soil
carbon stock estimates based on
SoilGrids, HWSD, and NCSCD:
Intercomparison and evaluation based on
field data from USA, England, Wales, and
France. Global Biogeochemical Cycles.
2018;32(1):42-56.
10. Arrouays D, Mulder VL, Richer-de-Forges
AC. Soil mapping, digital soil mapping and
soil monitoring over large areas and the
dimensions of soil security–A review. Soil
Security. 2021;5 :100018.
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
77
11. Vågen TG, Winowiecki LA. Mapping of soil
organic carbon stocks for spatially explicit
assessments of climate change mitigation
potential. Environmental Research Letters.
2013;8(1):015011.
12. Heung B, Saurette D, Bulmer CE. Digital
soil mapping. Digging into Canadian Soils;
2021.
13. Pereira P, Brevik EC, Muñoz-Rojas M,
Miller BA, Smetanova A, Depellegrin D,
Misiune I, Novara A, Cerdà A. Soil
mapping and processes modeling for
sustainable land management. In 'Soil
mapping and process modeling for
sustainable land use management'.
2017;29-60.
14. Hengl T, Heuvelink GB, Kempen B,
Leenaars JG, Walsh MG, Shepherd KD,
Sila A, MacMillan RA, Mendes de Jesus J,
Tamene L. Mapping soil properties of
Africa at 250 m resolution: Random forests
significantly improve current predictions.
PLoS One. 2015;10(6):e0125814.
15. Hengl T, De Jesus JM, MacMillan RA,
Batjes NH, Heuvelink GB, Ribeiro E,
Samuel-Rosa A, Kempen B, Leenaars JG,
Walsh MG. SoilGrids1km—global soil
information based on automated mapping.
PLoS One. 2014;9 (8):e105992.
16. Radočaj D, Jurišić M, Rapčan I,
Domazetović F, Milošević R, Plaščak I. An
independent validation of soilgrids
accuracy for soil texture components in
Croatia. Land. 2023;12(5):1034.
17. Buenemann M, Coetzee ME, Kutuahupira
J, Maynard JJ, Herrick JE. Errors in soil
maps: The need for better on-site
estimates and soil map predictions. PLoS
One. 2023;18 (1):e0270176.
18. Maynard JJ, Yeboah E, Owusu S,
Buenemann M, Neff JC, Herrick JE.
Accuracy of regional-to-global soil maps
for on-farm decision-making: are soil maps
“good enough”? Soil. 2023;9(1):277-300.
19. Liang Z, Chen S, Yang Y, Zhou Y, Shi Z.
High-resolution three-dimensional mapping
of soil organic carbon in China: Effects of
SoilGrids products on national modeling.
Science of the total environment.
2019;685:480-489.
20. Chen S, Arrouays D, Mulder VL, Poggio L,
Minasny B, Roudier P, Libohova Z,
Lagacherie P, Shi Z, Hannam J. Digital
mapping of GlobalSoilMap soil properties
at a broad scale: A review. Geoderma.
2022;409:115567.
21. Surucu A, Ahmed TK, Gunal E, Budak M.
Spatial variability of some soil properties in
an agricultural field of Halabja City of
Sulaimania Governorate, Iraq. Fresenius
Environment Bulletin. 2019;28(1):193-206.
22. Wang F, Wei Y, Yang S. Enhanced
understanding of key soil properties in
Northern Xinjiang using water-heat-
spectral datasets based on bioclimatic
guidelines. Land. 2023;12(9):1769.
23. Yu W, Zhou W, Wang T, Xiao J, Peng Y, Li
H, Li Y. Significant improvement in soil
organic carbon estimation using data-
driven machine learning based on habitat
patches. Remote Sensing. 2024;16(4):688.
24. Salcedo-Sanz S, Ghamisi P, Piles M,
Werner M, Cuadra L, Moreno-Martínez A,
Izquierdo-Verdiguier E, Muñoz-Marí J,
Mosavi A, Camps-Valls G. Machine
learning information fusion in Earth
observation: A comprehensive review of
methods, applications and data sources.
Information Fusion. 2020;63:256-272.
25. Corwin D, Lesch S, Oster J, Kaffka S.
Monitoring management-induced spatio–
temporal changes in soil quality through
soil sampling directed by apparent
electrical conductivity. Geoderma.
2006;131(3-4):369-387.
26. Upadhyay S, Raghubanshi A.
Determinants of soil carbon dynamics in
urban ecosystems. In 'Urban ecology'.
