Rev. Peru. Investig. Agropecu. 3(2), e63
ISSN: 2955-8530
e-ISSN: 2955-831X
DOI: 10.56926/repia.v3i2.63
Universidad Nacional Autónoma de Alto Amazonas
Original article / Artículo original
Received: 15/06/2024
Accepted: 29/09/2024
Published: 10/10/2024
* carevalo@unaaa.edu.pe (corresponding author)
© Authors. This is an open access article
distributed under the terms of the Creative Commons
Attribution 4.0 International License.
Prediction of soil fertility using machine learning in Alto
Amazonas province, Peru
Predicción de la fertilidad del suelo mediante aprendizaje automático en la
provincia de Alto Amazonas, Perú
César Oswaldo Arévalo-Hernández 1* ; Enrique Arévalo-Gardini 1 ; Luis Alberto Arévalo-López 1 ;
Oscar Tuesta-Hidalgo 1 ; Dayani Shirley Romero-Vela 1; Claudia Elizabeth Ruiz-Camus 1
3Facultad de Ciencias, Universidad Nacional Autónoma de Alto Amazonas, Yurimaguas, Perú
ABSTRACT
The objective of the work was to predict soil fertility in the province of Alto Amazonas with the use of satellite images and
machine learning techniques. The study was in the province of Alto Amazonas in Peru. Soil sampling was carried out in all
the provinces, totalling 100 samples. Afterwards, soil physical (texture) and chemical analyses were performed. Satellite
images were obtained from USGS, and vegetation indexes were calculated based on these images. Finally, descriptive
analysis and machine learning modelling using 06 algorithms (GLM, CUBIST, KKNN, SVM, Random Forest and NN) were
used and selected based on their R2 and rmse. In this work, we observed that most soils in the province have low pH, P,
Mg, K and high acidity. We also managed to achieve good predictions for pH, Ca, Mg and CEC, and we observed that the
most successful algorithm was Random Forest. Nevertheless, for Al, CUBIST performed better. This is one of the first works
using machine learning to predict soil fertility in the Peruvian Amazon, and we hope it may serve as a base for future
projects.
Keywords: Random Forest; amazon soils; soil modelling; acid soils
RESUMEN
El objetivo del trabajo fue predecir la fertilidad del suelo en la provincia de Alto Amazonas con el uso de imágenes satelitales
y técnicas de aprendizaje automático. El estudio se ubicó en la provincia de Alto Amazonas en Perú. Se realizaron muestreos
de suelos en toda la provincia, totalizando 100 muestras. Posteriormente se realizaron análisis físicos (textura) y químicos
del suelo. Las imágenes satelitales se obtuvieron del USGS y los índices de vegetación se calcularon con base en estas
imágenes. Finalmente, se utilizó análisis descriptivo y modelado de aprendizaje automático utilizando 06 algoritmos (GLM,
CUBIST, KKNN, SVM, Random Forest y NN) que se seleccionaron en función de su R2 y RMSE. En este trabajo observamos
que la mayoría de los suelos de la provincia tienen bajos pH, P, Mg, K y alta acidez. También se lograron obtener buenas
predicciones para pH, Ca, Mg y CIC y se observó que el algoritmo más exitoso fue Random Forest. Sin embargo, para Al,
Cubist tuvo mejores resultados. Este es uno de los primeros trabajos que utiliza aprendizaje automático para predecir la
fertilidad del suelo en la Amazonía peruana y se espera que pueda servir como base para futuros proyectos.
Palabras clave: Random Forest; suelos amazónicos; modelamiento de suelos; suelos ácidos
Citation / Cómo citar: Arévalo-Hernández, C.O., Arévalo-Gardini, E., Arévalo-López, L. A., Tuesta-Hidalgo, O., Romero-Vela, D.S., & Ruiz-
Camus, C.E. (2024). Prediction of soil fertility using machine learning in Alto Amazonas province, Peru.
Revista Peruana de Investigación
Agropecuaria
,
3
(2), e63. https://doi.org/10.56926/repia.v3i2.63
Editor: Dr. Fred William Chu Koo
2 Revista Peruana de Investigación Agropecuaria
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
1. INTRODUCTION
Soil is one of the main natural components in the production of agricultural and forestry crops. To
evaluate their physical, chemical or biological characteristics, random sampling is carried out to
represent the areas of interest; however, this type of analysis is often expensive and spatially
unrepresentative (Watt et al., 2019). However, this has been the classic way of analyzing and
diagnosing the characteristics related to soil fertility and its productive capacity in agricultural and
forestry crops (Delgado-Caballero et al., 2009).
