G. 2086

Page 1

Global Research journal of Natural Science

& Technology (GRJNST)

Volume: 04 - Issue 3 (2026), 2083

ISSN P: 2790-7643 ISSN E: 2790-7651

www.grjnst.net

https://doi.org/10.53762/grjnst.04.03.07

Integrating Artificial Intelligence, Remote Sensing, and GIS for Sustainable

Agro-Forestry Management and Land Resource Optimization

Received: 31 March 2026. Accepted: 29 April 2026. Published: 12 May, 2026

Muhammad Kashif Majeed (Corresponding Author)

Faculty of Engineering Science and Technology, Iqra University, Karachi, Pakistan.

Email: mkashif@iqra.edu.pk

Kashif Akbar

Department of Industrial Engineering, University of Padua, Italy

Email: kashif.akbar@studenti.unipd.it

Muhammad Hassan Ali

Forest and Range Management

Shaheed Benazir Bhutto University of Veterinary & Animal Sciences

Sakrand, Sindh, Pakistan

Email: soilscience1070@gmail.com

Muhammad Essa Siddique

PhD (IT) Scholar

Dr. A H S Bukhari Centre of ICT, Faculty of Engineering & Technology, University of Sindh, Jamshoro, Pakistan.

Email: Essasiddique@live.com

Murtaza Ali

Department of Horticulture, Sindh Agriculture University, Tandojam, Sindh, Pakistan.

Email: murtazal46@gmail.com

Gul Muhammad Shah

Department of Soil Science, Sindh Agriculture University, Tandojam, Sindh, Pakistan.

Email: syedgm7@gmail.com

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use and distribution

G. 2086

Page 2

Abstract: The increasing pressure on land resources due to climate change, population growth, and

unsustainable land-use practices necessitates advanced approaches for sustainable agro-forestry

management. This study proposes an integrated framework that combines Artificial Intelligence (AI),

Remote Sensing (RS), and Geographic Information Systems (GIS) to support land suitability analysis

and resource optimization. Multi-source geospatial data, including satellite imagery, environmental

variables, and topographic information, are processed through a structured pipeline involving pre-

processing, feature extraction, AI-based modelling, and GIS-driven spatial analysis. Machine learning

and deep learning models are employed to perform land use classification and predict agro-forestry

suitability, while multi-criteria decision analysis (MCDA) is used to integrate environmental factors

and generate suitability maps. The results indicate that a significant portion of the study area falls

within highly and moderately suitable categories, demonstrating strong potential for agro-forestry

development. Model evaluation using 5-fold cross-validation shows stable performance across all

models, with deep learning approaches providing slightly higher accuracy, while traditional models

remain computationally efficient. The integration of NDVI-based environmental analysis with spatial

modelling further validates the reliability of the framework. The resulting decision support outputs

enable the identification of priority zones and optimized land-use strategies. Overall, the proposed

framework offers a scalable and data-driven solution for sustainable land resource management and

supports informed decision-making for agro-forestry planning.

Keywords: Artificial Intelligence (AI), Remote Sensing (RS), Geographic Information Systems (GIS),

Agro-Forestry, Land Suitability Analysis, Land Resource Optimization, Land Use and Land Cover

(LULC), Spatial Analysis, Multi-Criteria Decision Analysis (MCDA), Machine Learning, Deep

Learning, NDVI, Environmental Monitoring, Decision Support Systems (DSS)

1. Introduction

1.1 Background and Motivation

The growing pressure on land resources has become one of the defining environmental challenges of

the 21st century. Rapid population growth, expanding agricultural demand, and increasing

urbanization are continuously reshaping natural landscapes, often at the expense of ecological stability.

At the same time, climate change is intensifying these pressures by altering rainfall patterns, increasing

temperatures, and amplifying the frequency of extreme weather events. These changes have profound

implications for land productivity, biodiversity, and long-term sustainability. In many regions,

traditional land management practices are no longer sufficient to cope with these complex and

interconnected challenges, creating a strong need for more adaptive, data-driven approaches to resource

management.

Within this broader context, agro-forestry has gained considerable attention as a sustainable land-use

strategy that integrates trees with crops and livestock systems. Unlike conventional monoculture-based

agriculture, agro-forestry systems promote ecological balance by enhancing soil fertility, improving

water retention, reducing erosion, and increasing carbon sequestration. These systems also offer socio-

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 3

economic benefits by diversifying income sources and improving resilience against climate variability.

However, the successful implementation of agro-forestry requires a detailed understanding of spatial

heterogeneity, environmental constraints, and land suitability conditions—factors that are difficult to

assess using traditional methods alone.

In recent years, geospatial technologies such as Remote Sensing (RS) and Geographic Information

Systems (GIS) have emerged as powerful tools for monitoring and managing environmental systems.

Remote sensing provides continuous, large-scale observations of the Earth’s surface through satellite

imagery, enabling the assessment of vegetation health, land-use dynamics, and environmental changes

over time. GIS, on the other hand, facilitates the integration and analysis of spatial data, allowing

researchers and practitioners to develop models for land-use planning and resource allocation.

Together, these technologies form the backbone of modern environmental monitoring systems and

have significantly improved the ability to analyze complex spatial phenomena [1], [2].

1.2 Role of Geospatial Technologies and AI Integration

The integration of Artificial Intelligence (AI) with geospatial technologies has further strengthened the

capability of environmental monitoring and land management systems. Machine learning and deep

learning algorithms enable automated extraction of patterns from large datasets, improving both

efficiency and accuracy in geospatial analysis. AI-driven models have shown strong performance in

land-use classification, vegetation monitoring, and environmental prediction tasks, often outperforming

traditional statistical approaches [3], [4]. To better understand the complementary roles of these

technologies, Table 1 provides a comparative overview of AI, Remote Sensing, and GIS in agro-forestry

applications.

Table 1: Comparative Role of AI, Remote Sensing, and GIS in Agro-Forestry Systems

Core

Application

in

Technology

Data Type

Strengths

Limitations

Capability

Agro-Forestry

Earth

observation

Sensing (RS) &

monitoring

Large-scale

coverage,

temporal

Satellite

imagery,

spectral data

Cloud

interference,

resolution limits

Vegetation health

(NDVI), land

cover mapping

Remote

monitoring

Multi-layer

analysis,

decision

Spatial

analysis

integration

Vector

& raster spatial

data

&

Requires quality Land suitability,

GIS

input data

zoning, planning

modelling

Data

Artificial

Intelligence

(AI)

Pattern

recognition

& prediction

Structured & High accuracy,

Classification,

prediction,

optimization

dependency,

computational

cost

unstructured

data

automation,

scalability

Intelligent

geospatial

High

Decision Support

Integrated

Multi-source

precision, real- computational

Systems

(DSS),

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 4

AI–RS–GIS

analytics

datasets

time insights

requirements

precision forestry

1.3 Global Challenges and Sustainable Agro-Forestry

The urgency of adopting sustainable agro-forestry practices is closely linked to several global

environmental challenges. Climate change remains a central concern, as it directly affects agricultural

productivity and ecosystem stability. Variations in temperature and precipitation patterns can disrupt

crop cycles, reduce yields, and increase vulnerability to pests and diseases. At the same time,

deforestation continues to contribute significantly to greenhouse gas emissions while reducing

biodiversity and ecosystem services [5].

