Smart Urban Tree Valorization: An AI-Blockchain-Based Application

for the Preservation of Remarkable Trees

Hajer Nabli

1 a

, Issra Jegham

, Yasmine Zorgati

, Rania Ajmi

, Raoudha Ben Djemaa

and

Layth Sliman

University of Sousse, Higher Institute of Computer Science and Communication Technologies of Sousse/Miracl Lab,

Sousse, Tunisia

University of Sousse, High Institute of Agronomic Science of Chott Mariem, Sousse, Tunisia

Paris Pantheon-Assas University, Efrei Research Lab, Villejuif, France

Keywords:

Remarkable Trees, Artiﬁcial Intelligence, Chatbot, Blockchain, NFT, Urban Environmental Monitoring.

Abstract:

Urban trees contribute signiﬁcantly to the well-being of city dwellers by improving air quality, reducing heat,

and offering psychological and cultural value. Among them, remarkable trees—due to their rarity, size, age,

or symbolic importance—deserve special attention, yet they often remain poorly documented or undervalued.

This paper introduces a smart application that uses artiﬁcial intelligence (AI) and blockchain to enhance public

awareness and long-term recognition of these trees. The system features an AI-driven chatbot that interacts

with users by asking questions about tree types or suggesting trees based on their desired beneﬁts—such as

relaxation, biodiversity, or carbon absorption—thus guiding users toward relevant and meaningful discoveries.

In parallel, each remarkable tree is assigned a blockchain-based digital certiﬁcate in the form of a non-fungible

token (NFT), ensuring that its identity, characteristics, and location are securely recorded and veriﬁable. By

combining participatory exploration with digital certiﬁcation, this approach offers a novel tool for cities to

promote biodiversity, support environmental education, and foster citizen involvement in urban green heritage.

The proposed system ﬁlls a gap between static tree mapping platforms and dynamic, secure, and intelligent

urban biodiversity applications.

1 INTRODUCTION

Urban ecosystems are vital for sustainable develop-

ment, providing ecological, health, and social bene-

ﬁts. Among these, urban trees help ﬁlter air pollu-

tants, mitigate heat islands, sequester carbon, support

biodiversity, and contribute to well-being and cultural

identity (Torres and McDonald, 2021). Remarkable

trees—distinguished by age, size, rarity, or symbolic

value—hold particular ecological and heritage impor-

tance but are often under-identiﬁed or overlooked in

urban planning (Yoon and Fischer, 2023).

Existing digital platforms for urban tree manage-

ment rely on static inventories or geographic infor-

mation systems (GIS), offering basic information but

lacking personalized, interactive features. They rarely

engage citizens in discovering, monitoring, or pre-

serving urban tree heritage, limiting opportunities

https://orcid.org/0000-0002-3603-251X

for public participation and environmental education

(Mullaney et al., 2015; Li and Wu, 2022).

Advances in Artiﬁcial Intelligence (AI) and

Blockchain can address these gaps. AI enables natural

language understanding, personalized recommenda-

tions, and context-aware interaction (Radhakrishnan

et al., 2022), while Blockchain ensures transparent,

immutable, and veriﬁable data records (Treiblmaier,

2018; Abdelghani et al., 2023; Abdelhamid et al.,

2024).

This paper presents a smart AI- and blockchain-

based system for valorizing and preserving remark-

able urban trees, comprising two main components:

• An AI-powered chatbot that allows citizens to ex-

plore trees via natural language queries and re-

ceive personalized suggestions based on ecolog-

ical, recreational, or carbon-sequestration criteria.

• A blockchain-backed certiﬁcation mechanism

that Assigns each tree a non-fungible token (NFT)

Nabli, H., Jegham, I., Zorgati, Y., Ajmi, R., Ben Djemaa, R. and Sliman, L.

Smart Urban Tree Valorization: An AI-Blockchain-Based Application for the Preservation of Remarkable Trees.

DOI: 10.5220/0013948300003982

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 22nd International Conference on Informatics in Control, Automation and Robotics (ICINCO 2025) - Volume 2, pages 609-620

ISBN: 978-989-758-770-2; ISSN: 2184-2809

609

storing its identity, characteristics, and geoloca-

tion in a secure, tamper-proof format.

By combining intelligent interaction with secure dig-

ital certiﬁcation, the system promotes urban biodi-

versity, environmental education, and citizen engage-

ment.