With the improvement of technology and the use of geographic information systems, it is possible
to carry out sampling considering variations in climate, topography, mineralogy and others, allowing
mapping areas to be better identified (Campos et al., 2019). In general, the sampling scheme and
sampling design are fundamental in digital soil mapping since they allow for obtaining the greatest
possible representativeness of an area with the smallest number of samples (Brus, 2019). Thus,
technological advances and the possibility of obtaining covariates such as those related to the DEM
(digital elevation model) and climate, among others, have allowed the use of new sampling and point
selection techniques such as the Latin hypercube. conditional, which also allows sampling points to
be selected at the lowest cost (Yang et al., 2020), is currently one of the preferred design methods.
On the other hand, soil fertility is important for agricultural and forestry production and plays a
fundamental role in food security. Soil fertility is divided into three main components: Physical,
Chemical and Biological, the interaction between the three being essential to ensure the quality and
sustainability of the crops (Bünemann et al., 2018). Of all these attributes, the easiest to diagnose by
traditional methods are the physical and chemical ones, the former being the least variable in time
and space and commonly used in many cases as covariates for predicting chemical attributes (Di
Raimo et al., 2022).
Chemical attributes are highly variable in space, especially micronutrients or heavy elements, which
increases the probability of misdiagnosis in an area with high variability, as is the case of tropical
soils (Macedo Neto et al., 2020). In this way, it is possible to observe nutritional deficiencies in areas
where fertilizers or amendments were applied due to the low dose suggested using traditional soil
fertility diagnostic methods.
In this way, computer systems have evolved quite a bit, allowing the use of powerful statistical
techniques such as machine learning, which base and adjust their predictions from models based on
experience or, in this case, from data, one of its applications being agricultural sciences and soil
science (Wadoux et al., 2020). Machine learning algorithms are diverse, and to date, more than 300
different models have been registered that can be adjusted and/or adapted to agricultural sciences
to be used for the prediction of relationships between different climatic conditions, topography,
management and soil fertility, with quite promising results (Dharumarajan et al., 2022; Wadoux et al.,
2020).
Some cases that use this type of technique for predicting soil fertility can be found in India
(Dharumarajan et al., 2022), Europe (Lu et al., 2023), Africa (Hounkpatin et al., 2022), Brazil (Vieira et
al., 2021) and even for the prediction of textural classes in Antarctic soils (Siqueira et al., 2023) in all
Arévalo-Hernández et al. 3
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
cases with high prediction indices, which allowed having an idea of the spatial variability of soil
characteristics and how these are related to the landscape and environment. In this way, the present
research aimed to predict soil fertility in the province of Alto Amazonas with the use of satellite
images and machine learning techniques.
2. MATERIALS AND METHODS
2.1. Localization
The research will be carried out in the Province of Alto Amazonas, which has an Am type climate
(Köppen, 1931). This region of the country has average annual maximums and minimums of 31.7°C
and 21.8°C, respectively. The average annual accumulated precipitation is 2086.2 mm.
2.2. Imaging processing
Images will be obtained freely from the United States Geological Service (USGS), considering the
date of collection of soil samples. Images from satellite Sentinel-2 will be selected, taking into
consideration the cartographic base available from the company that was visualized in Google Earth.
Sentinel-2 has a regular multispectral camera with 13 bands in the spectrum's visible, near-infrared
and short-wave infrared parts with main applications such as agriculture, land ecosystems, forest
management, and others. To improve values obtained from the satellite, calibrations and conversion
will be performed to suppress the effect of atmospheric gases.
2.3. Vegetation Indexes
Proposed Vegetation Indexes (VI) are presented in Table 1 and are based on previous work executed
for oil palm (Oliveira Teixeira, 2022). These VI will be calculated to predict their relationship with Alto
Amazonas – Loreto soil characteristics.
Table 1.