Land degradation is another critical issue, affecting a large portion of the world’s land area.

Unsustainable agricultural practices, deforestation, and overexploitation of natural resources have led to

soil erosion, nutrient depletion, and desertification. These processes not only reduce land productivity

but also threaten food security. Agro-forestry systems, through their integrated approach, offer a viable

solution by improving soil quality, enhancing carbon sequestration, and restoring ecological balance [6].

Food security is intrinsically tied to these environmental issues. With increasing population demands,

there is a growing need to optimize land use without further degrading natural ecosystems. This

requires advanced tools that can analyze environmental conditions and support informed decision-

making.

1.4 Key Factors Influencing Agro-Forestry Suitability

Effective agro-forestry planning depends on a wide range of environmental, spatial, and socio-economic

factors. These factors must be carefully analyzed to determine the suitability of land for sustainable

practices. The integration of multi-source datasets allows for a more comprehensive evaluation of these

variables. Table 2 summarizes the key factors influencing agro-forestry suitability and their relevance in

land resource optimization.

Table 2: Key Environmental and Geospatial Factors Influencing Agro-Forestry Suitability

Factor

Importance

Forestry

in

Agro-

Parameters

Data Source

Category

Rainfall,

Humidity

Temperature, Meteorological

satellite

data, Determines

crop/tree

Climatic

Soil

growth patterns

Soil type, pH, organic Soil

matter

surveys,

FAO Affects

productivity

fertility

and

datasets

DEM (Digital Elevation Influences water flow and

Topographic

Vegetation

Elevation, slope, aspect

NDVI, biomass density

Model)

erosion

Remote sensing imagery

Indicates vegetation health

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 5

Water

drainage

availability,

land

Critical

planning

for

irrigation

Hydrological

GIS hydrological models

Survey/GIS layers

Socio-

economic

Accessibility,

ownership

Impacts

adoption

feasibility

and

1.5 Emergence of AI in Geospatial Analytics

The application of AI in geospatial analytics has significantly improved the ability to process and

interpret complex environmental datasets. Machine learning techniques such as Random Forest and

Support Vector Machines are widely used for classification and prediction tasks, while deep learning

models like Convolutional Neural Networks (CNN’s) provide enhanced performance in image-based

analysis [7], [8]. These models enable the extraction of high-level features from satellite imagery,

facilitating more accurate land-use classification and environmental monitoring.

Furthermore, AI enables predictive modelling, allowing researchers to forecast changes in land use,

vegetation patterns, and environmental conditions. This capability is particularly valuable for proactive

planning and sustainable resource management. However, the effectiveness of these models depends on

the availability of high-quality data and appropriate feature selection [9], [10].

1.6 Problem Statement

Despite the availability of advanced geospatial and AI technologies, many land management practices

still rely on fragmented and traditional approaches. These methods often fail to capture the spatial and

temporal complexity of environmental systems, leading to inefficient resource utilization and

suboptimal decision-making. In particular, traditional approaches lack the ability to integrate diverse

datasets, analyze large volumes of information, and provide real-time insights [11].

Moreover, the absence of integrated frameworks that combine AI, RS, and GIS limits the effectiveness

of current decision-support systems. While individual technologies have demonstrated significant

potential, their isolated application often results in incomplete or inconsistent analyses. This highlights

the need for a unified approach that leverages the strengths of these technologies to provide

comprehensive and accurate insights for sustainable land management [12].

1.7 Research Objectives and Contributions

In light of these challenges, this study aims to develop an integrated framework that combines AI,

remote sensing, and GIS for sustainable agro-forestry management and land resource optimization. The

primary objective is to utilize multi-source geospatial data and advanced analytical techniques to

support informed decision-making. Specifically, the study focuses on developing models for land

suitability assessment, optimizing resource allocation, and enhancing environmental monitoring

capabilities through AI-driven analysis.

This research contributes to the field by proposing a unified AI–RS–GIS framework that addresses the

limitations of existing approaches. The study emphasizes multi-source data integration, combining

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 6

satellite imagery, environmental variables, and spatial datasets to improve analytical accuracy. It also

introduces AI-driven predictive modelling techniques for land suitability assessment, enabling more

precise and reliable decision-making. Additionally, the development of a decision support system

(DSS) provides practical tools for policymakers and practitioners to implement sustainable agro-

forestry strategies. These contributions collectively advance the integration of geospatial technologies

and AI in environmental management [1], [3].

1.8 Organization of the Paper

The remainder of this paper is organized as follows. Section 2 presents a detailed review of the existing

literature on agro-forestry, geospatial technologies, and AI applications. Section 3 describes the

proposed methodology, including data collection, pre-processing, and model development. Section 4

discusses the results and findings, followed by a comprehensive discussion in Section 5. Finally, Section

6 concludes the paper and outlines directions for future research.

To provide a clear overview of the proposed integrated system, the conceptual framework illustrating

the interaction between Artificial Intelligence, Remote Sensing, and GIS components is presented in

Fig. 1.

Fig. 1. Conceptual framework of integrating Artificial Intelligence (AI), Remote Sensing (RS), and

Geographic Information Systems (GIS) for sustainable agro-forestry management and land resource

optimization

Fig. 1 presents the overall architecture of the proposed AI–RS–GIS integrated framework for

sustainable agro-forestry management. The process begins with multi-source data acquisition, including

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 7

satellite imagery, environmental variables, and socio-economic data. This is followed by data

preprocessing steps such as correction, feature extraction, and integration. In the next stage, AI-based

models are applied for land use classification, vegetation monitoring, and predictive analysis. The

outputs are then incorporated into GIS-based spatial analysis techniques, including multi-criteria

decision analysis and suitability mapping. Finally, a decision support system (DSS) generates actionable

insights, such as optimal land allocation and sustainable management recommendations. The

framework also includes a feedback loop for continuous monitoring and improvement, ensuring

adaptive and data-driven decision-making.

2. Literature Review

2.1 Agro-Forestry and Sustainable Land Management

Agro-forestry has evolved as a multidisciplinary approach that integrates agricultural and forestry

practices to enhance both ecological sustainability and economic productivity. At its core, agro-forestry

emphasizes the deliberate combination of trees with crops and livestock within the same land-use

system, creating synergistic interactions that improve overall system performance. Over the years,

researchers have increasingly highlighted the importance of agro-forestry in addressing environmental

challenges such as soil degradation, biodiversity loss, and climate change. Studies indicate that agro-

forestry systems contribute significantly to carbon sequestration, improve soil fertility through organic

matter accumulation, and enhance water retention capacity, thereby reducing vulnerability to climate

variability [13], [14].

Beyond environmental benefits, agro-forestry also plays a critical role in improving rural livelihoods. By

diversifying production systems, farmers can reduce economic risks associated with single-crop

dependency while increasing income stability. Socio-economic studies have shown that agro-forestry

contributes to food security by providing multiple outputs such as fruits, timber, fodder, and fuelwood

[15]. Moreover, these systems support sustainable land management by promoting efficient resource

utilization and reducing the need for chemical inputs. However, despite these advantages, the adoption

of agro-forestry practices remains uneven due to challenges such as lack of awareness, insufficient

technical support, and limited access to decision-making tools.