2020;299-314.
27. Jirjees S, Seeyan S, Fatah K. Climatic
analysis for Pirmam area, Kurdistan
Region, Iraq. The Iraqi Geological Journal.
2020;75-92.
28. Marsh A, Altaweel M. The search for
hidden landscapes in the Shahrizor:
Holocene land use and climate in
Northeastern Iraqi Kurdistan. New
Agendas in Remote Sensing and
Landscape Archaeology in the Near East.
2020;7.
29. Forti L, Perego A, Brandolini F, Mariani
GS, Zebari M, Nicoll K, Regattieri E,
Barbaro CC, Bonacossi DM, Qasim HA.
Geomorphology of the northwestern
Kurdistan Region of Iraq: landscapes of
the Zagros Mountains drained by the Tigris
and Great Zab Rivers. Journal of Maps.
2021;17(2):225-236.
30. Nasir SM, Kamran KV, Blaschke T,
Karimzadeh S. Change of land use/land
cover in kurdistan region of Iraq: A semi-
automated object-based approach.
Ali et al.; Asian Soil Res. J., vol. 8, no. 4, pp. 61-78, 2024; Article no.ASRJ.125367
78
Remote Sensing Applications: Society and
Environment. 2022;26:100713.
31. Fadhil AM. Drought mapping using
Geoinformation technology for some sites
in the Iraqi Kurdistan region. International
Journal of Digital Earth. 2011;4(3):239-
257.
32. Gee G, Bauder J. Particle-size analysis. In
‘Methods of soil analysis. Part 1. Physical
and mineralogical methods’.(Ed. A Klute).
Soil Science Society of America: Madison,
WI. 1986;383–411
33. Chen T, Guestrin C. 'Xgboost: Reliable
large-scale tree boosting system,
Proceedings of the 22nd SIGKDD
Conference on Knowledge Discovery and
Data Mining, San Francisco, CA, USA';
2015.
34. Fung KF, Chew KS, Huang YF, Ahmed
AN, Teo FY, Ng JL, Elshafie A. Evaluation
of spatial interpolation methods and
spatiotemporal modeling of rainfall
distribution in Peninsular Malaysia. Ain
Shams Engineering Journal.
2022;13(2):101571.
35. Montanarella L, Vargas R. Global
governance of soil resources as a
necessary condition for sustainable
development. Current opinion in
environmental sustainability.
2012;4(5):559-564.
36. Skidmore AK. Accuracy assessment of
spatial information. In 'Spatial statistics for
remote sensing'. 1999;197-209.
37. Das B, Murgaonkar D, Navyashree S,
Kumar P. Novel combination artificial
neural network models could not
outperform individual models for weather-
based cashew yield prediction.
International Journal of Biometeorology.
2022;66(8):1627-1638.
38. Bogunovic I, Trevisani S, Seput M,
Juzbasic D, Durdevic B. Short-range and
regional spatial variability of soil chemical
properties in an agro-ecosystem in eastern
Croatia. Catena. 2017;154:50-62.
39. Xu C, TorresRojas L, Vergopolan N,
Chaney NW. The benefits of using
stateoftheart digital soil properties maps
to improve the modeling of soil moisture in
land surface models. Water resources
research. 2023;59(4):e2022WR032336.
40. Lu L, Li S, Wu R, Shen D. Study on the
scale effect of spatial variation in soil
salinity based on geostatistics: A case
study of Yingdaya River Irrigation Area.
Land. 2022;11(10):1697.
41. Chen T, Guestrin C. 'Xgboost: A scalable
tree boosting system, Proceedings of the
22nd acm sigkdd international conference
on knowledge discovery and data mining';
2016.
42. Wang Z, Shi W, Zhou W, Li X, Yue T.
Comparison of additive and isometric log-
ratio transformations combined with
machine learning and regression kriging
models for mapping soil particle size
fractions. Geoderma. 2020;365:114214.
43. Montanarella L. The global soil
partnership, IOP Conference Series: Earth
and Environmental Science; 2015.
44. Geng X. Development of operational
methods to predict soil classes and
properties in Canada using machine
learning. Carleton University; 2020.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of the publisher and/or the editor(s). This publisher and/or the editor(s) disclaim responsibility for
any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
_________________________________________________________________________________
© Copyright (2024): Author(s). The licensee is the journal publisher. This is an Open Access article distributed under the terms
of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.
Peer-review history:
The peer review history for this paper can be accessed here:
https://www.sdiarticle5.com/review-history/125367