Proposed VI for predicting soil fertility in Alto Amazonas - Loreto
Index
Equation
Atmospherically Resistant Vegetation
Index (ARVI) (Kaufman & Tanré, 1992)
Difference Vegetation Index (DVI)
(Tucker, 1980)
Green Chlorophyll Index (GCI) (Gitelson
et al., 2003)
Green Difference Vegetation Index
(GDVI) (Sripada et al., 2006)
Leaf Area Index (LAI) (Boegh et al., 2002)
Normalized Difference Vegetation Index
(NDVI) (Rouse et al., 1974)
Optimized Soil Adjusted Vegetation
Index (OSAVI) (Rondeaux et al., 1996)
Soil Adjusted Vegetation Index (SAVI)
(Huete, 1988)
4 Revista Peruana de Investigación Agropecuaria
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
Specific Leaf Area Vegetation Index
(SLAVI) (Lymburner et al., 2000)
Simple Ratio or Ration Vegetation Index
(SR) (Jordan, 1969)
Triangular Vegetation Index (TVI) (Broge
& Leblanc, 2001)
Visble Atmospherically Resistant Index
(VARI) (Gitelson et al., 2003)
Vegetativen (VEG)
2.4. Soil sampling and analysis
Soil sampling was performed at 0-20 cm depth in different regions of Alto Amazonas province to
obtain the maximum information from each site, totalling 100 soil samples.
The physical and chemical properties of the soil, such as pH, E.C., organic matter, textural fractions
(sand, clay, and lime), exchangeable bases (Ca, Mg, Na, and K), exchangeable acidity, available P, and
CEC were determined before and after liming the soil. The chemical methods used for assessing soil
characteristics are reported in previous publications (Arévalo-Hernández et al., 2022). Soil texture
analysis was performed with the Bouyoucos method, using 1 M L−1 of NaOH as a dispersant. Soil
pH (1:2.5 H2O) was measured with a potentiometer, electrical conductivity (EC) with a conductivity
meter, and organic matter (OM) concentration with the Walkey and Black method by titration. CEC
and base cations (Ca2+, Mg2+, Na+, K+) were determined using extraction with 1 M NH4OAc and,
after, determined in flame atomic absorption spectrophotometer—FAAS. Yuan's method was used
to determine exchangeable acidity (Al3+, H+). Available P was extracted with the Olsen method (0.5
M NaHCO3 pH 8.5) and determined in a UV-VIS spectrophotometer. The selected microelements
(Cu, Fe, Mn, Zn) were extracted by DTPA and then analyzed by AAS.
2.5. Modelling and statistical analyses
For modelling, data will be divided into training and testing data. For the training phase, 75% of data
from farms will be used, while the remaining 25% will be for prediction assessment (testing data).
Descriptive statistics (minimum, quartile-1, mean, median, quartile-3, standard deviation, maximum,
interquartile range and coefficient of variation) will be performed for all soil characteristics and
vegetation indexes (VI). The data and descriptive statistics will be used for modelling procedures;
however, to avoid high correlated variables (r>0.90), Spearman correlation (5% confidence) will be
performed, and only low correlated variables will be stored.
Afterwards, data will be submitted to six models as follows: Cubist (C), General Linear Model (GLM),
Random Forest (RF), Weighted K-Nearest Neighbor Classifier (KKNN), Support Vector Machine (SVM)
and Neural Network (NN). The root mean square error (RMSE) and the coefficient of determination
(R2) will be calculated for training and validation to compare and select the best algorithm with the
higher R2 and lower RMSE. From the results of the models, an important analysis will be performed
to select the most influential variables to produce the final prediction model. All modelling
procedures and statistical analyses will be performed in R, version 4.1.2 (R Core Team, 2021).
Arévalo-Hernández et al. 5
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
3. RESULTS AND DISCUSSION
3.1. Soil physical and chemical characteristics
Table 2 presents the descriptive statistics of soil physical and chemical characteristics. In the case of
physical characteristics, texture indicated that mean values of sand were higher in comparison to silt
or clay, indicating a predominance of clay loam to sandy loam soils, representing 74.7 % of the
sampled soils.
Table 2.