2.2 Remote Sensing in Environmental Monitoring

Remote sensing has become an indispensable tool in environmental monitoring, offering the ability to

observe and analyze large-scale spatial patterns over time. Satellite imagery from platforms such as

Landsat, Sentinel, and MODIS provides high-resolution data that can be used to assess land-use

changes, vegetation health, and environmental degradation. One of the most widely used techniques in

remote sensing is the calculation of vegetation indices, particularly the Normalized Difference

Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI). These indices provide quantitative

measures of vegetation density and health, enabling researchers to monitor crop conditions, detect

stress factors, and evaluate ecosystem productivity [16], [17].

In forestry applications, remote sensing has been extensively used for land cover classification, biomass

estimation, and deforestation monitoring. Machine learning algorithms applied to satellite imagery have

significantly improved classification accuracy, allowing for more precise mapping of land-use categories.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 8

Furthermore, time-series analysis of remote sensing data enables the detection of temporal changes,

which is crucial for understanding environmental dynamics and assessing the impact of human activities

[18]. Despite its advantages, remote sensing faces limitations such as cloud interference and variability

in spatial resolution, which can affect data quality and interpretation.

2.3 GIS for Spatial Analysis and Land Use Planning

Geographic Information Systems (GIS) play a central role in spatial analysis and land-use planning by

providing tools for data integration, visualization, and modelling. GIS enables the combination of

multiple spatial datasets, including topography, soil characteristics, climate variables, and land-use

information, to support comprehensive environmental assessments. One of the key strengths of GIS lies

in its ability to perform spatial modelling, which allows researchers to analyze relationships between

different environmental factors and predict future scenarios.

A widely used approach in GIS-based decision-making is Multi-Criteria Decision Analysis (MCDA),

which integrates multiple factors to evaluate land suitability for specific applications. MCDA

techniques, such as weighted overlay analysis, allow for the prioritization of criteria based on their

relative importance, enabling more informed and transparent decision-making processes [19], [20].

GIS-based environmental assessment has been successfully applied in various domains, including

agriculture, forestry, and urban planning. However, the effectiveness of GIS depends heavily on the

quality and accuracy of input data, as well as the selection of appropriate modelling techniques.

2.4 Artificial Intelligence in Geospatial Applications

The application of Artificial Intelligence (AI) in geospatial analysis has gained significant momentum

in recent years, driven by the increasing availability of large datasets and advances in computational

power. Machine learning algorithms such as Random Forest (RF), Support Vector Machines (SVM),

and Artificial Neural Networks (ANN) have been widely used for classification and prediction tasks in

environmental studies. These algorithms are particularly effective in handling high-dimensional data

and capturing complex, non-linear relationships between variables [21], [22].

Deep learning techniques, especially Convolutional Neural Networks (CNN’s), have further enhanced

the capabilities of geospatial analysis by enabling automated feature extraction from satellite imagery.

CNN-based models have demonstrated superior performance in image classification, object detection,

and change detection tasks, making them highly suitable for applications in land-use mapping and

environmental monitoring [23]. In addition to classification tasks, AI is increasingly being used for

predictive analytics, allowing researchers to forecast changes in land use, vegetation patterns, and

environmental conditions. These predictive capabilities are essential for proactive decision-making and

sustainable land management [24]. To summarize the key AI techniques and their applications in

geospatial studies, Table 3 provides a comparative overview.

Table 3: AI Techniques in Geospatial and Agro-Forestry Applications

AI Technique

Type

Application Area

Strengths

Limitations

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 9

Machine

Learning

Land

classification

High

robust to noise

accuracy,

Random Forest (RF)

Less interpretable

Support

Machine (SVM)

Vector Machine

Learning

Suitability

mapping

Effective

small datasets

with Computational

complexity

Artificial

Network (ANN)

Neural Neural

Prediction

modelling

Captures

linear relationships

non-

Risk of overfitting

Network

Convolutional Neural Deep

Image

High accuracy for Requires

large

Network (CNN)

Gradient Boosting

Learning

classification

imagery data

datasets

Ensemble

Learning

Environmental

prediction

Strong

predictive

Sensitive to noise

performance

2.5 Integrated AI–RS–GIS Approaches

The integration of AI, Remote Sensing, and GIS has emerged as a promising approach for addressing

complex environmental challenges. Several studies have proposed frameworks that combine these

technologies to improve land-use classification, environmental monitoring, and decision-making

processes. For instance, integrated models have been used to analyze multi-source data, enabling more

accurate and comprehensive assessments of land suitability and resource distribution [25], [26].

Comparative analyses of prior studies reveal that integrated approaches consistently outperform single-

technology methods in terms of accuracy and efficiency. The synergy between AI, RS, and GIS allows

for the automation of data processing, the extraction of meaningful patterns, and the generation of

actionable insights. However, existing frameworks often lack scalability and real-time capabilities,

limiting their applicability in dynamic environments. To highlight the differences between traditional

and integrated approaches, Table 4 presents a comparative analysis.

Table 4: Comparison of Traditional vs Integrated AI–RS–GIS Approaches

Data

Real-Time

Capability

Decision

Support

Approach

Accuracy

Scalability

Integration

Traditional

Methods

Low

Moderate

Limited

No

Basic

RS-based Methods

Moderate

Good

High

Moderate

High

Limited

Partial

GIS-based Methods High

Moderate

AI-based Methods

Moderate

Integrated AI–RS–

Very

High

Very High

High

Emerging

Advanced

GIS

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 10

2.6 Research Gaps

Despite the significant progress in geospatial technologies and AI, several research gaps remain. One of

the primary challenges is the lack of unified frameworks that seamlessly integrate AI, RS, and GIS into

a cohesive system. Many existing studies focus on individual components rather than exploring their

combined potential, resulting in fragmented solutions. Additionally, there is a limited availability of

real-time decision support systems that can provide timely and actionable insights for land

management.

Another important gap is the need for scalable and adaptive models that can handle large datasets and

dynamic environmental conditions. Current approaches often struggle with computational complexity

and data heterogeneity, which can limit their effectiveness in real-world applications. Addressing these

challenges requires the development of integrated frameworks that leverage the strengths of AI, RS, and

GIS while ensuring scalability, adaptability, and real-time performance. To systematically organize the

reviewed literature and highlight existing research gaps, the taxonomy of AI, Remote Sensing, and GIS

applications is illustrated in Fig. 2.

Fig. 2. Literature taxonomy and research gap analysis of AI–RS–GIS approaches for sustainable agro-

forestry management.

Fig. 2 presents a structured taxonomy of the literature on agro-forestry and land resource optimization

by categorizing existing studies into five major domains: agro-forestry and sustainable land

management, remote sensing, GIS-based spatial analysis, artificial intelligence techniques, and integrated

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 11

AI–RS–GIS frameworks. Each domain outlines key concepts, methodologies, and applications such as

vegetation monitoring, spatial modelling, machine learning, and predictive analytics.

