A Prototype Decision Support System for Tree Selection and Plantation with a Focus on Agroforestry and Ecosystem Services

Abstract

This study presents the development and application of a prototype decision support system (DSS) for tree selection specifically for Punjab, India, a region facing challenges of low forest cover and an increasing demand for sustainable land use practices. The DSS developed using the R Shiny framework integrates ecological, social, and agro-commercial criteria to facilitate scientific knowledge decision making in tree plantation. The modules in this DSS include a tree selection tool based on comprehensive species attributes, a GIS-based tree suitability map module utilizing an Analytical Hierarchical Process (AHP), and a silviculture practice information module sourced from authoritative databases. Combining sophisticated statistical and spatial analysis, such as NMDS and AHP-GIS, this DSS mitigates data redundancy in SDM while incorporating extensive bibliographic research in dataset processing. The study highlights the necessity of fundamental niche-based suitability in comparison to realized niche suitability. It emphasizes on the importance of addressing ecosystem services, agro-commercial aspects, and enhancing silvicultural knowledge. Additionally, the study underscores the significance of local stakeholder engagement in tree selection, particularly involving farmers and other growers, to ensure community involvement and support. The DSS supports agroforestry initiatives and finds applications in urban tree management and governmental programs, emphasizing the use of scientific literature at each step, in contrast to relying solely on local knowledge.

1. Introduction

Tree plantation decision making requires expert scientific knowledge and rule-based processes to address three core objectives: species selection, location, and value assessment, facilitated by guided user inputs through computerized interfaces [1]. The development of a data-driven decision support system for tree selection is crucial as it must integrate comprehensive databases on geographical, soil, climatic, commercial, and physiological factors. The adoption of agroforestry practices is often hindered by a lack of scientific information and poor awareness among farmers and landowners [2].

India’s climate exhibits significant variations compared to other countries at similar latitudes, with temperature ranges playing a major role in the distribution of tree species [3].

Commercial agroforestry, which involves cultivating commercial cash crops alongside intercrops, is becoming increasingly prevalent across India. This practice is particularly significant for timber, pulp, and firewood production, with an estimated value of USD 480 million covering 5 million hectares. Common tree species used include Eucalyptus, Populus, Casuarina, Leucaena, Ailanthus, Melia, Anthocephalus, Acacia, and Bombax [4]. This trend is expected to address India’s wood shortage and enhance agricultural income nationwide.

It is crucial to expand and enrich the knowledge base and address knowledge gaps particularly in emerging areas of agroforestry research. These areas include the value of ecosystem services, biofuel production, and carbon sequestration, the long-term sustainability of trees in subsistence farming, and climate resilience and adaptation [5]. However, there is lack of a robust decision support system (DSS) for tree selection and plantation in India. Additionally, the limited focus on ecosystem services has led to the preference for commercial exotic tree species [6], which has hindered forest departments from making decisions based on comprehensive scientific knowledge. The commercial benefits of agroforestry and the enhancement of ecosystem services depend on scientific knowledge-based decision making. This approach promotes agroforestry systems, such as alley cropping, which require critical information on plantation spacing, precipitation requirements, soil characteristics, and topography for optimal decision making. By ensuring informed decisions, agroforestry can increase ecosystem service benefits, including nutrient enrichment, soil erosion control, and soil quality improvement [7].

Previous DSSs for tree selection have focused on various ecosystem services, including urban heat island (UHI) mitigation (23), air quality improvement (29), runoff reduction and erosion control (26), non-timber forest products (NTFPs) (28), timber production (30), water quality enhancement (19), slope protection (25), and biodiversity and cultural services (27). This study aims to internalize and address all these critical ecosystem services as part of the DSS development.

In this study, the DSS is focused on Punjab (50,362 km²), a northwestern state in India with very low forest cover (5%) overall. Punjab has experienced a small but significant loss in forest cover in recent years (−1.9 % change). To address this, promoting agroforestry plantations is crucial as over 81% of the land in Punjab is agriculture cropland. This makes agroforestry the only viable option for increasing forest cover or trees outside of forested areas (Forest Service of India; FSI) [8].