Soil chemical and physical attributes mean, median, Interquartile Range and Range in Alto Amazonas
province
Soil variables
Mean (SD)
Median (IQR)
Range
pH
4.8 (0.9)
4.6 (1.0)
3.2 - 7.9
CE dS/cm
0.2 (0.4)
0.1 (0.1)
0.0 - 2.0
CaCO3 %
0.0 (0.3)
0.0 (0.0)
0.0 - 2.6
Organic Matter %
2.7 (4.3)
1.8 (2.0)
0.0 - 31.9
P mg/kg
5.8 (7.6)
3.7 (4.0)
0.0 - 54.9
Sand %
48.3 (20.4)
47.7 (31.6)
7.0 - 88.0
Silt %
25.9 (11.2)
25.5 (16.5)
5.3 - 52.0
Clay %
25.9 (13.6)
25.4 (17.2)
2.0 - 57.0
CEC cmol+/kg
10.3 (9.5)
7.5 (9.3)
1.1 - 42.8
Ca cmol+/kg
5.2 (8.0)
1.0 (5.8)
0.0 - 38.0
Mg cmol+/kg
0.7 (0.9)
0.3 (1.0)
0.0 - 3.5
K cmol+/kg
0.1 (0.1)
0.1 (0.1)
0.0 - 0.5
Na cmol+/kg
0.1 (0.1)
0.1 (0.0)
0.1 - 0.4
Al cmol+/kg
2.8 (2.9)
2.0 (3.4)
0.0 - 15.2
For chemical characteristics, the mean and median value of pH in the soil (4.8) was acidic. However,
some places showed high pH values (7.9), indicating that a great region area may require lime
amendments to achieve better yields. In the case of CE (dS/cm), all the values remain low, with no
saline soils observed in this region. In the case of Carbonates, low to zero values were observed,
being the mean and median near zero. For organic matter (%), mean values were observed in the
medium range and the mean slightly below the critical limit (2%); however, due to wetlands, higher
values were also observed in the magnitude of 31.9%. In the case of nutrients, P mean and median
values were low (<7 mg kg-1), indicating the high need for P fertilizers for crop production.
For exchangeable bases (Ca, Mg, K and Na), Ca had mean and median with very different values
indicating a non-normal distribution of data, while mean values were in the range considered as
medium (3-6 cmol+/kg). For Mg, mean values were low (<1.0 cmol+/kg), indicating the need to
apply high Mg amendments. Finally, for K and Na, mean values were very low (<0.1 cmol+/kg),
indicating the need for high K fertilizers.
Finally, in the case of exchangeable acidity (Al), mean values were high (>2.5 cmol+/kg), indicating
the need for the use of acidic tolerant species or lime amendments to reduce Al toxicity in crop
production in this province.
6 Revista Peruana de Investigación Agropecuaria
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
3.2. Prediction models
The results of the different machine learning models applied to the study for the prediction of soil
fertility in Alto Amazonas province in the Loreto region are presented in Table 3.
Table 3.
Prediction models (GLM, CUBIST, KKNN, SVM, RF and NN) R2, RMSE and MAE for main soil physical and
chemical characteristics in Alto Amazonas province
Soil variables
GLM*
CUBIST
KKNN
SVM
RF
NN
R2
rmse
MAE
R2
rmse
MAE
R2
rmse
MAE
R2
rmse
MAE
R2
rmse
MAE
R2
rmse
MAE
pH
0.21
0.99
0.66
0.60
0.73
0.49
0.18
0.96
0.64
0.01
1.11
0.74
0.71
0.73
0.49
0.01
3.82
2.55
Organic Matter
0.12
2.54
1.69
0.23
1.71
1.14
<0.01
2.08
1.39
0.12
3.06
2.04
0.19
2.83
1.89
0.20
1.77
1.18
P mg/kg
0.05
9.51
6.34
0.01
6.20
4.13
0.02
4.56
3.04
0.07
5.67
3.78
0.03
4.49
2.99
0.09
4.91
3.27
Ca cmol+/kg
0.46
6.12
4.08
0.06
9.31
6.21
0.52
4.91
3.27
0.42
6.61
4.41
0.75
3.53
2.35
0.28
8.20
5.47
Mg cmol+/kg
0.35
0.62
0.41
0.48
0.52
0.35
0.42
0.56
0.37
0.37
0.61
0.41
0.57
0.48
0.32
0.47
0.56
0.37
K cmol+/kg
0.27
35.76
23.84
0.15
33.37
22.25
0.07
32.76
21.84
0.21
38.71
25.81
0.29
27.94
18.63
0.10
60.10
40.07
Al cmol+/kg
0.47
3.36
2.24
0.66
2.26
1.51
0.09
3.45
2.30
0.44
3.11
2.07
0.34
3.02
2.01
0.05
4.20
2.80
CEC cmol+/kg
0.91
1.88
1.25
0.65
4.29
2.86
0.78
3.15
2.10
0.89
2.19
1.46
0.92
2.32
1.55
0.01
10.49
6.99
* Cubist (C), General Linear Model (GLM), Random Forest (RF), Weighted K-Nearest Neighbor Classifier (KKNN), Support
Vector Machine (SVM) and Neural Network (NN)
It was possible to observe that, in general, to predict soil variables, the Random Forest algorithm was
satisfactory compared to other models, obtaining the higher R2 and lower RMSE in the Alto
Amazonas province. However, in the case of organic matter and Aluminum, the CUBIST algorithm
was slightly superior.