The figure also highlights critical research gaps, including the lack of unified frameworks, limited real-

time decision-support systems, data heterogeneity, and scalability challenges. Based on these gaps, the

need for an integrated, intelligent, and scalable AI–RS–GIS framework is emphasized. Additionally,

future research directions are presented, focusing on real-time processing, advanced AI models, and

improved decision-support capabilities for sustainable agro-forestry management.

3. Methodology

3.1 Proposed Framework Overview

This study adopts an integrated methodological framework that combines Artificial Intelligence (AI),

Remote Sensing (RS), and Geographic Information Systems (GIS) to support sustainable agro-forestry

management and land resource optimization. Building on the research gaps identified in Section 2 and

the conceptual architecture illustrated in Figure 1, the proposed methodology is designed as a multi-

stage pipeline that systematically transforms raw multi-source data into actionable decision-support

outputs. The framework begins with data acquisition from heterogeneous sources, followed by pre-

processing and feature extraction, AI-based modelling, GIS-driven spatial analysis, and finally the

generation of decision-support insights. This structured workflow ensures that both spatial and non-

spatial dimensions of agro-forestry systems are effectively captured and analyzed.

The integration of AI with geospatial technologies enables the handling of high-dimensional datasets

and supports the development of predictive models for land suitability, biomass estimation, and

environmental monitoring. Furthermore, the incorporation of GIS-based multi-criteria decision analysis

(MCDA) ensures that multiple environmental and socio-economic factors are considered

simultaneously, leading to more informed and reliable decision-making. This holistic approach

addresses the limitations of traditional methods by providing a scalable, adaptive, and data-driven

solution for sustainable land management [27], [28]. The simulation workflow of the proposed AI–

RS–GIS integrated framework is illustrated in Fig. 3.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 12

Fig. 3. Simulation workflow of the proposed AI–RS–GIS framework for agro-forestry suitability

analysis and land resource optimization

Fig. 3 presents the end-to-end workflow of the proposed framework for agro-forestry suitability

analysis and land resource optimization. The process begins with defining objectives and collecting

multi-source data, including satellite imagery, climate, soil, topographic, and socio-economic datasets.

These inputs undergo preprocessing and feature extraction, where key indicators such as NDVI,

elevation, and environmental variables are generated.

The framework then performs parallel analysis through two main components: AI-based modeling and

GIS-based spatial analysis. AI models (e.g., RF, SVM, ANN, CNN) are used for classification and

prediction, while GIS techniques such as multi-criteria decision analysis (MCDA) and weighted overlay

are applied for spatial suitability mapping. The outputs from both components are integrated and

evaluated based on performance metrics and spatial consistency.

An iterative refinement loop ensures continuous improvement of model performance. Once acceptable

results are achieved, the system proceeds to validation and decision support generation. Finally, the

framework produces actionable outputs, including agro-forestry suitability maps, priority zones, land

resource optimization strategies, and management recommendation.

3.2 Study Area Description

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 13

The study area is selected based on its relevance to agro-forestry practices and environmental variability.

Typically, such regions are characterized by diverse land-use patterns, varying climatic conditions, and

heterogeneous soil properties. The selected area includes agricultural lands, forest cover, and

transitional zones, making it suitable for evaluating the effectiveness of integrated modelling

approaches. Key parameters such as temperature, rainfall, elevation, and vegetation density are

considered to capture the environmental complexity of the region. The spatial extent of the study area

is defined using GIS boundaries, ensuring consistency in data integration and analysis.

3.3 Data Acquisition and Sources

The proposed framework utilizes multi-source datasets to capture the complexity of agro-forestry

systems. Remote sensing data, including satellite imagery from platforms such as Landsat and Sentinel,

provides high-resolution spatial information for land-use classification and vegetation analysis.

Environmental datasets, including climate variables (temperature, rainfall), soil characteristics, and

topographic features, are obtained from publicly available databases and integrated into the analysis.

Additionally, socio-economic data such as land-use patterns, infrastructure, and accessibility are

incorporated to enhance decision-making. The integration of these diverse datasets ensures a

comprehensive representation of the study area, enabling more accurate modelling and analysis. Table 5

summarizes the key datasets used in this study.

Table 5: Data Sources and Description

Data Type

Source

Parameters

Resolution

Purpose

Land

classification

use

Satellite Imagery Landsat, Sentinel-2

Multispectral bands

10–30 m

Meteorological

Climate Data

Temperature,

rainfall

Environmental

analysis

Monthly/Annual

Varies

databases

FAO,

surveys

national Soil

type,

organic carbon

pH,

use,

Suitability

assessment

Soil Data

Topographic

Data

DEM (SRTM)

Elevation, slope

30 m

Terrain analysis

Decision support

Socio-economic

Data

Land

infrastructure

GIS layers, surveys

Varies

3.4 Data Pre-processing and Feature Engineering

Data pre-processing is a critical step in ensuring the quality and consistency of input datasets. Remote

sensing imagery undergoes radiometric and atmospheric correction to remove noise and improve data

accuracy. Geometric correction and co-registration are applied to align multiple datasets spatially,

ensuring consistency across different data layers. Cloud masking techniques are used to eliminate cloud-

covered pixels, which can affect the reliability of analysis.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 14

Feature engineering involves the extraction of meaningful variables from raw data. Vegetation indices

such as NDVI, EVI, and NDWI are computed to assess vegetation health and water content.

Topographic features, including slope and aspect, are derived from Digital Elevation Models (DEM),

while soil and climatic variables are standardized for integration into the modelling process. These

features serve as input variables for AI-based models, enhancing their predictive capabilities [29]. The

data processing and feature engineering pipeline used to prepare multi-source inputs for modeling is

illustrated in Fig. 4.

Fig. 4. Data preprocessing and feature engineering pipeline for transforming multi-source geospatial

data into structured inputs for AI-based modeling and GIS analysis

Fig. 4 presents a systematic pipeline for converting raw multi-source geospatial data into a clean and

structured dataset suitable for AI-based modelling and GIS analysis. The process begins with data

acquisition from multiple sources, including remote sensing imagery, environmental, soil, topographic,

and socio-economic data.

In the preprocessing stage, operations such as radiometric and geometric correction, cloud masking, and

normalization are applied to ensure data quality and consistency. This is followed by feature extraction,

where relevant indicators such as spectral indices (NDVI, EVI), textural features, topographic

variables, and soil and climate attributes are derived.

The feature engineering stage further refines the dataset through layer stacking, feature selection,

normalization, handling of missing values, and dataset splitting for training and testing. Finally, a

structured feature matrix is generated, which serves as input for AI models and GIS-based spatial

analysis. The pipeline also produces intermediate outputs such as feature raster’s, statistical summaries,

and correlation analysis to support model development and validation.

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 15

3.5 AI-Based Modelling and Analysis

The core analytical component of the framework involves the application of AI-based models for

classification, prediction, and optimization. Machine learning algorithms such as Random Forest (RF),

Support Vector Machines (SVM), and Artificial Neural Networks (ANN) are used for land use and

land cover (LULC) classification. These models are trained using labelled datasets and validated

through cross-validation techniques to ensure robustness.