Furthermore, soil characteristics are crucial for evaluating tree suitability as neglecting these factors increases the likelihood of plantation failure [9]. The soil datasets provided by International Soil Reference and Information Centre (ISRIC) are the most comprehensively mapped soil data available globally. These datasets are produced using advanced machine learning techniques based on extensive soil observations and more than 400 environmental covariates, chosen for their relationship with the major soil formation factors worldwide [10]. Moreover, ISRIC provides important variables that represent key factors such as rootable depth and water holding capacity of soil, which directly affect tree growth [11].

In various future climate change scenarios, it is increasingly important to have a DSS that strategizes a selection of climate-tolerant and climate-resilient tree species. This approach enhances tree survival and ensures the continuance of ecosystem services [12].

Temperature is one of the most critical variables determining the survival of any species [13]. With rising temperatures, it becomes essential to consider temperature variables in terms of tree species distribution and suitability, especially in lowland areas where the impact of climate change is also significant [14].

There are two widely used approaches to species distribution models (SDM) or suitability models: (i) correlative SDM (e.g., MaxEnt) and (ii) multi-criteria analysis. Correlative SDMs have become more widely adopted in recent years due to their simple data requirements as they utilize statistical functions to link species occurrence locations with environmental geospatial datasets [15,16]. However, correlative SDMs are limited by time and space, and changes in variables over time are not properly internalized in the models [17]. Multi-criteria analysis, such as the Analytic Hierarchy Process (AHP), involves considering various factors that influences species suitability and survival, including climate, soil, and topography, and assigning weights to each factor [18].

Tabassum et al. [19] utilized species distribution models (SDMs) to map the suitability of numerous individual species. However, in a geographically vast area like Australia, with a limited number of data points, predicting suitability becomes challenging. They employed niche modeling, but these models were based on realized niches (only areas where species occurrence data were available) rather than fundamental niches (areas where the environment is actually suitable for these species). Consequently, using SDMs for tree suitability and survival predictions introduces significant biases, as observed in the “Which Plant Where” project [19]. For example, many species of Acacia were limited to very small areas in Northern Queensland despite having higher climatic and soil suitability across the semi-arid/arid landscape of Australia. Additionally, many of these SDMs have not utilized soil datasets for assessing tree species soil suitability. Most SDMs typically depend on realized niches based on the equilibrium assumption that natural occurrences and distributions of tree species reflect their climatic requirements. However, it is now essential to recognize that the climatic suitability or adaptability of a tree species beyond their natural distributions must be modeled using other datasets, such as trials [20].

Therefore, addressing the concerns regarding the reliability and accuracy of correlative models such as MaxEnt SDMs, process-based models have been proposed in the past. These models focus on fundamental niches, as realized niches are constrained by various variables and represent only a subset of the fundamental niche [21]. Moreover, the clustering of tree species has recently been applied to summarize extensive distribution data and identify ecologically significant patterns among species. For example, a study identified 11 clusters out of 390 tree species in Borneo [22]. Similarly, an earlier study applied broader faunal clustering for taxonomic groups such as primates, bats, and other carnivores, grouping species based on climatic similarities and then assessing habitat suitability [23]. As a result, we advocate for employing AHP-based species group mapping, which utilizes a clustering of species approach. This approach utilizes similar climatic and soil parameters sourced from the scientific literature, such the CABI digital library and Plantation Trees (a comprehensive species-wise textbook on Indian Plantation Trees; [24]).

The emerging trend in DSS development shows a preference for R and Python frameworks, as they facilitate data-driven tasks [6]. R Shiny dashboards offer extensive capabilities for effective dashboard development, leveraging the shinydashboard package. The DT R package enhances functionality via its integration with the Javascript library DataTables, which aids in database creation and management. Moreover, Shiny simplifies development processes by translating R code into HTML, CSS, and JavaScript, thereby making bioinformatics web application development accessible even to non-programmers within the bioinformatics community [25].

