digital mapping of soil particle size distribution in an
TRANSCRIPT
Digital Mapping of Soil Particle Size Distribution in
an Alluvial Plain Using the Random Forest Algorithm
Fuat KAYA*; Levent BAŞAYİĞİT
Isparta University of Applied Sciences, Faculty of Agriculture, Department of Soil
Science and Plant Nutrition, Çünür, Isparta 32260, Turkey [email protected],
Introduction
The spatial distribution of physical soil
properties is an important requirement
in practice as basic input data. Soil
texture is one of the most important
physical properties affecting water
holding capacity, nutrient availability,
and crop development. A spatial
distribution map of soil texture at a high
spatial resolution is essential data for
crop planning and management.
Traditionally, soil texture can be determined in the field by hand feel method,
however, it is confirmed by laboratory analysis of particle size fractions
(sand, silt, and clay) using the hydrometer method [1]. Conventional soil
texture analysis methods are expensive. Also, it requires a large number of
samples to obtain higher resolution spatial distribution of soil texture over
large areas.
Introduction
Digital soil maps can be used to show the soil’s ability to perform certain
functions. The digital soil mapping technique provides solutions for estimating
the soil properties and soil texture classes based on quantitative soil-landscape
models [2-3].
Introduction
Topography controls the
distribution and accumulation
of water and energy in the
pedosphere, so it has been
widely accepted in the
literature as the dominant
factor affecting soils and soil
properties [4-5].
Introduction
Method
Digital soil mapping is the method of spatial data generation with the
advantages of current technologies. It supplies fast, accurate, and
reproducible results.
In this study, a soil texture map with 30 m spatial resolution was produced
for an alluvial plain covering an area of approximately 10,000 ha. In the
study, 11 Topographic Environmental Variables obtained from NASA's
ASTER Global Digital Elevation model were used. Another input
parameters were clay, silt, and sand values determined for 91 soil samples
obtained through field studies.
Random Forest Algorithm offers interpretability for pedological information
extraction by determining the importance of environmental variables in
digital soil mapping.
Method
Environmental variable
extraction,
modeling,
and spatial mapping
R Core Environment (3.6.1) [7]
and related packages were used for
environmental variable extraction,
modeling, and spatial mapping.
Results - Soil properties descriptive statistics
n Mean Standart D Min. Maks. Skewness. Kurtosis Coefficient
Variation (%)
Clay(%) 91 36.91 13.68 13.28 73.76 0.82 -0.22 37.06
Silt (%) 91 25.81 7.16 12.47 51.94 1.13 1.64 27.72
Sand(%) 91 37.29 15.01 6.35 69.3 -0.34 -0.87 40.26
If it is desired to evaluate the degree of variability for soil properties by
considering the coefficient of variation, it is reported that there is a low
variation of 0-15%, medium variation of 16-35% and high variation coefficient
values (CV) of more than 36% [6]. Among the soil properties, clay and sand
have high variability in an alluvial plain.
Results -Model Accuracy
Results – Variable Importance
Results – Spatial Mapping
Conclusion
As a result of the study, digital soil maps obtained with the random forest
algorithm show compatibility with the study area.
The idea that the Random Forest Algorithm is compatible with the
compliance problem in small data sets is controversial.
Pixel-based soil properties maps obtained within the framework of soil
science can be controlled by creating soil texture class maps using
geographic information systems.
In this way, the evaluation of models produced in small data sets with
overfitting interpretation can be pedologically more effective.
References
1. Bouyoucos, G. J.: Hydrometer method improved for making particle size analyses of soils.
Agronomy J. 54(5), 464-465 (1962). doi: 10.2134/agronj1962.00021962005400050028x
2. McBratney, A.B. Santos, M.M. Minasny, B.: On digital soil mapping. Geoderma. 117, 3–52
(2003). doi: 10.1016/S0016-7061(03)00223-4
3. Dharumarajan, S., Kalaiselvi, B., Suputhra, A., Lalitha, M., Hegde, R., Singh, S. K.,
Lagacherie, P.: Digital soil mapping of key GlobalSoilMap properties in Northern Karnataka
Plateau. Geoderma Reg. 20, e00250 (2020). doi: 10.1016/j.geodrs.2019.e00250
4. Hewitt, A.E.: Predictive modeling in soil survey. Soils Fertil. 3, 305–315 (1993)
5. Hudson, B.D.: The soil survey as paradigm-based science. Soil Sci. Soc. Am. J. 56, 836–841
(1992). doi: 10.2136/sssaj1992.03615995005600030027x
6. Oku, E., Essoka, A., & Thomas, E. (2010). Variability in soil properties along an Udalf
toposequence in the humid forest zone of Nigeria. Agriculture and Natural Resources, 44(4),
564-573.
7. R Core Team (2019). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.