The remainder of the paper is structured as fol-

lows: Section 2 reviews related work, Section 3 de-

tails the system architecture, Section 4 presents im-

plementation and evaluation, and Section 5 concludes

with key ﬁndings and future directions.

2 RELATED WORK

The integration of digital tools in the monitoring

and preservation of urban trees has gained trac-

tion in recent years (Varveris et al., 2023; Nandhini

et al., 2021). This section reviews existing works

across three key dimensions relevant to our appli-

cation: (1) urban tree mapping platforms, (2) AI-

powered chatbots for environmental engagement, and

(3) blockchain and NFTs for ecological and cultural

heritage certiﬁcation.

2.1 Urban Tree Mapping Platforms

In recent years, a growing number of digital platforms

have been developed to document, share, and promote

urban tree inventories. These tools support biodiver-

sity awareness, engage local communities, and often

aim to inform urban planning and conservation ef-

forts.

Among the most notable is Moabi

, a French mo-

bile application speciﬁcally designed for the partici-

patory referencing of remarkable trees. Moabi em-

powers both citizens and professionals to geolocate

trees, upload photographs, and contribute descriptive

information such as species, dimensions, or historical

anecdotes. The platform seeks to build a nationwide,

crowdsourced inventory of ecologically and cultur-

ally signiﬁcant trees, thus fostering broader public in-

volvement in environmental heritage preservation.

Other important initiatives include TreeMap LA,

developed by TreePeople

, which enables users in

Los Angeles to view, add, and update data related to

street and park trees. It emphasizes civic participa-

tion in tree maintenance and urban forestry. Similarly,

TreeTalk London

provides interactive green walking

https://www.moabi.re/

https://www.treemap.in/#home

https://www.treetalk.co.uk/

routes, combining cartographic interfaces with biodi-

versity data to enhance public engagement with the

urban forest. On a broader scale, OpenTrees

ag-

gregates tree inventory datasets from over 60 cities

worldwide and offers an API for basic visualization

and data access.

While these platforms clearly demonstrate the

beneﬁts of community engagement and geospatial vi-

sualization in urban ecology, they present several lim-

itations in the context of long-term heritage recogni-

tion and digital certiﬁcation. Most systems:

• Lack of interactive or intelligent educational com-

ponents: these platforms do not include features

like conversational agents or other tools that en-

courage user learning and exploration

• Lack of data veriﬁcation or authentication mech-

anisms: there are no reliable processes to conﬁrm

tree uniqueness, origin, or historical signiﬁcance

• Absence of user feedback or annotation options:

users cannot provide input or add notes, limiting

engagement and the dynamic nature of environ-

mental storytelling

• Lack of durable, tamper-proof records: these sys-

tems fail to maintain persistent, secure identities

for trees over time

These limitations reduce their effectiveness for

supporting institutional recognition, long-term trace-

ability, and the symbolic valorization of individual

trees — particularly in cases where trees are consid-

ered urban natural heritage assets. In contrast, our

proposed platform introduces a proactive AI chat-

bot and a blockchain-backed certiﬁcation system, en-

abling more secure, informative, and user-centered in-

teraction with urban trees.

2.2 AI Chatbots for Agriculture and

Environmental Awareness

AI-driven chatbots are increasingly adopted in agri-

culture to provide farmers with real-time informa-

tion, facilitate crop management, and support in-

formed decision-making (Pravinkrishnan et al., 2022;

Ong et al., 2021; Niranjan et al., 2019; Chathurya

et al., 2023). These systems typically combine natu-

ral language processing (NLP) techniques to interpret

user queries, classiﬁcation algorithms to detect crop-

related issues or suggest interventions, and rule-based

methods to organize domain knowledge and generate

context-aware responses.

AgronomoBot (Mostaco et al., 2018) is developed

to assist farmers by retrieving real-time data from

http://opentrees.org/

TISAS 2025 - Special Session on Trustworthy and Intelligent Smart Agriculture Systems: AI, Blockchain, and IoT Convergence

610

wireless sensor networks deployed in vineyards. Us-

ing the Telegram messaging platform, it interacts with

users via natural language and leverages IBM Watson

services for intent recognition. Its goal is to facili-

tate fast and context aware access to environmental

data, including soil and climate conditions, thus en-

hancing agricultural decisions. However, it is highly

domain-speciﬁc, lacks entity-level interactions, and

does not support static, descriptive data like botanical

attributes or historical context. It also lacks integra-

tion with web platforms or map-based visualization,

which are key to our project.