Deep learning models, particularly Convolutional Neural Networks (CNN’s), are employed for

advanced image analysis tasks, including feature extraction and pattern recognition. CNN’s are

particularly effective in handling high-resolution satellite imagery, enabling accurate classification of

land-use categories and detection of subtle changes in vegetation patterns. In addition to classification,

predictive models are developed to estimate land suitability, biomass, and agricultural productivity. The

AI-based modeling framework used for land suitability prediction is presented in Fig. 5.

Fig. 5. AI-based modeling framework for land suitability prediction using machine learning and deep

learning techniques

Fig. 5 presents the structured workflow of the AI-based modelling framework used for land suitability

prediction. The process begins with the input feature set derived from remote sensing, topographic,

soil, climate, and socio-economic data. These inputs undergo data preparation steps, including cleaning,

normalization, feature selection, and dataset splitting into training and testing subsets.

The framework then applies multiple machine learning and deep learning models, such as Random

Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Convolutional

Neural Network (CNN), using k-fold cross-validation to ensure robust model training. Model

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 16

performance is evaluated using standard metrics including accuracy, precision, recall, F1-score, Kappa

coefficient, and ROC-AUC.

Based on comparative analysis, the best-performing model is selected and used for land suitability

prediction. The outputs include suitability maps, probability maps, classified suitability zones (high,

moderate, marginal, and not suitable), and decision-support recommendations. An iterative

optimization loop is incorporated to refine model performance and improve prediction accuracy. To

evaluate the performance of these models, standard metrics such as accuracy, precision, recall, F1-score,

and Kappa coefficient are used. Table 6 presents the evaluation metrics applied in this study.

Table 6: Model Evaluation Metrics

Metric

Description

Purpose

Accuracy

Precision

Recall

Overall correctness of predictions

True positives over predicted positives

True positives over actual positives

Harmonic mean of precision and recall

Agreement beyond chance

General performance

Reliability of predictions

Sensitivity

F1-Score

Kappa Coefficient

Balanced performance

Model robustness

3.6 GIS-Based Spatial Analysis

GIS-based spatial analysis is used to integrate the outputs of AI models with environmental and socio-

economic datasets. Multi-criteria decision analysis (MCDA) is applied to evaluate land suitability for

agro-forestry by assigning weights to different factors based on their importance. Techniques such as

Analytical Hierarchy Process (AHP) and weighted overlay analysis are used to combine multiple

criteria and generate suitability maps.

Spatial modelling is further employed to identify priority zones for agro-forestry interventions. This

involves the analysis of spatial patterns and relationships between different variables, enabling the

identification of areas with high potential for sustainable land use. The integration of AI outputs with

GIS analysis enhances the accuracy and reliability of spatial decision-making [30]. The GIS-based

multi-criteria decision analysis framework used for land suitability assessment is illustrated in Fig. 6.

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 17

Fig. 6. GIS-based multi-criteria decision analysis (MCDA) framework for agro-forestry suitability

assessment and spatial decision-making

Fig. 6 presents the GIS-based multi-criteria decision analysis (MCDA) framework used for agro-

forestry suitability assessment. The process begins with the identification of key evaluation criteria,

including land use, vegetation, slope, elevation, soil type, and climatic factors. These criteria are

standardized to a common scale to ensure consistency across different data layers. The Analytical

Hierarchy Process (AHP) is then applied to assign relative weights to each factor based on their

importance, followed by a consistency check to validate the weighting scheme.

The standardized and weighted criteria are integrated using a weighted overlay analysis to generate a

composite suitability index. This index is subsequently classified into different suitability zones, such as

highly suitable, moderately suitable, marginally suitable, and not suitable areas. The final outputs

include suitability maps, priority zones, and decision-support recommendations, which provide a

structured basis for sustainable agro-forestry planning and efficient land resource management.

3.7 Decision Support System (DSS)

The final stage of the framework involves the development of a Decision Support System (DSS) that

translates analytical results into actionable insights. The DSS provides interactive tools for

visualization, analysis, and reporting, enabling stakeholders to explore different scenarios and make

informed decisions. Outputs include land suitability maps, priority zones for agro-forestry, and

recommendations for resource allocation. The DSS is designed to be user-friendly and adaptable,

allowing for the incorporation of new data and continuous updates. This ensures that the system

remains relevant and responsive to changing environmental conditions.

3.8 Workflow Integration and Continuous Monitoring

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 18

An important feature of the proposed methodology is the incorporation of a feedback loop for

continuous monitoring and improvement. The system continuously updates models based on new data,

enabling adaptive decision-making and improving predictive accuracy over time. This dynamic

approach ensures that the framework remains scalable and applicable to different regions and

environmental conditions.

4. Results

4.1 Land Use and Land Cover Classification

The land use and land cover (LULC) classification provide a comprehensive overview of the spatial

distribution of different land categories within the study area. Using the proposed AI–RS–GIS

framework, the region was classified into five major categories: forest, agricultural land, water bodies,

built-up areas, and barren land. The classification results indicate a diverse landscape, where agricultural

land occupies the largest portion, followed by forested areas concentrated in relatively less disturbed

regions. Built-up areas are mainly located near transportation corridors and urban centres, while barren

land appears in zones with poor soil quality or limited vegetation cover.

To evaluate the classification performance, standard accuracy metrics were computed based on

validation samples. The results demonstrate a balanced performance across all classes, with relatively

higher accuracy observed for water bodies and agricultural areas due to their distinct spectral

characteristics. Slight confusion was observed between barren land and sparse vegetation, which is

common in heterogeneous environments. The classification accuracy results are summarized in Table 7.

Table 7: Land Use and Land Cover Classification Accuracy

Class

Producer Accuracy (%)

User Accuracy (%)

F1-Score (%)

91.1

Forest

90.7

92.9

95.8

88.9

86.5

90.8

0.88

91.5

92.1

96.3

87.6

85.9

—

Agriculture

Water Bodies

Built-up

92.5

96.0

88.2

Barren Land

Overall Accuracy

Kappa Coefficient

86.2

—

The results indicate that the classification framework achieves reliable performance across all land-use

categories, making it suitable for subsequent spatial and suitability analysis. The spatial distribution of

land use and land cover classes derived from the classification process is shown in Fig. 7.

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 19

Fig. 7. Land use and land cover (LULC) classification map of the study area generated using the

proposed AI–RS–GIS framework

Fig. 7 presents the land use and land cover (LULC) classification map of the study area generated using

the proposed AI–RS–GIS framework. The map categorizes the region into major classes, including

forest, agricultural land, water bodies, built-up areas, and barren land. Forest cover is predominantly

observed in the northern and elevated regions, while agricultural land is widely distributed across the

central valleys. Built-up areas are concentrated around major urban centers such as Mingora, indicating

patterns of urban expansion.

The spatial distribution highlights a diverse landscape influenced by topography and human activity.

The presence of dense vegetation in higher altitudes and agricultural dominance in lowland areas

reflects the suitability of the region for agro-forestry practices. Additionally, the identification of barren

and sparsely vegetated areas provides useful insights for land restoration and resource management.

