Assessing Soil and Land Suitability of an Olive–Maize Agroforestry System Using Machine Learning Algorithms

Authors: ASIF, HAYAT - National University Of Sciences And Technology; JAVED, IQBAL - National University Of Sciences And Technology; Ashworth, Amanda; Owens, Phillip

Submitted to: Crops
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/14/2024
Publication Date: 7/9/2024
Citation: Asif, H., Javed, I., Ashworth, A.J., Owens, P.R. 2024. Assessing Soil and Land Suitability of an Olive–Maize Agroforestry System Using Machine Learning Algorithms. Crops. 4(3):308-323. https://doi.org/10.3390/crops4030022.
DOI: https://doi.org/10.3390/crops4030022

Interpretive Summary: Population growth is having consequences on the environment and natural resources, particularly in Pakistan. Agroforestry “is a collective name for land-use systems and practices in which woody perennials are deliberately integrated with crops and/or animals on the same land-management unit.” Agroforestry practices attains more sustainable land use and can enhance optimize production per unit area and improve the livelihood of rural economies. Machine learning models can also strategically combine data from multiple sources and identify relationships by learning from massive datasets. Scientist from USDA-ARS and National University of Sciences and Technology in Islamabad Pakistan set out to identify suitable land classes for an integrated olive and maize agroforestry system. Information about soil physical and chemical properties were obtained from 701 soil samples, along with climatic and topographic data. After determination of land suitability classes for this integrated olive and maize crop agroforestry system, it was then mapped using three machine learning algorithms. The land suitability classes predicted through various methods varied greatly. Major crop suitability limitations of the study region included high elevation, slope, pH, and large gravel content. Climate change and shifting agricultural land use threaten agriculture production and food security. Agroforestry systems can enhance land productivity, while reducing climate change effects such as improved soil moisture and reduced below canopy temperature. Production potential from rainfed agriculture is low but this dual-use cropping system may increase productivity and improve the livelihood of farmers. Future land suitability assessments can be used to analyze suitable areas for various agroforestry systems in different regions of Pakistan according to the suitability of climate, soil, and topography.

Technical Abstract: Exponential population increases are threating food security, particularly in mountainous areas. One potential solution to this is dual-use cropping systems. Olive (Olea europaea) and maize (Zea mays) in an agroforestry-cropping system may mitigate risk through providing multiple market sources (oil and grain crop) for small holder producers. Several studies have conducted integrated agroforestry land suitability analyses, however few studies used machine learning (ML) algorithms to evaluate multiple variables for the selection of suitable rainfed sites under mountainous terrain. This study aims to identify suitable land classes for an integrated olive and maize agroforestry system based on FAO's “land suitability assessment Framework” for 1,757 km² in Khyber Pakh-tunkhwa province, Pakistan. Information about soil physical and chemical properties were obtained from 701 soil samples, along with climatic and topographic data. After determination of land suitability classes for this integrated olive and maize crop agroforestry system and then mapped through ML algorithm using random forest (RF), support vector machine (SVM), and traditional techniques of weighted overlay (WOL). The land suitability classes predicted by the two different techniques show a remarkable difference. For example, the area of the S1 class (optimum suitability) classified through RF were (˜9%') than that of SVM, and less (8%') than the class classified through WOL. The area of S2 class classified through RF was (18%') than that of SWM and was (17%') than the area classified through WOL, similarly, the S3 class area classified by RF was (27%') than that of SVM, and (45%') than the area classified through WOL. Conversely, the area of N2 area classified through RF and SVM was (6%') than the area classified through WOL. Model performance was assessed through the overall accuracy and Kappa Index and indicated the RF performed better than the SVM and WOL. Major crop suitability limitations of the study area include high elevation, slope, pH, and large gravel content.