Hence, in this study, we discuss the development, implementation, and benefits of a decision support system (DSS) for tree selection and plantation in Punjab, India. This DSS utilizes the R Shiny framework and incorporates a detailed AHP-based grouping and clustering of tree species along with their respective map analysis for tree species suitability. Additionally, the broader objectives addressed by this DSS, compared to other existing systems, are briefly appraised to elucidate its accomplishments.

2. Materials and Methods

The DSS was developed for Punjab, India, using the R Shiny Web Framework. The Shinydashboard package was employed to create the tree DSS dashboard. This DSS has been made available online as an open-access resource in its alpha stage of testing and development.

To streamline usability, the dashboard consists of three main modules as summarized in the workflow in Figure 1:

(i).

Tree selection tool;

(ii).

GIS-based tree suitability map module;

(iii).

Silviculture practices information module.

These modules were designed to provide comprehensive support for tree selection, a visualization of tree suitability via GIS, and access to information on silviculture practices within Punjab.

2.1. Tree Selection Tool Development

This tool was specifically designed to facilitate tree selection based on a wide range of attributes from the tree database. It was developed using the general filtering capabilities of the DT R package, applied to a CSV file containing the tree database, which was developed utilizing the institute’s own tree datasets as further discussed in detail further in this section. The tool includes a form that retrieves filtered tree species based on criteria specified in the tree selection form. The tree species list for the database was sourced from the tree directory of Punjab (Punjab Biodiversity Board), encompassing both native and exotic tree species.

Previously, the FRI developed Haryana Forest Flora (www.haryanaforestflora.in, accessed on 14 June 2024) which contains a comprehensive database of ligneous plants found in Haryana, a neighboring state to Punjab with a similar climate. The tree attributes from this database were utilized for the development of the current tool, as most tree species in Punjab are also covered in this database. Information for each tree species was drawn from this database, including phenological characteristics, topography, distribution, and high-resolution field-captured photographs. Additionally, important attributes such as salinity tolerance, flowering season, native–exotic species classification, carbon sequestration potential, and other relevant details were incorporated based on the comprehensive database available.

Furthermore, information regarding the agroforestry commercial uses and urban plantation purposes of each tree species in the database was sourced from Ecofriendly trees for urban beautification. This publication aims to meet the needs of planters and growers by guiding them to choose the appropriate tree species based on their possible uses. Without this crucial information, planters often struggle to select the right tree species [26]. The index of uses derived from this publication is provided as a separate tab in the dashboard, facilitating easy access to information on the various uses of each tree species for agroforestry and urban plantation purposes.

2.2. GIS-based Tree Suitability Map Module

This module was developed based on suitability maps of tree species grouped according to climate, soil, and topography. The data values used were derived from the extensive literature sourced from the CABI digital library and the Plantation Trees book by Luna (1996) [24], which covers most species in this region.

These suitability maps were created using the Analytical Hierarchical Process (AHP) within the QGIS Open Source Environment. Geographic dataset layers were obtained from two main sources: climatological data from CHELSA-BIOCLIM+ Version 1.2 (1979–2013 mean) [27] and soil datasets from ISRIC soil grids. The ISRIC soil grids datasets, with a resolution of 250 m, are considered the latest and most relevant information available with global coverage [28].