This classification serves as a foundational input for subsequent suitability analysis and decision-

support modelling. The classification performance is further evaluated using a confusion matrix, as

presented in Fig. 8.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 20

Fig. 8. Confusion matrix illustrating the classification accuracy across different land-use classes

Fig. 8 presents the confusion matrix used to evaluate the performance of the LULC classification. The

matrix compares actual and predicted classes across five categories: forest, agriculture, water bodies,

built-up areas, and barren land. The diagonal elements represent correctly classified samples, showing

strong classification performance across all classes. Water bodies exhibit the highest classification

accuracy, while slight misclassification is observed between built-up areas, agriculture, and barren land,

which is expected due to spectral similarities.

The overall classification accuracy is 90.8%, with a Kappa coefficient of 0.88, indicating a high level of

agreement beyond chance. User and producer accuracies further confirm the reliability of the

classification, with most classes achieving values above 85%. These results demonstrate that the

proposed AI–RS–GIS framework provides robust and consistent classification performance, forming a

reliable basis for subsequent suitability analysis and spatial modelling.

4.2 Agro-Forestry Suitability Mapping

The agro-forestry suitability mapping results reveal the spatial variation in land potential for sustainable

agro-forestry practices. By integrating environmental, topographic, and socio-economic factors, the

study identifies areas that are highly suitable, moderately suitable, marginally suitable, and not suitable

for agro-forestry implementation.

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 21

Highly suitable zones are generally characterized by favourable soil conditions, adequate rainfall, and

moderate slopes, which support both crop growth and tree development. Moderately suitable areas

show minor environmental constraints, while marginal zones are affected by factors such as poor soil

quality or limited water availability. Not suitable areas are primarily associated with steep slopes,

degraded land, or urbanized regions. The spatial distribution of agro-forestry suitability zones is

illustrated in Fig. 9.

Fig. 9. Agro-forestry suitability map showing highly suitable, moderately suitable, marginally suitable,

and unsuitable zones

Fig. 9 presents the agro-forestry suitability map of the study area, classifying land into highly suitable,

moderately suitable, marginally suitable, and unsuitable zones. The map reveals that highly suitable

areas are primarily located in regions with dense vegetation, favourable soil conditions, and moderate

terrain, while moderately suitable zones dominate a large portion of the landscape. Marginally suitable

areas are scattered across transitional zones, whereas unsuitable regions are mainly concentrated in

steep, barren, or environmentally constrained areas.

The spatial distribution highlights the influence of topography, vegetation cover, and environmental

factors on land suitability. The central and northern regions exhibit higher suitability due to better

ecological conditions, while peripheral and elevated areas show lower suitability. This classification

provides a clear basis for identifying priority zones for agro-forestry development, supporting targeted

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 22

land-use planning and efficient resource management. The distribution of land across these suitability

categories is presented in Table 8.

Table 8: Agro-Forestry Suitability Distribution

Suitability Class

Highly Suitable

Moderately Suitable

Marginally Suitable

Not Suitable

Area (km²)

298.4

Percentage (%)

26.9

365.7

33.0

247.3

22.3

197.8

17.8

Total

1109.2

100

The results indicate that a significant portion of the study area falls within highly and moderately

suitable categories, highlighting strong potential for agro-forestry expansion and sustainable land-use

planning. The distribution of land area across different agro-forestry suitability classes is illustrated in

Fig. 10.

Fig. 10. Distribution of land area across agro-forestry suitability classes in the selected study area

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 23

Fig. 10 illustrates the distribution of land area across different agro-forestry suitability classes in the

study area. The results show that moderately suitable areas occupy the largest proportion, followed by

highly suitable and marginally suitable zones. In contrast, unsuitable and not suitable areas represent a

relatively smaller share of the total land area, indicating that a significant portion of the region has

potential for agro-forestry development.

The distribution highlights a favorable overall suitability pattern, where more than half of the study

area falls within high to moderate suitability categories. This suggests strong potential for sustainable

land-use planning and agro-forestry expansion. The presence of marginal and unsuitable zones also

provides important insights for targeted interventions, such as land rehabilitation and environmental

management strategies.

4.3 Model Performance Analysis

The performance of the developed models was evaluated to assess their effectiveness in classification

and prediction tasks. A 5-fold cross-validation strategy was applied to ensure robustness and minimize

the influence of data variability. This approach provides a more reliable estimate of model performance

by averaging results across multiple training and testing splits.

As shown in Table 9, all models demonstrate stable and competitive performance, with accuracy values

ranging from approximately 86% to 90%. The Convolutional Neural Network (CNN) achieved the

highest overall accuracy of 89.7%, along with the best F1-score and Kappa coefficient. This indicates

its ability to effectively capture complex spatial patterns and relationships within the dataset. The

dataset was split into training (70%) and testing (30%) subsets before applying 5-fold cross-validation

Random Forest (RF) performed consistently well, particularly in precision, suggesting strong capability

in minimizing false positives. The Artificial Neural Network (ANN) showed slightly higher recall than

precision, indicating its effectiveness in identifying true positive instances. The Support Vector

Machine (SVM) exhibited comparatively lower recall, which may be attributed to its sensitivity to

parameter tuning in complex datasets.

Table 9: Performance Comparison of AI Models (5-Fold Cross-Validation Results)

Model

Accuracy (%)

88.6 ± 1.4

86.9 ± 1.7

87.8 ± 1.5

89.7 ± 1.2

Precision (%)

89.2 ± 1.6

87.5 ± 1.9

86.9 ± 1.7

88.6 ± 1.4

Recall (%)

87.3 ± 1.8

85.6 ± 2.1

88.4 ± 1.6

90.3 ± 1.3

F1-Score (%)

88.2 ± 1.5

86.5 ± 1.8

87.6 ± 1.4

89.4 ± 1.2

Kappa

0.82

0.79

0.80

0.84

Random Forest (RF)

SVM

ANN

CNN

The relatively low variation across folds indicates stable model performance, with CNN demonstrating

slightly better generalization capability compared to other models. The comparative performance of the

evaluated machine learning and deep learning models is illustrated in Fig. 11.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 24

Fig. 11. Performance comparison of AI models based on accuracy, precision, recall, F1-score, and

Kappa coefficient using 5-fold cross-validation

Fig. 11 presents the performance comparison of different AI models, including Random Forest (RF),

Support Vector Machine (SVM), Artificial Neural Network (ANN), and Convolutional Neural

Network (CNN), evaluated using 5-fold cross-validation. The results show that all models achieve

strong and consistent performance across multiple metrics, including accuracy, precision, recall, F1-

score, and Kappa coefficient. Among them, the CNN model demonstrates the highest overall

performance, particularly in accuracy and recall, indicating its effectiveness in capturing complex spatial

patterns.

The inclusion of standard deviation values highlights the stability and reliability of the models across

validation folds. While CNN slightly outperforms the other models, Random Forest also shows

competitive performance with strong precision values. In contrast, SVM and ANN exhibit slightly

lower but still acceptable performance. Overall, the results confirm that the proposed AI-based

framework provides robust and reliable predictions for agro-forestry suitability assessment.