Even though the use of these algorithms in the prediction of soil variables is not new, some
algorithms have performed better than others, such as random forest, since it has great capacity in
the use of nonlinear data and is robust against possible errors (Breinman, 2001). Also, Smith et al.
(2020) have obtained better performance with random forest in soil texture prediction in agricultural
soils. In the case of Organic Matter, Mosaid et al. (2024) showed that the use of random forest
performed better, as shown in the present study.
Even though the perspectives of using machine learning have been discussed elsewhere by Sujatha
et al. (2023), it remains an interesting tool to have results in areas where the logistics and costs are
often expensive, improving the information and decision-making.
CONCLUSIONS
Alto Amazonas province has high climatic and geologic differences, generating different types of
soils and formations; also, access to many areas is limited. The use of machine learning algorithms
to predict poses as an alternative to improve information on soil fertility. In this work, we observed
that most soils in the province have low pH, P, Mg, K and high acidity. We also managed to achieve
good predictions for pH, Ca, Mg and CEC, and we observed that the most successful algorithm was
Random Forest. Nevertheless, for Al, Cubist performed better. This is one of the first works using
machine learning to predict soil fertility in the Peruvian Amazon, and we hope it may serve as a base
for future projects.
Arévalo-Hernández et al. 7
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
FUNDING
The authors thank the Universidad Nacional Autónoma de Alto Amazonas for the funding within the
scope of the call for Research Projects 2023 with Presidential Resolution N°245-UNAAA/P. We also
appreciate the technical and logistical facilities for field research and laboratory analytical facilities.
CONFLICT OF INTEREST
Los autores declaran que no existe ningún tipo de conflicto de intereses.
AUTHORSHIP CONTRIBUTION
Conceptualization: Arévalo-Hernández, C.O. y Arévalo-Gardini, E.
Formal analysis: Arévalo-López, L. A. y Tuesta-Hidalgo, O.
Research: All authors
Methodology: Arévalo-Hernández, C.O., Arévalo-Gardini, E. y Arévalo-López, L. A.
Project management: Romero-Vela, D.S.
Supervision: Arévalo-Gardini, E. y Ruiz-Camus, C.E.
Validation: Arévalo-López, L. A. y Tuesta-Hidalgo, O.
Visualization: Romero-Vela, D.S.
Writing - original draft: All authors
Writing - proofreading and editing: All authors
REFERENCES
Arévalo-Gardini, E., Canto, M., Alegre, J., Loli, O., Julca, A., & Baligar, V. (2015). Changes in soil
physical and chemical properties in long term improved natural and traditional agroforestry
management systems of cacao genotypes in Peruvian Amazon.
PLoS ONE
, 10(7).
https://doi.org/10.1371/journal.pone.0132147
Bendig, J., Yu, K., Aasen, H., Bolten, A., Bennertz, S., Broscheit, J., Gnyp, M. L., & Bareth, G. (2015).
Combining UAV-based plant height from crop surface models, visible, and near infrared
vegetation indices for biomass monitoring in barley.
International Journal of Applied Earth
Observation and Geoinformation
, 39, 79–87. https://doi.org/10.1016/j.jag.2015.02.012
Brus, D. J. (2019). Sampling for digital soil mapping: A tutorial supported by R scripts.