2.2.1. Climatological and Soil Datasets

The suitability maps were prepared based on multiple layers crucial for determining the survivability of tree species. These layers include the following:

• Tmax May (Maximum Temperature of the Hottest Month): Alongside Bio1, Tmax May is recognized as a robust predictor of species distribution, as evidenced by previous studies [29,30]. This layer was particularly crucial for assessing tree species tolerance to temperature, especially during the hottest month of the year. Similar AHP-based suitability modeling studies, such as those conducted for Mango Mangifera indica in India [31], have underscored its significance in understanding and predicting species distributions.
• Tmax Jan (Max Temperature of Coldest Month): This layer is crucial for assessing both cold tolerance and maximum temperature tolerance during the coldest month of the year. It plays a significant role in understanding how well tree species can endure cold temperatures and precipitation levels [32]. This attribute is essential for their survival and adaptation strategies during winter months.
• Bio1 (Mean Annual Temperature): This variable is widely recognized as a strong predictor in species distribution models [33]. It provides crucial information about the overall thermal conditions experienced by tree species throughout the year. Additionally, Bio1, along with Bio12, is considered highly important in larger ecological studies. However, research indicates that temperature variables often hold greater significance than rainfall variables in determining species distributions [34]. This is because temperature variability plays a critical role in shaping the geographic range and adaptation strategies of species.
• Bio12 (Mean Annual Precipitation): This variable was included in the suitability mapping process because rainfall distribution is crucial for grouping tree species based on their water requirements and ecological preferences. Bio12 represents the mean annual precipitation, providing essential data on the amount of rainfall received by a region annually. In similar dry deciduous regions, such as those in Sub-Saharan Africa, Bio12 has been identified as a critical variable influencing the distribution and ecological niche of tree species [35]. Understanding precipitation pattern helps in assessing the suitability of different tree species to specific climatic conditions, ensuring their successful growth and survival.
• SOC (Soil Organic Carbon): Soil organic carbon, specifically its content in the fine earth fraction, was included in the suitability mapping process. This variable is crucial because it reflects the organic matter content in the soil, which plays a significant role in supporting tree growth and nutrient availability. In studies conducted in dry deciduous regions, such as in Ecuador, SOC has been identified as an important variable influencing tree species distribution [36]. It provides insights into soil fertility, water retention capacity, and overall soil health, all of which are critical factors affecting the suitability of tree species to specific environments.
• Sand: Sand content, also described as soil organic carbon content in the fine earth fraction, which is crucial for many tree species’ suitability, was included in the suitability mapping process. According to Luna 1996, many tree species in Punjab favor sandy soils. This attribute was chosen due to the significant ecological association observed between sand content and plant species. The ISRIC sand data, known to closely match field measurements, were particularly valued for their reliability in enhancing model accuracy [37]. Sand content was utilized in the model to ensure an accurate assessment of tree species suitability, aligning with ecological preferences documented in regional tree species information and validated by robust field data.
• Bulk Density: Bulk density is a crucial variable included in the model to assess soil compaction, drainage conditions, and overall soil quality [31]. It provides insights into the density and porosity of the soil, influencing root growth, water movement, and nutrient availability for tree species. Studies have highlighted the high environmental correlation of bulk density with perennial plants [38]. This correlation underscores its importance in understanding the ecological requirements and distribution patterns of tree species. Incorporating bulk density in the suitability mapping process enhances the model’s ability to accurately predict the optimal habitats and growth conditions for different tree species based on soil characteristics.

2.2.2. Statistical Analysis and Tree Species Grouping/Clustering

Furthermore, the AHP-processed maps depicting grouped/clustered species distributions were visualized using leaflet and raster R packages. These species were categorized into three major groups based on their ecological preferences:

(i).

Temperate, warm climate with clayey mixed soil species;

(ii).

Temperate, hot climate with sandy loam soil species;

(iii).

Arid, hot climate with sandy loam soil species.

This classification was aligned with the Köppen–Geiger climate classification map (1980–2016) as described by Beck et al. [39], and soil classification properties were derived from “Plantation Trees” [24]. Visualizing these grouped species distributions aids in understanding the spatial patterns of tree species suitability across Punjab, considering both climate conditions and soil types. The leaflet and raster R packages were instrumental in creating interactive and informative maps, facilitating effective decision making in tree selection and plantation planning. In addition, a non-metric multi-dimensional scaling (NMDS) approach was utilized to visualize these three clusters based on climatological data. This computation was performed using the R ‘vegan’ package to identify potential groupings of tree species based on the climatic variables Tmax May, Tmax Jan, and Bio1.

