Assessing the performance of decision tree and neural network models in mapping soil properties

Hateffard, Fatemeh; Dolati, Payam; Heidari, Ahmad; Zolfaghari, Ali Asghar

doi:10.1007/s11629-019-5409-8

Cited by 30 publications

(20 citation statements)

References 72 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Model testing was based on 15% of the data while 15% was used for model validation. As the performance of the DT is better than ANN (Appendix 1 and 2 ), the output of the DT model was adopted as the main input for calculating the K factor 59 .…”

Section: Methodsmentioning

confidence: 99%

CMIP5 climate projections and RUSLE-based soil erosion assessment in the central part of Iran

Hateffard

Mohammed

Alsafadi

et al. 2021

Sci Rep

Self Cite

View full text Add to dashboard Cite

Soil erosion (SE) and climate change are closely related to environmental challenges that influence human wellbeing. However, the potential impacts of both processes in semi-arid areas are difficult to be predicted because of atmospheric variations and non-sustainable land use management. Thus, models can be employed to estimate the potential effects of different climatic scenarios on environmental and human interactions. In this research, we present a novel study where changes in soil erosion by water in the central part of Iran under current and future climate scenarios are analyzed using the Climate Model Intercomparison Project-5 (CMIP5) under three Representative Concentration Pathway-RCP 2.6, 4.5 and 8.5 scenarios. Results showed that the estimated annual rate of SE in the study area in 2005, 2010, 2015 and 2019 averaged approximately 12.8 t ha−1 y−1. The rangeland areas registered the highest soil erosion values, especially in RCP2.6 and RCP8.5 for 2070 with overall values of 4.25 t ha−1 y−1 and 4.1 t ha−1 y−1, respectively. They were followed by agriculture fields with 1.31 t ha−1 y−1 and 1.33 t ha−1 y−1. The lowest results were located in the residential areas with 0.61 t ha−1 y−1 and 0.63 t ha−1 y−1 in RCP2.6 and RCP8.5 for 2070, respectively. In contrast, RCP4.5 showed that the total soil erosion could experience a decrease in rangelands by − 0.24 t ha−1 y−1 (2050), and − 0.18 t ha−1 y−1 (2070) or a slight increase in the other land uses. We conclude that this study provides new insights for policymakers and stakeholders to develop appropriate strategies to achieve sustainable land resources planning in semi-arid areas that could be affected by future and unforeseen climate change scenarios.

show abstract

Section: Methodsmentioning

confidence: 99%

CMIP5 climate projections and RUSLE-based soil erosion assessment in the central part of Iran

Hateffard

Mohammed

Alsafadi

et al. 2021

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…Using Normalized Difference Salinity Index to measure soil salinity, [30] selected and evaluated several indices that had the highest response to salinity, based on the reports from various studies conducted on plants and mineral salts. The images used in their study were Landsat satellite images orbiting the Earth once every 99 minutes.…”

Section: Introductionmentioning

confidence: 99%

Monitoring Soil Salinity Changes, Comparison of Different Maps and Indices Extracted from Landsat Satellite Images (Case Study: Atabieh, Khuzestan)

Golabkesh¹,

Ghanavati²,

Nazarpour³

et al. 2021

Pol. J. Environ. Stud.

View full text Add to dashboard Cite

Soil salinity is the dominant process in the degradation of arid and semi-arid soils, which in turn reduces crop yields, increases erosion, and exacerbates desertification. In recent years, soil salinity has affected much of the land in the Atabieh area located in the west of Khuzestan province in Iran. The purpose of this study was thus to evaluate and map soil salinity changes in the region over 15 years using Landsat 7 and 8 satellite images. To that end, the spectra of saline soils in the study area were extracted from the satellite data, and after the initial pre-processing in EVNI software version 5.3, the SI1, SI2, SI3, BI, NDVI, and NDSI indices were prepared. Using the supervised classification method, the salinity map with four different classes was then plotted in Arc GIS version 10.2, and the changes in saline soil area were investigated. Moreover, field surveys, surface soil sampling, soil EC measurement and identification of available minerals were performed by X-ray diffraction (XRD) technique and satellite images. Among the studied indices, the BI index with the highest correlation (0.71) was considered as the best index, and NDVI with the correlation coefficient of 0.35 at the 95% confidence level, was the best index for vegetation cover. Examination of changes in BI index by Landsat 7 images showed that the non-saline land area decreased from 1023.54 ha in 2000 to 143.43 ha in 2010, while the area with medium salinity increased by 14.57%. Besides, the salinity severity in the NDVI index had a growth rate of 72.86%. In turn, XRD studies confirmed the presence of abundant evaporate minerals (Halites, Calcite and Dolomite) corresponding to the values (real numbers) of salinity and mineralogical maps obtained from the Landsat 8 images.

show abstract

“…Consequently, ML algorithms such as Decision Tree (DT), Random Forest (RF), Support Vector Machines (SVM), Artificial Neural Network (ANN), and Discriminant Analysis (DA) are still widely used [35][36][37]. Owing to their relatively high speed and accuracy, together with the ability to accommodate non-linearity and multicollinearity [10,38,39], ML methods have become popular in soil erosion studies over the past few years [25,[40][41][42][43]. Among these ML methods, SVM and RF consistently showed better performance relative to other ML methods [44], but when compared among themselves (SVM and RF), it is still unclear which method can outperform the other.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning for Gully Feature Extraction Based on a Pan-Sharpened Multispectral Image: Multiclass vs. Binary Approach

Phinzi

Abriha

Bertalan

et al. 2020

IJGI

View full text Add to dashboard Cite

Gullies reduce both the quality and quantity of productive land, posing a serious threat to sustainable agriculture, hence, food security. Machine Learning (ML) algorithms are essential tools in the identification of gullies and can assist in strategic decision-making relevant to soil conservation. Nevertheless, accurate identification of gullies is a function of the selected ML algorithms, the image and number of classes used, i.e., binary (two classes) and multiclass. We applied Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and Random Forest (RF) on a Systeme Pour l’Observation de la Terre (SPOT-7) image to extract gullies and investigated whether the multiclass (m) approach can offer better classification accuracy than the binary (b) approach. Using repeated k-fold cross-validation, we generated 36 models. Our findings revealed that, of these models, both RFb (98.70%) and SVMm (98.01%) outperformed the LDA in terms of overall accuracy (OA). However, the LDAb (99.51%) recorded the highest producer’s accuracy (PA) but had low corresponding user’s accuracy (UA) with 18.5%. The binary approach was generally better than the multiclass approach; however, on class level, the multiclass approach outperformed the binary approach in gully identification. Despite low spectral resolution, the pan-sharpened SPOT-7 product successfully identified gullies. The proposed methodology is relatively simple, but practically sound, and can be used to monitor gullies within and beyond the study region.

show abstract

Assessing the performance of decision tree and neural network models in mapping soil properties

Cited by 30 publications

References 72 publications

CMIP5 climate projections and RUSLE-based soil erosion assessment in the central part of Iran

CMIP5 climate projections and RUSLE-based soil erosion assessment in the central part of Iran

Monitoring Soil Salinity Changes, Comparison of Different Maps and Indices Extracted from Landsat Satellite Images (Case Study: Atabieh, Khuzestan)

Machine Learning for Gully Feature Extraction Based on a Pan-Sharpened Multispectral Image: Multiclass vs. Binary Approach

Contact Info

Product

Resources

About