Geoderma
,
338, 464–480. https://doi.org/10.1016/J.GEODERMA.2018.07.036
Bünemann, E. K., Bongiorno, G., Bai, Z., Creamer, R. E., De Deyn, G., de Goede, R., Fleskens, L.,
Geissen, V., Kuyper, T. W., Mäder, P., Pulleman, M., Sukkel, W., van Groenigen, J. W., &
Brussaard, L. (2018). Soil quality – A critical review. In Soil Biology and Biochemistry (Vol.
120, pp. 105–125). https://doi.org/10.1016/j.soilbio.2018.01.030
Canero, F. M., Rodriguez-Galiano, V., & Aragones, D. (2024). Machine Learning and Feature
Selection for soil spectroscopy. An evaluation of Random Forest wrappers to predict soil
8 Revista Peruana de Investigación Agropecuaria
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
organic matter, clay, and carbonates.
Heliyon
,
10
(9).
https://doi.org/10.1016/j.heliyon.2024.e30228
Campos, A. R., Giasson, E., Costa, J. J. F., Machado, I. R., Silva, E. B. da, & Bonfatti, B. R. (2019).
Selection of Environmental Covariates for Classifier Training Applied in Digital Soil Mapping.
Revista Brasileira de Ciência Do Solo
, 42, e0170414.
https://doi.org/10.1590/18069657RBCS20170414
Chen, J. M. (1996). Evaluation of vegetation indices and a modified simple ratio for boreal
applications.
Canadian Journal of Remote Sensing
, 22(3), 229–242.
https://doi.org/10.1080/07038992.1996.10855178
Chuvieco, E., Martín, M. P., & Palacios, A. (2002). Assessment of different spectral indices in the red-
near-infrared spectral domain for burned land discrimination.
International Journal of
Remote Sensing
, 23(23), 5103–5110. https://doi.org/10.1080/01431160210153129
Dash, J., & Curran, P. J. (2007). Evaluation of the MERIS terrestrial chlorophyll index (MTCI).
Advances in Space Research, 39(1), 100–104. https://doi.org/10.1016/j.asr.2006.02.034
Delgado-Caballero, C. E., Gómez-Guerrero, A., Valdez-Lazalde, J. R., De los Santos- Posadas, H.,
Fierros-González, A. M., & Horwath, W. R. (2009). Site index and soil properties in young
plantations of Eucalyptus grandis and E. urophylla in southeastern México.
Agrociencia
,
43(1). https://www.agrociencia-colpos.org/index.php/agrociencia/article/view/697
Dharumarajan, S., Lalitha, M., Niranjana, K., & Hegde, R. (2022). Evaluation of digital soil mapping
approach for predicting soil fertility parameters—a case study from Karnataka Plateau,
India.
Arabian Journal of Geosciences
, 15(5), 1–21. https://doi.org/10.1007/S12517-022-
09629-8
Di Raimo, L. A. D. L., Couto, E. G., de Mello, D. C., Demattê, J. A. M., Amorim, R. S. S., Torres, G. N.,
Bocuti, E. D., Veloso, G. V., Poppiel, R. R., Francelino, M. R., & Fernandes-Filho, E. I. (2022).
Characterizing and Modeling Tropical Sandy Soils through VisNIR-SWIR, MIR Spectroscopy,
and X-ray Fluorescence. Remote Sensing 2022,14(19), 4823.
https://doi.org/10.3390/RS14194823
EMBRAPA. (2009). Manual de análises químicas de solos, plantas e fertilizantes (2nd ed.). Embrapa.
www.sct.embrapa.br/liv
Frampton, W. J., Dash, J., Watmough, G., & Milton, E. J. (2013). Evaluating the capabilities of
Sentinel-2 for quantitative estimation of biophysical variables in vegetation.
ISPRS Journal of
Photogrammetry and Remote Sensing
, 82, 83–92.
https://doi.org/10.1016/j.isprsjprs.2013.04.007
Gamon, J. A., & Surfus, J. S. (1999). Assessing leaf pigment content and activity with a reflectometer.
New Phytologist, 143(1), 105–117. https://doi.org/10.1046/j.1469-8137.1999.00424.x
Gitelson, A. A., Gritz, Y., & Merzlyak, M. N. (2003). Relationships between leaf chlorophyll content
and spectral reflectance and algorithms for non-destructive chlorophyll assessment in
Arévalo-Hernández et al. 9
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
higher plant leaves. Journal of Plant Physiology, 160(3), 271–282.
https://doi.org/10.1078/0176-1617-00887
Gitelson, A. A., Merzlyak, M. N., & Chivkunova, O. B. (2001). Optical Properties and Nondestructive
Estimation of Anthocyanin Content in Plant Leaves.