For the AHP model, a pair-wise comparison matrix (

Table 1. Pair-wise comparison matrix for comparison of degree of importance of different variables. The last column shows the weights of each variable.

Variables	Tmax May	Bio1	SOC	Sand	Tmax Jan	Bio12	BD	Weight
Tmax May	1	5	6	4	3	6	5	0.44
Bio1	0.2	1	0.333	2	0.5	2	2	0.087
SOC	0.167	3	1	2	0.4	0.8	1.667	0.099
Sand	0.25	0.5	0.5	1	0.143	1	1.25	0.055
Tmax Jan	0.333	2	2.5	7	1	4	2.5	0.19
Bio12	0.167	0.5	1.25	1	0.25	1	0.5	0.054
BD	0.2	0.5	0.6	0.8	0.4	2	1	0.064

) was constructed using a comparison scale ranging from 1 to 9 to assess relative importance. These weights were derived from expert knowledge and informed by the scientific literature [40]. The pair-wise comparisons in the matrix allow for the prioritization of criteria and factors influencing tree species suitability, ensuring robust decision making based on systematic evaluations of their relative significance. Additionally, sub-factors (4 quantile ranges) of each variable were ranked as well (

Table 2. The suitability ranking of criteria across various groups is denoted as follows: 4 = very high, 3 = high, 2 = moderate, 1 = low. These rankings are based on environmental and soil information obtained from CABI digital library (CABI 2024) and Plantation Trees [24].

Variables	Sub-Factors	Group 1	Group 2	Group 3
Tmax May (in °C)	0–37	4	3	1
	37–39	3	4	2
	39–40	1	2	4
	40–50	2	1	3
Bio1	18–20	4	3	3
(in °C)	20–22	3	4	4
	22–24	2	2	2
	24–50	1	1	1
SOC (dg/kg)	0–111	4	1	1
	111–132	3	2	4
	132–180	2	3	3
	180–2000	1	4	2
Sand	0–348	1	1	4
(g/kg)	348–399	2	3	3
	399–438	4	4	2
	438–800	3	2	1
Tmax Jan	−20–0	4	2	3
(in °C)	0–10	2	4	4
	10–12	3	3	2
	12–50	1	1	1
Bio12	0–450	1	1	3
(in mm)	450–1000	2	2	4
	1000–1700	3	4	2
	1700–5000	4	3	1
Bulk Density	0–134	4	4	3
(cg/cm3)	134–144	3	3	4
	144–146	2	2	2
	146–200	1	1	1

).

2.3. Silviculture Practice Information Module

The silvicultural practice information module incorporates textual information sourced from [24] as mentioned in

Table 3. Summary table showing list of datasets utilized in each module developed in DSS.

Component	Data Source
Tree database–tree selection module	Primary: Haryana Forest Flora (www.haryanaforestflora.in, accessed on 14 June 2024); additional: India Biodiversity portal (www.indiabiodiversity.org accessed on 14 June 2024) [41]; Plantation Trees [24]
GIS-based suitability module	AHP based on ISRIC and CHELSA-BIOCLIM + datasets
Index of uses	Ecofriendly trees for urban beautification [26]
Silviculture information module (ongoing)	Plantation Trees [24]

. To efficiently manage this information, a form was developed in PHP to create a MySQL database. This database contains detailed text-rich silvicultural guidelines specific to each tree species. Although the module in the R Shiny app is subject to updates and the entry of additional silviculture literature data, the concept is illustrated here with a sample page for Dalbergia sissoo.

Each species information page within this module provides comprehensive silvicultural guidelines, covering aspects such as phenology, soil, and site factors, natural and artificial regeneration methods, seed treatments, nursery and planting techniques, and growth and yield considerations, as well as pests and diseases management. This structured approach ensures that detailed information crucial for effective tree management and cultivation practices is readily accessible and organized for users.