4.4 Environmental Monitoring Insights

The integration of remote sensing data enabled detailed environmental monitoring, particularly in

assessing vegetation health and land degradation patterns. Vegetation indices such as NDVI were used

to analyze spatial and temporal variations in vegetation density. The results indicate that areas classified

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 25

as highly suitable exhibit consistently higher NDVI values, reflecting healthier vegetation and better

environmental conditions.

Conversely, regions identified as marginally suitable or unsuitable show lower NDVI values, indicating

reduced vegetation cover and potential environmental stress. These patterns are consistent with

observed soil conditions and rainfall distribution, reinforcing the reliability of the integrated analysis.

Additionally, zones with declining vegetation trends were identified as areas prone to land degradation.

These findings provide valuable insights for targeted interventions, such as soil restoration and

sustainable land management practices. The spatial variation in vegetation health across the study area

is illustrated in Fig. 12.

Fig. 12. Spatial distribution of vegetation health based on NDVI values across the study area

Fig. 12 illustrates the spatial distribution of vegetation health across the study area based on NDVI

values. Areas with higher NDVI values, represented by dark green shades, indicate dense and healthy

vegetation, primarily observed in the northern and elevated regions. In contrast, lower NDVI values,

shown in yellow to red shades, correspond to sparse vegetation or degraded land, mainly concentrated

around urban areas and low-lying regions.

The NDVI patterns highlight a strong relationship between vegetation health and environmental

conditions, such as topography and land use. Regions with high NDVI values align with highly

suitable agro-forestry zones, while areas with lower values correspond to marginal or unsuitable regions.

This analysis supports the reliability of the suitability mapping and provides valuable insights for

environmental monitoring and sustainable land management.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 26

4.5 Decision Support Outputs

The final outputs of the proposed framework are designed to support practical decision-making in

agro-forestry planning and land resource optimization. The Decision Support System (DSS) integrates

the results of AI modelling and GIS-based analysis to generate actionable recommendations for

stakeholders.

The system identifies priority zones for agro-forestry implementation, enabling efficient allocation of

resources and targeted interventions. Highly suitable areas are recommended for expansion of agro-

forestry practices, while marginal zones are suggested for rehabilitation through soil improvement and

conservation measures. Scenario-based analysis further allows decision-makers to evaluate different

land-use strategies and their potential outcomes. The key decision-support outputs are summarized in

Table 10.

Table 10: Decision Support Outputs

Output Type

Description

Suitability Maps

Spatial representation of agro-forestry potential

Identification of high-potential areas

Efficient land allocation strategies

Guidelines for sustainable land use

Priority Zones

Resource Optimization Plans

Management Recommendations

These outputs provide a comprehensive foundation for informed decision-making, supporting

sustainable land management and long-term environmental resilience. The final decision-support

outputs for optimized land-use planning and agro-forestry implementation are illustrated in Fig. 13

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 27

Fig. 13. Decision support output map showing priority zones and optimized land-use strategies for

sustainable agro-forestry management

Fig. 13 presents the decision-support output map derived from the integrated AI–RS–GIS framework.

The map highlights priority zones for agro-forestry development, categorized into high, moderate, and

low priority areas, along with rehabilitation and restricted zones. High-priority areas are mainly

concentrated in regions with favorable environmental conditions, while moderate and low-priority

zones are distributed across transitional landscapes. Rehabilitation zones are identified in degraded

areas, and restricted zones correspond to environmentally sensitive or unsuitable regions.

The spatial distribution of these zones provides actionable insights for sustainable land-use planning

and resource optimization. High-priority areas are recommended for immediate agro-forestry

implementation, while marginal and degraded regions require restoration and soil improvement

strategies. This decision-support output enables policymakers and planners to make informed, data-

driven decisions, ensuring efficient allocation of resources and long-term environmental sustainability.

5. Discussion

The results demonstrate that the proposed AI–RS–GIS framework is effective for land suitability

analysis and sustainable agro-forestry planning. The LULC classification reveals a heterogeneous

landscape dominated by agricultural and forest areas, indicating strong potential for agro-forestry

expansion. However, the presence of barren land and built-up areas highlights environmental pressures

and the need for balanced land-use strategies.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 28

The suitability analysis shows that a significant portion of the study area falls within highly and

moderately suitable categories, primarily due to favorable vegetation, soil, and terrain conditions. The

NDVI analysis further supports these findings, showing a clear relationship between vegetation health

and suitability classes. Areas with higher NDVI values align with highly suitable zones, while lower

values correspond to degraded or unsuitable regions, confirming the reliability of the integrated

framework.

Model performance evaluation indicates that all models provide stable and reliable results, with CNN

achieving the highest overall performance. However, the improvement over traditional models is

moderate, suggesting that simpler models remain practical and efficient. The decision-support outputs

identify priority zones for development, conservation, and rehabilitation, offering actionable insights

for policymakers. Overall, the framework provides a scalable and data-driven approach for sustainable

land resource optimization, with future improvements possible through temporal and climate-based

analysis.

6. Conclusion

This study presented an integrated AI–RS–GIS framework for agro-forestry suitability analysis and

land resource optimization. The results demonstrate that combining remote sensing data, spatial

analysis, and artificial intelligence enables accurate identification of suitable zones and supports

informed land-use planning. A significant portion of the study area was found to be suitable for agro-

forestry, highlighting its potential for sustainable development.

The model evaluation shows that while deep learning approaches provide slightly better performance,

traditional machine learning models remain effective and computationally efficient. The consistency

between NDVI-based environmental analysis and suitability mapping further validates the reliability of

the proposed approach.

Moreover, the framework provides a practical and scalable decision-support tool for policymakers and

planners, facilitating efficient resource allocation, environmental conservation, and sustainable agro-

forestry management.

7. Future Work

Future research can enhance the proposed framework by incorporating temporal analysis and multi-year

satellite data to capture seasonal and long-term land-use dynamics. The integration of climate change

projections and weather variability can further improve predictive accuracy and adaptability. Advanced

deep learning architectures and hybrid models may be explored to better handle complex spatial

patterns. Additionally, the inclusion of socio-economic and policy-driven factors can strengthen

decision-making relevance. Expanding the framework to larger geographic regions and real-time

monitoring systems would also improve its scalability and practical implementation.

References

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 29

1. Gu, Z., & Zeng, M. (2024). The use of artificial intelligence and satellite remote sensing in land

cover change detection: Review and perspectives. Sustainability, 16(1), 274.

2. Aziz, D., Rafiq, S., Saini, P., Ahad, I., Gonal, B., Rehman, S. A., ... & Nabila Iliya, M. (2025).

Remote

sensing

and

artificial

intelligence:

revolutionizing

pest

management

in

agriculture. Frontiers in Sustainable Food Systems, 9, 1551460.

3. Sadenova, M., Beisekenov, N., & Varbanov, P. S. (2024). Assessing the effectiveness of RS, GIS,

and AI data integration in analysing agriculture performance to enable sustainable land

management. Discover Sustainability, 5(1), 478.

4. Madane, D. A., Tarate, S. B., & Yousuf, A. (2026). Integration of Artificial Intelligence, Remote

Sensing, and GIS for Sustainable Land Management and Planning. In Advancements in Soil

Conservation: From Traditional Practices to AI-Driven Solutions (pp. 539-559). Cham: Springer

Nature Switzerland.