Photochemistry and Photobiology
,
74(1), 38–45. https://doi.org/10.1562/0031-8655(2001)0740038opaneo2.0.co2
Hounkpatin, K. O. L., Bossa, A. Y., Yira, Y., Igue, M. A., & Sinsin, B. A. (2022). Assessment of the soil
fertility status in Benin (West Africa) – Digital soil mapping using machine learning.
Geoderma Regional
, 28, e00444. https://doi.org/10.1016/J.GEODRS.2021.E00444
Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment, 25(3),
295–309. https://doi.org/10.1016/0034-4257(88)90106-X INEI. (2012). IV Censo Nacional
Agropecuario . In Resultados Definitivos. IV Censo Nacional Agropecuario.
http://proyectos.inei.gob.pe/web/documentospublicos/resultadosfinalesivcenagro.pdf
Jordan, C. F. (1969). Derivation of Leaf-Area Index from Quality of Light on the Forest Floor.
Ecology
, 50(4), 663–666. https://doi.org/10.2307/1936256
Köppen, W. (1931). Grundriß der Klimakunde. In Grundriß der Klimakunde. De Gruyter.
https://doi.org/10.1515/9783111667751
Lang, P., Qin, C. zhi, Zhu, A. xing, Hou, Z. wei, Fan, N. qing, & Wang, Y. jie. (2020). A case- based
method of selecting covariates for digital soil mapping.
Journal of Integrative Agriculture
,
19(8), 2127–2136. https://doi.org/10.1016/S2095-3119(19)62857-1
Louhaichi, M., Borman, M. M., & Johnson, D. E. (2001). Spatially located platform and aerial
photography for documentation of grazing impacts on wheat.
Geocarto International
, 16(1),
65–70. https://doi.org/10.1080/10106040108542184
Lu, Q., Tian, S., & Wei, L. (2023). Digital mapping of soil pH and carbonates at the European scale
using environmental variables and machine learning.
Science of The Total Environment
, 856,
159171. https://doi.org/10.1016/J.SCITOTENV.2022.159171
Lymburner, L., Beggs, P. J., & Jacobson, C. R. (2000). Estimation of canopy-average surface-specific
leaf area using Landsat TM data. In Photogrammetric Engineering and Remote Sensing (Vol.
66, Issue 2, pp. 183–191).
Macedo Neto, A. A. L.; Farias, P. R. S.; Matos, G. S. B.; Silva, G. B.; Santos, A. V. F. Anhê, B. B. (2020).
Diagnosis and spatial variability of soil fertility and crop production in a teak area in Eastern
Pará State.
CERNE
, 26(1), 37-47. https://doi.org/10.1590/01047760202026012683
Mosaid, H., Barakat, A., John, K., Faouzi, E.H., Bustillo, V., El Garnaoui, M., & Heung, B. (2024).
Improved soil carbon stock spatial prediction in a Mediterranean soil erosion site through
robust machine learning techniques.
Environmental monitoring and assessment, 196 2
, 130.
https://doi.org/10.1007/s10661-024-12294-x
10 Revista Peruana de Investigación Agropecuaria
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
Qi, J., Chehbouni, A., Huete, A. R., Kerr, Y. H., & Sorooshian, S. (1994). A modified soil adjusted
vegetation index.
Remote Sensing of Environment
, 48(2), 119–126.
https://doi.org/10.1016/0034-4257(94)90134-1
R Core Team. (2021). R Core Team (2021). R: A language and environment for statistical computing.
R Foundation for Statistical Computing, Vienna, Austria. URL Http://Www.R-Project.Org/
Richardson, A. J., & Wiegand, C. L. (1977). Distinguishing Vegetation from Soil Background
Information. A gray mapping technique allows delineation of any Landsat scene into
vegetative cover stages, degrees of soil brightness, and water.
Photogrammetric
Engineering and Remote Sensing
, 43(13), 1541-1552. https://www.asprs.org/wp-
content/uploads/pers/1977journal/dec/1977_dec_1541-1552.pdf
Rondeaux, G., Steven, M., & Baret, F. (1996). Optimization of soil-adjusted vegetation indices.