3. Results

The tree decision support system (DSS) for Punjab is published online and can be accessed at the following link: (http://rakholias.shinyapps.io/Tree_PDSS).

3.1. Tree DSS Interface: Intro Tab, Tree Selection form Filters and Index of Uses

Firstly, a small step-by-step guide to use the tree selection DSS, along with information on ecosystem services and agroforestry systems, is provided in the first tab of the dashboard (Figure 2). It is particularly important to display the available ecosystem services and agroforestry types in the subsequent tree selection form tab.

The step-by-step guide includes the following steps for proper usage of this DSS:

• Introduction Page: read the introduction page to understand how to use the step-by-step guide, and access information on agrofestry systems and ecosystem services.
• Index of Uses Tab: identify the unique code for each commercial/non-commercial uses type, which will be useful to enter in the next tab.
• Tree Selection Form Tab: enter the identified codes and other criteria into the form fields to filter suitable species based on the entered criteria.

Selection Filter Results: Review the filtered results based on the selection criteria to choose the most suitable tree species. This structured approach ensures users can effectively utilize the DSS to select tree species that best meet their specific requirements for ecosystem services and agroforestry applications.

Secondly, the second tab, labeled “Index of Uses”, displays index codes for both commercial and non-commercial uses of tree species. Given the lengthy Index of Uses list, this page provides detailed information on various uses, highlighting the commercial benefits of agrofrestry and urban tree plantations. Users can employ the index numbers as single or multiple criteria in the subsequent tree selection form tab. The Index of Uses is particularly important as it offers a wide range of options for use selection, including but not limited to fruits, fodder, landscaping, shade, dyes, oil, wood, paper, etc. This detailed index ensures users can make informed decisions about tree species based on specific use cases, enhancing the commercial and ecological benefits of their selections (Figure 3).

Lastly, the tree selection form contains several important attribute fields including the following: Origin (Native/Exotic), Salinity tolerance, Flowering period, Agroforestry system types, ecosystem services, Uses (index number based on Index of Uses in the previous tab), and Carbon Sequestration Potential. This tree database, with all its attributes and relevant information, was compiled from the Haryana Forest Flora species database, as mentioned earlier. Based on the selected values of these fields, the form fetches filtered tree species metadata from the master data table. These metadata include attribute information such as the following: Tree species name, Group (geographical suitability), Origin, Tree height, Tree canopy crown spread, etc., as shown in Figure 4. Additionally, the form can be expanded to include many other attributes contained in the tree species master data CSV file. This flexibility ensures that users can tailor their tree selections based on a comprehensive set of criteria, optimizing both the ecological and commercial outcomes of their planting decisions (Figure 4).

3.2. Tree DSS Interface GIS-Based Tree Suitability Map Module

The species grouping/clustering approach is also highlighted in the NMDS plot to show climatic similarity between various species that are further mapped in this module (Figure 5). This was performed to assess the clusters prior to grouping each tree species based on key climatic variable data [24] in the AHP model, including Tmax Jan, Tmax May, and Bio1. The stress value for these clusters was 0.1043, indicating a good representation and ordination with a non-existent risk of false inferences, as stress values < 0.1 are considered highly reliable [42].

The NMDS plot primarily showed clustering in higher temperature regimes, with genera such as Acacia and Terminalia forming Cluster 3. In contrast, genera such as Ficus and Eucalyptus were clustered together along with various other species. Additionally, genera Albizia, one species of Ficus, and Pinus were grouped together in Cluster 1. These clusters were based on climatic data as per the literature [24], excluding soil type as soil type is qualitative and thus not included in NMDS computation. Consequently, some overlaps and exclusions were expected, which were further refined based on expert knowledge. Nevertheless, the NMDS plot was crucial in the natural clustering of species based on important climatic information.