Similarly, Agribot (Arora et al., 2020) employs

an LSTM-based conversational model combined with

CNNs for plant disease detection and weather fore-

casting. Though technically advanced, it is built for

broad agricultural tasks and lacks support for user-

generated feedback, location-speciﬁc recommenda-

tions, or interactive interfaces beyond Telegram.

A third system, AgroBot (Marla et al., 2023),

trained on India’s Kisan Call Center data, is a voice-

enabled chatbot designed to answer farming-related

questions using RNNs and deep learning. However,

it is based on a closed dataset, produces pre-scripted

responses, and provides no interface for visualizing or

contextualizing environmental entities, making it less

suitable for heritage or educational goals.

Adding to these, (Usip et al., 2022) propose a

mobile chatbot using ontologies and similarity algo-

rithms to assist Nigerian farmers. While effective for

localized knowledge retrieval, it lacks multimodal in-

teraction, emotional or well-being-based inputs, and

does not manage individual entities with rich meta-

data (such as speciﬁc trees).

Despite their contributions, current agricultural

chatbots share key limitations in the context of our

goals:

• No support for entity-speciﬁc interaction: These

systems do not manage individualized, geo-

referenced environmental objects like remarkable

trees

• Absence of user well-being perspective: They

are focused on factual problem-solving, not on

proposing nature-based recommendations tied to

emotional or ecological beneﬁts

• Lack of participatory or community layers: Ex-

isting chatbots do not engage citizens in co-

producing data or contributing to environmental

heritage awareness

• Minimal interface diversity: Most are mobile-

or Telegram-based with no map-based, web-

integrated experiences.

Our chatbot addresses these gaps by reversing the

interaction ﬂow: it asks users about their desired ben-

eﬁts (e.g., shade, air puriﬁcation, relaxation), then

recommends speciﬁc remarkable trees that match

those goals. This design supports exploratory en-

vironmental learning, enhances citizen engagement,

and aligns with broader urban biodiversity and well-

being objectives.

2.3 Blockchain for Digital

Environmental Certiﬁcation

In recent years, blockchain technology has emerged

as a powerful tool for ensuring the transparency, trace-

ability, and authenticity of data in agricultural and

environmental contexts (Kamble et al., 2020; Slama

et al., 2024; Nabli et al., 2025). Several blockchain-

based frameworks have been proposed, particularly

in the domains of organic certiﬁcation, supply chain

tracking, and customer engagement through tokeniza-

tion. While these systems have shown technical and

conceptual maturity, their design and goals differ sig-

niﬁcantly from applications intended to valorize ur-

ban natural heritage, such as remarkable trees.

One notable contribution is the SAFE platform

(Tegeltija et al., 2022), designed to automate and

secure organic agriculture certiﬁcation by combin-

ing IoT sensors and blockchain smart contracts. By

streamlining data collection and validation, SAFE re-

duces administrative burdens and enhances consumer

trust. However, the platform focuses on standardized

certiﬁcations in agricultural production and does not

address subjective or cultural valuation, which is cen-

tral to heritage tree recognition. Additionally, it lacks

support for non-fungible digital identities like NFTs

that can uniquely represent individual natural entities.

Another approach by Santos et al. (Santos

et al., 2023) introduces a blockchain-based system for

certifying agri-food harvests through ERC-standard

NFTs. Their solution emphasizes ﬁne-grained trace-

ability and anti-fraud measures within the food sup-

ply chain. Yet, it assumes tokens are consumable and

temporary, as they are burned post-sale—an approach

incompatible with the long-term tracking and preser-

vation needed for the digital identity of heritage trees.

Furthermore, the model focuses on fungible products,

whereas urban trees require rich, non-fungible repre-

sentations embedded with ecological, historical, and

visual data.

A third work explores smart agriculture assur-

ance through blockchain and IoT integration (Hasan

et al., 2024). It proposes an architecture involving

real-time sensors, oracles, IPFS, and permissioned

blockchain for sustainability certiﬁcations (e.g., or-

ganic, non-GMO). While technically robust, the sys-

Smart Urban Tree Valorization: An AI-Blockchain-Based Application for the Preservation of Remarkable Trees

611

tem targets dynamic agricultural conditions, not static

environmental assets. Its reliance on costly hardware,

complex infrastructure, and continuous data streams

makes it poorly suited for urban biodiversity valoriza-

tion, where the emphasis is on accessibility, longevity,

and public interaction rather than real-time monitor-

ing.