Remote Sensing of Environment
, 55(2), 95–107. https://doi.org/10.1016/0034-
4257(95)00186-7
Rouse, J. W., Haas, R. H., Schell, J. A., & Deering, D. W. (1974). Monitoring vegetation systems in the
great plains with ERTS. NASA Goddard Space Flight Center 3d ERTS-1 Symposium, 1, 309–
317. https://doi.org/10.1021/jf60203a024
Roy, D. P., Boschetti, L., & Trigg, S. N. (2006). Remote sensing of fire severity: Assessing the
performance of the normalized burn ratio.
IEEE Geoscience and Remote Sensing Letters
,
3(1), 112–116. https://doi.org/10.1109/LGRS.2005.858485
Siqueira, R. G., Moquedace, C. M., Francelino, M. R., Schaefer, C. E. G. R., & Fernandes- Filho, E. I.
(2023). Machine learning applied for Antarctic soil mapping: Spatial prediction of soil
texture for Maritime Antarctica and Northern Antarctic Peninsula.
Geoderma
, 432, 116405.
https://doi.org/10.1016/J.GEODERMA.2023.116405
Sripada, R. P., Heiniger, R. W., White, J. G., & Meijer, A. D. (2006). Aerial color infrared photography
for determining early in-season nitrogen requirements in corn.
Agronomy Journal
, 98(4),
968–977. https://doi.org/10.2134/agronj2005.0200
Sujatha, M., & Jaidhar, C. D. (2023). Machine learning-based approaches to enhance the soil
fertility—A review.
Expert Systems with Applications
, 240, 122557.
https://doi.org/10.1016/j.eswa.2023.122557
Tucker, C. J. (1980). A spectral method for determining the percentage of green herbage material in
clipped samples.
Remote Sensing of Environment
, 9(2), 175–181.
https://doi.org/10.1016/0034-4257(80)90007-3
Vescovo, L., & Gianelle, D. (2008). Using the MIR bands in vegetation indices for the estimation of
grassland biophysical parameters from satellite remote sensing in the Alps region of
Trentino (Italy).
Advances in Space Research
, 41(11), 1764–1772.
https://doi.org/10.1016/j.asr.2007.07.043
Vieira, R. M. da S. P., Tomasella, J., Barbosa, A. A., Polizel, S. P., Ometto, J. P. H. B., Santos, F. C.,
Ferreira, Y. da C., & Toledo, P. M. de. (2021). Land degradation mapping in the MATOPIBA
Arévalo-Hernández et al. 11
Rev. Peru. Investig. Agropecu. 3(2): e63; (Jul-Dic, 2024). e-ISSN: 2955-831X
region (Brazil) using remote sensing data and decision-tree analysis.
Science of The Total
Environment
, 782, 146900. https://doi.org/10.1016/J.SCITOTENV.2021.146900
Wadoux, A. M. J. C., Minasny, B., & McBratney, A. B. (2020). Machine learning for digital soil
mapping: Applications, challenges and suggested solutions.
Earth-Science Reviews
, 210,
103359. https://doi.org/10.1016/J.EARSCIREV.2020.103359
Wang, F., Huang, J., Tang, Y., & Wang, X. (2007). New Vegetation Index and Its Application in
Estimating Leaf Area Index of Rice.
Rice Science
, 14(3), 195–203.
https://doi.org/10.1016/s1672-6308(07)60027-4
Watt, M. S., Pearse, G. D., Dash, J. P., Melia, N., & Leonardo, E. M. C. (2019). Application of remote
sensing technologies to identify impacts of nutritional deficiencies on forests.
ISPRS Journal
of Photogrammetry and Remote Sensing
, 149, 226–241.
https://doi.org/10.1016/j.isprsjprs.2019.01.009
Wu, W. (2014). The Generalized Difference Vegetation Index (GDVI) for dryland characterization.
Remote Sensing
, 6(2), 1211–1233. https://doi.org/10.3390/rs6021211
Yang, L., Li, X., Shi, J., Shen, F., Qi, F., Gao, B., & Chen, Z. (2020). Geoderma Evaluation of conditioned
Latin hypercube sampling for soil mapping based on a machine learning method.
Geoderma
, 369, 114337. https://doi.org/10.1016/j.geoderma.2020.114337