The tree suitability map module tab (Figure 6) displays a map with raster values ranging from 3 to 8, where 3 indicates low suitability and 8 indicates high suitability. Values of 1 and 2, which represent very low suitability, were excluded from the scale. Originally, the data range was from 1 to 4, but it was doubled to better visualize continuous values (Figure 4). Therefore, values in the range of 2 to 4 are visualized as follows:

• 2 (3–4): low to moderate suitability;
• 3 (5–6): moderate suitability;
• 4 (7–8): high suitability.

The suitability of the tree species groups/clusters mapped for different regions of Punjab (Figure 7) is summarized as follows:

• Group 3 (arid, hot, sandy loam soil type species): shows better suitability in Southern Punjab.
• Group 2 (temperate, hot, sandy loam soil type species): has high to moderate suitability in northern and Center–East Punjab.
• Group 1 (temperate, warm, clayey mixed soil species): is especially suitable in districts bordering the Indian Himalayan Region.

3.3. Tree DSS Interface Silviculture Information Module

The tree silvicultural information module (Figure 8) provides detailed information on silvicultural practices. This includes guidelines and the best practices for the cultivation, management, and care of tree species, ensuring users have access to critical information for successful tree planting and maintenance.

3.4. Summary Appraisal of Functionalities of Prototype DSS

It is essential to compare the functionalities of our DSS with respect to other existing DSSs, which were extensively reviewed in our recent publication focusing on five major objectives: climate resilience, ecosystem services, space optimization, agroforestry, and urban sustainability [6]. Based on these objectives, we further evaluate our DSS in a detailed manner, addressing these aspects along with other objectives that were previously unaddressed, as mentioned in

Table 4. The major objectives of various DSS and the degree to which they were addressed in our prototype DSS.

Objective	Appraisal (Addressal)	Description
Climate Resilience	Partially	Utilized climatic dataset for mapping suitable areas for tree species in GIS suitability map module
Ecosystem Services	Extensively	Multiple ecosystem services of tree species are included in tree selection module
Space Optimization	Not addressed	Other than tree height and crown size, this DSS does not address this objective much
Agroforestry	Extensively	Index of Uses (Agrocommercial) and Agroforestry systems are internalized in DSS
Urban Sustainability	Partially	Urban sustainability aspects in Index of Uses (urban avenues, fruit trees, etc.) are internalized in DSS; GIS suitability map module also covers all urban areas as well
Additional Objective: Silviculture Practice Information	Extensively (Ongoing)	Silviculture information module was added (ongoing) as this objective was not assessed in Yadav et al., 2014 [6], as most DSSs lacked this type of tool

.

4. Discussion

This study not only addresses the development of a decision support system (DSS) for tree selection but also integrates ecological concepts of agroforestry, ecosystem services, and climatic and soil suitability using data driven modules, including spatial data. Additionally, it introduces the concept of providing literature-based silviculture information, which is crucial for tree care and management but has not been extensively addressed in the past. The silviculture practice information module was added to offer guidance on plantation techniques specific to each tree species available in the state.

Although current research indicates that MaxEnt provides relatively better accuracy in distribution modeling compared to AHP-GIS methods, this advantage primarily stems from the availability of extensive occurrence records for a few selected species only [43]. Moreover, there is a significant data deficit in GBIF occurrence data for India, particularly for various tree species native to Punjab and surrounding states. Additionally, many studies underestimate the spatial bias and the importance of initial parameters in species distribution modeling, as proper non-random distribution modeling necessitates extensive data correction [44].