Finally, Hosseinalibeiki and Zaree (Hosseinal-

ibeiki and Zaree, 2023) present a model for NFT-

based loyalty programs in agribusiness, leveraging

smart contracts to reward customer behavior. Their

NFTs include metadata (photos, certiﬁcates) to re-

inforce personalization and trust. While conceptu-

ally closer to symbolic valorization, their system re-

mains rooted in commercial use cases and lacks fea-

tures critical to environmental education, public en-

gagement, or urban planning integration. The absence

of interactive maps, botanical classiﬁcation, or user-

centered ecological narratives limits its relevance to

urban tree heritage.

Across these studies, several limitations emerge in

relation to our objectives:

• Context mismatch: Most works focus on agri-

culture and product certiﬁcation, not on natural,

static assets like urban trees or their cultural and

ecological importance

• Temporal design: Many systems rely on

ephemeral tokens tied to transactions or seasonal

data, whereas our NFTs must persist over time to

reﬂect ongoing tree status and legacy

• Lack of rich metadata: Existing NFTs often

lack botanical, historical, or geographic attributes,

which are essential for tree valorization

• No citizen engagement layer: The reviewed sys-

tems do not include interactive, participatory plat-

forms involving municipal authorities, citizens, or

educators

• Overly complex architecture: Several solutions

rely on heavy technical stacks (e.g., IoT, IPFS,

CRM), which are ill-suited for lightweight, edu-

cational, and publicly accessible web/mobile plat-

forms.

In response, our work introduces a blockchain-

based system for certifying remarkable urban trees

through NFTs enriched with multi-dimensional meta-

data (e.g., age, species, photos, location, cultural sto-

ries). Certiﬁcation is triggered by a smart contract

upon validation by municipal authorities and stored

immutably on the blockchain. Unlike agricultural

systems, this framework is not designed for eco-

nomic traceability but for heritage recognition, citizen

awareness, and digital conservation.

3 SYSTEM DESIGN AND

ARCHITECTURE

The proposed application is a modular, AI- and

blockchain-based system designed for the digital val-

orization of remarkable urban trees. Its architecture

integrates three complementary technological pillars,

each addressing a critical aspect of the valorization

process. Figure 1 illustrates the high-level architec-

ture and interaction between system components.

• Web-Based Interface: This component provides

an interactive map interface allowing users, cit-

izens, urban planners, and environmentalists, to

explore and contribute geo-referenced data on re-

markable urban trees. It supports dynamic visual-

ization of tree locations, species, and health status,

fostering community awareness and engagement.

• AI-Driven Recommendation Chatbot: Integrated

with the geographic interface, the AI chatbot of-

fers personalized recommendations and informa-

tive responses regarding tree care, species identi-

ﬁcation, and environmental beneﬁts.

• Blockchain-Powered Certiﬁcation Engine: To en-

sure the integrity and transparency of tree-related

data, a blockchain certiﬁcation engine registers

and certiﬁes remarkable trees on a decentralized

ledger. Each certiﬁed tree is represented as a Non-

Fungible Token (NFT), enabling trusted digital

recognition and long-term data preservation.

Figure 1: Smart Urban Tree Valorization System Architec-

ture.

3.1 Web-Based Interface

The web-based interface serves as the central module

of the platform, offering users a structured, interac-

tive, and intuitive environment to explore and engage

with remarkable urban trees. Designed as a respon-

sive web-mobile application, this interface supports

the valorization, management, and digital certiﬁca-

tion of ecologically or culturally signiﬁcant trees. It

TISAS 2025 - Special Session on Trustworthy and Intelligent Smart Agriculture Systems: AI, Blockchain, and IoT Convergence

612

bridges the gap between citizens, urban planners, and

municipal authorities by providing accurate botani-

cal data, geolocation tools, and participatory features

within a uniﬁed interface.

Upon launching the platform, users are welcomed

by a public homepage presenting the goals of the

project and its environmental importance. Authenti-

cated users are granted access to personalized dash-

boards, with functionalities that vary depending on

their assigned roles (Regular User, Admin, Super Ad-

min). To clarify the interactions between the system

and its different user types, Figure 2 presents a use

case diagram illustrating the functionalities available

to each role.

Three primary actors are deﬁned: User, Admin,

and Super Admin, as well as two external compo-

nents: an AI Chatbot and a Blockchain Platform.

Each actor interacts with the system through a set of

functionally distinct use cases.