5. Selmy, S. A., Kucher, D. E., & Moursy, A. R. (2025). Integrating remote sensing, GIS, and AI

technologies in soil erosion studies. In Modern Geospatial Approaches for Environmental

Monitoring and Management. IntechOpen.

6. Farbo, A., Sarvia, F., De Petris, S., Basile, V., & Borgogno-Mondino, E. (2024). Forecasting corn

NDVI through AI-based approaches using sentinel 2 image time series. ISPRS Journal of

Photogrammetry and Remote Sensing, 211, 244-261.

7. Judith, J., Tamilselvi, R., Beham, M. P., Lakshmi, S., Panthakkan, A., Mansoori, S. A., & Ahmad,

H. A. Remote Sensing Based Crop Health Classification Using NDVI and Fully Connected

Neural Networks. arXiv 2025. arXiv preprint arXiv:2504.10522.

8. Escobar-López, A., Castillo-Santiago, M. Á., Mas, J. F., Hernández-Stefanoni, J. L., & López-

Martínez, J. O. (2024). Identification of coffee agroforestry systems using remote sensing data: a

review of methods and sensor data. Geocarto International, 39(1), 2297555.

9. Giannetti, F., Puletti, N., Puliti, S., Travaglini, D., & Chirici, G. (2020). Assessment of UAV

photogrammetric DTM-independent variables for modelling and mapping forest structural indices

in mixed temperate forests. Ecological Indicators, 117, 106513.

10. Zhu, X. X., Tuia, D., Mou, L., Xia, G. S., Zhang, L., Xu, F., & Fraundorfer, F. (2017). Deep

learning in remote sensing: A comprehensive review and list of resources. IEEE geoscience and

remote sensing magazine, 5(4), 8-36.

11. Kamran, M., Tanveer, K., Khalid, N., Khalil, M. N., Ahmad, B., Arooj, A., ... & Mir, S. Z. (2025).

Integrating advanced deep learning algorithms for climate systems: Enhancing weather forecast

accuracy, real-time climate monitoring, and long-term climate predictions. Spectrum of

Engineering Sciences, 365-403.

12. Cian, F., Marconcini, M., & Ceccato, P. (2018). Normalized Difference Flood Index for rapid

flood mapping: Taking advantage of EO big data. Remote sensing of environment, 209, 712-730.

13. Belgiu, M., & Drăguţ, L. (2016). Random forest in remote sensing: A review of applications and

future directions. ISPRS journal of photogrammetry and remote sensing, 114, 24-31.

14. Mehmood, B., Rani, G., Khalid, N., Majeed, M. K., Ahmad, B., Qamar, S., & Saeed, M. (2025).

Development of a hybrid artificial intelligence framework for accurate forecasting of solar power

generation using machine learning algorithms and time-series analysis. Spectrum of Engineering

Sciences, 613-636.

15. Foody, G. M. (2002). Status of land cover classification accuracy assessment. Remote sensing of

environment, 80(1), 185-201.

16. Sharma, A., Paliwal, K. K., & Onwubolu, G. C. (2006). Class-dependent PCA, MDC and LDA: A

combined classifier for pattern classification. Pattern Recognition, 39(7), 1215-1229.

GRJNST, Volume: 04 - Issue 2 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07

G. 2086

Page 30

17. Grill, A., & Cleary, D. F. (2003). Diversity patterns in butterfly communities of the Greek nature

reserve Dadia. Biological Conservation, 114(3), 427-436.

18. Koklu, M., Cinar, I., & Taspinar, Y. S. (2021). Classification of rice varieties with deep learning

methods. Computers and electronics in agriculture, 187, 106285.

19. Boles, S. H., Xiao, X., Liu, J., Zhang, Q., Munkhtuya, S., Chen, S., & Ojima, D. (2004). Land

cover characterization of Temperate East Asia using multi-temporal VEGETATION sensor

data. Remote Sensing of Environment, 90(4), 477-489.

20. Aslam, M. F., Ahmed, M. U., Akbar, K., & Amjad, A. (2025). THE INFLUENCE OF

DIGITAL LEADERSHIP ON SUSTAINABLEPERFORMANCE: THE MEDIATING

ROLES OF DIGITAL ORGANIZATIONAL CULTURE, AND THE MODERATING

ROLE OF ORGANIZATIONAL AGILITY. Journal of Management Science Research

Review, 4(2), 62-90.

21. Zhang, C., Kerner, H., Wang, S., Hao, P., Li, Z., Hunt, K. A., ... & Shen, Y. (2025). Remote

sensing for crop mapping: A perspective on current and future crop-specific land cover data

products. Remote Sensing of Environment, 330, 114995.

22. Singh, A. K., Ghosh, A., Sannagoudar, M. S., Kumar, R. V., Kumar, S., Singh, P. D., & Ahamad, S.

(2022). Sustainable agriculture systems and technologies. Sustainable Agriculture Systems and

Technologies, 279-294.

23. Ennouri, K. (2026). USE OF REMOTE SENSING FOR PRECISION. AI and Deep Learning

Enabled Surveillance System Using Image Processing, 173.

24. Mohammed, K., Kpienbaareh, D., Kerr, R. B., Wang, J., Luginaah, I., Lupafya, E., ... &

Mkandawire, M. (2026). Integrating participatory GIS, remote sensing, and explainable machine

learning to assess forest provisioning services. Environmental Impact Assessment Review, 117,

108245.

25. Akbar, K. (2021). How Eco-innovation & Green Manufacturing Affects Sustainability

Performance: A Quantitative Study In Pakistan. International Research Journal of Modernization

of Engineering Technology and Science, 3(6), 3785-3792.

26. Siddiqui, M. H. S., Abbasi, M. D., Mir, S. Z., Nadeem, G., Majeed, M. K., Khawer, S. K., ... &

Kashif, M. (2025). Prediction of maximum air temperature for defining heat wave in various

states of Pakistan using machine learning algorithm. Spectrum of Engineering Sciences, 942-960.

27. Nuwarapaksha, T. D., Udumann, S. S., Dissanayaka, N. S., & Atapattu, A. J. (2025). Future

Prospects and Emerging Trends in Agroforestry Research. Agroforestry for a Climate-Smart

Future, 497-518.

28. Bhardwaj, D. R., & Bhatia, A. K. (2025). Land use and land cover analysis in agroforestry and

other land use categories in Shimla district of Himachal Pradesh: a GIS-based analysis in North

Western Himalayas. Discover Forests, 1(1), 33.

29. Machireddy, S. R. (2023). Natural resource management using remote sensing and geographic

information systems. Environmental Science and Engineering, 2(2), 73-82.

30. Udumann, S. S., Dissanayaka, N. S., Nuwarapaksha, T. D., & Atapattu, A. J. (2025). Agroforestry

Education and Training for the Next Generation. In Agroforestry for a Climate-Smart Future (pp.

449-472). IGI Global Scientific Publishing.

GRJNST, Volume: 04 - Issue 3 (2026) / ISSN P: 2790-7643

Article ID: 2086

https://doi.org/10.53762/grjnst.04.03.07