Our results, based on the AHP, comprehensively focus on the fundamental niche aspect rather than the realized niche typically produced in the SDM. This approach significantly enhances the accuracy of potential tree species suitability, as seen in previous studies [45]. Furthermore, the relatively higher resolution of ISRIC Soil Grids datasets represents a recent practice in our modeling, unlike systems such as Which Plant Where, which have excluded these data. This exclusion is likely due to the higher importance of climatic conditions in determining the distribution and suitability of tree species. Furthermore, the clustering or grouping of tree species approach for mapping their geographical suitability was statistically analyzed in this study. This analysis demonstrates that statistically clustering species using NMDS provides insights into grouping species in climatological terms. The clustering approach was informed by the tree species literature and similar techniques that combine knowledge from bibliographic research and the multi-criteria evaluation capabilities of AHP [46]. More importantly, the ability of decision support systems to emphasize the separation of native and exotic species, along with their respective ecosystem services and agroforestry systems, becomes crucial in raising awareness among farmers and growers. In countries like India, exotic species can become invasive weeds, increase pest problems, disrupt avifauna, and colonize larger regions, as seen with species such as Prosopis juliflora and Lantana camara [47]. Moreover, in tree plantation decision making, it is important to involve local residents in tree selection to ensure shared decision-making power. This approach fosters the meaningful involvement of residents, particularly in the maintenance and care of trees [48]. As a result, this decision support system (DSS) was designed to enable multiple stakeholders to participate in tree selection decision making in the state, considering ecological, social, agro-commercial, and silvicultural aspects. Silvicultural interventions can ensure proportional gains in both economic and habitat values [49].

This DSS addressed major objectives such as agroforestry, ecosystem services, and silviculture practices (Table 4). Silviculture practices for tree species have been notably excluded from most other DSSs, as observed in previous review studies [6]. Additionally, this DSS incorporated aspects of urban sustainability in the suitability maps within the GIS module, as well as index codes in the Index of Uses. While climate resilience in future scenarios was not part of this study, current climatic suitability was mapped based on the extensive research literature.

The northern plains of Punjab are classified climatically as temperate, dry winter, and hot summer (Cwa), while the southern semi-arid region is classified as arid, steppe, and hot (BSh) as per the Koppen Geiger Climate Classification System [39]. Therefore, temperate agroforestry systems, including alley cropping, forest farming, riparian buffers, silvopasture, and windbreaks, are more suitable for this region rather than tropical agroforestry systems [50]. Hence, these systems were included in the tree selection form to provide a range of suitable options of agroforestry systems.

In addition, many DSSs in agroforestry have historically focused on yield and market economic objectives, neglecting social and environmental aspects. Therefore, it has been suggested that future DSSs should address environmental services and involve multiple stakeholders [51]. Nevertheless, this DSS has placed extensive emphasis on non-timber forest products (NTFPs), particularly with the Index of Uses component which lists numerous NTFPs and other uses. Incorporating the significance of NTFPs and forest values in the socio-cultural dimension can enhance sustainability [52].

Furthermore, in this study, scientific bibliographic literature produced by forestry experts such as R.K. Luna and R.S. Troup (Indian Forestry Service, formerly Imperial Forestry Service) in India was extensively utilized. Past studies on DSS development have cautioned that local knowledge alone is insufficient for decision making and should always be supplemented with the scientific literature to ensure appropriate tree selection [53]. Lastly, similar DSS developed using R programming can be highly valuable for decision making focused on ecosystem services and urban sustainability, as seen in projects like Right Place, Right Tree | Boston [54], which also inspired the development of the prototype DSS in this study.

The core objective of this study has been to provide a multi-module interface DSS that facilitates the integration of essential ecological and agro-commercial concepts. Additionally, it aims to extend its applications beyond agroforestry farmers to include urban tree management, government plantation programs, NGOs, and other stakeholders, ensuring that all major objectives are equally addressed.

5. Conclusions

In conclusion, our study has developed a robust decision support system (DSS) for tree selection that integrates ecological, social, and agro-commercial considerations. By addressing ecosystem services, it proposes the involvement of multiple stakeholders, including local residents, farmers, growers, forest officials and conservationists, as our DSS aims to foster sustainable tree plantation practices in Punjab and similar regions. The multi-module interface not only supports agroforestry farmers but also extends its utility to urban tree management, government initiatives, and non-governmental organizations. The use of sophisticated statistical analysis and spatial mapping techniques in this study provides new insights for the development of decision support systems, supported by the rigorous scientific literature. However, it is essential to address some broader objectives of tree selection and plantation DSS, particularly future climate suitability and space optimization.