The Regular User interacts with the system

through several core functionalities, each correspond-

ing to speciﬁc use cases within the system’s architec-

ture:

• Register and Authenticate use cases: Users can

create a personal account and securely authenti-

cate themselves using standard login procedures.

This ensures personalized access and enables par-

ticipation in interactive features such as reviews

and chatbot queries.

• Explore Trees use case: The user can browse

the complete inventory of remarkable urban trees

through two complementary interfaces: an inter-

active map view (Explore Trees on Map use case)

and a structured list view (Explore Trees on List

use case). The map supports both OpenStreetMap

and satellite layers, with each tree represented by

a clickable marker that reveals detailed proﬁle in-

formation, including the common names, botani-

cal family, age, and height (See Figure 3). In the

list view, users can apply search and ﬁltering op-

tions to reﬁne the inventory based on various cri-

teria, such as species, age, or name.

• Submit Review use case: After exploring a tree,

users can submit a rating and review through the

system. Each review allows users to assess sev-

eral aspects, including perceived air quality, am-

bient noise level, cleanliness of the surroundings,

accessibility and visibility of the tree, and its ap-

parent health and maintenance status. Users can

also provide open comments and suggestions (See

Figure 4). This participatory use case encourages

public engagement and contributes to the collec-

tive knowledge about remarkable urban trees.

• Ask Question use case: This use case introduces

a generative AI chatbot that enables users to ask

natural language questions about remarkable ur-

ban trees. Through the included Respond to Tree

Inquiry use case, the chatbot interprets user in-

tent and generates context-aware responses, rang-

ing from factual information about speciﬁc trees

(e.g., species, age, ecological value) to personal-

ized recommendations (e.g., quiet spots, accessi-

ble trees). This interactive feature fosters public

engagement in the preservation and appreciation

of urban natural heritage.

In addition to the regular user’s core functionali-

ties, the system provides dedicated use cases for the

Admin role, including:

• Manage Tree Data use case: The Admin can add,

update, or delete tree proﬁles within the system.

This includes managing detailed information such

as scientiﬁc and common names, species, age,

height, trunk circumference, location, and upload-

ing representative photos to ensure accurate and

up-to-date tree records.

• Manage Users use case: The Admin oversees user

accounts by creating, updating, or removing users.

• Certify Tree use case: The Admin submits a certi-

ﬁcation request for a tree by verifying its attributes

and conﬁrming it meets preliminary criteria for

recognition as a remarkable urban tree. The ﬁ-

nal validation and approval of the certiﬁcation are

performed by the Super Admin, ensuring the cred-

ibility of the system’s inventory.

Beyond Regular User and Admin functionalities,

the system includes specialized use cases reserved for

the Super Admin role, responsible for high-level man-

agement tasks, including:

• Manage Admins use case: The Super Admin over-

sees admin accounts by creating, updating, or re-

moving admin users.

• Validate Tree Certiﬁcation use case: The Su-

per Admin reviews and approves tree certiﬁca-

tion requests submitted by admins. The certiﬁ-

cation process involves a secure interaction with

the Blockchain Platform, represented by the Cre-

ate Tree Certiﬁcate use case.

3.2 AI-Driven Recommendation

Chatbot

The AI-driven chatbot module constitutes a key com-

ponent of the proposed system, designed to assist

users in discovering remarkable urban trees accord-

ing to their preferences, location, or well-being in-

Smart Urban Tree Valorization: An AI-Blockchain-Based Application for the Preservation of Remarkable Trees

613

Figure 2: Use Case Diagram: Functionalities by User Role.

Figure 3: OpenStreetMap Interface Displaying Details of a

Selected Tree.

tentions. Unlike traditional static query interfaces,

this chatbot dynamically interprets natural language

input, offering interactive and personalized sugges-

tions (Nabli et al., 2024). As illustrated in Figure 5,

the chatbot adopts a hybrid, multi-layered architec-

ture combining three intelligent modules: a Natu-

ral Language Processing (NLP) module for under-

standing user intent, a fuzzy keyword-based retrieval

system powered by Fuse.js, and a semantic vector

search integrated with a Retrieval-Augmented Gen-

eration (RAG) pipeline.

3.2.1 Natural Language Processing (NLP)

The NLP module functions as the chatbot’s inter-

pretive front-end, transforming informal, often am-

biguous natural language into structured represen-

tations compatible with downstream search and re-

trieval mechanisms. This transformation unfolds

across several stages:

Preprocessing and Normalization: The user input

is ﬁrst normalized through lowercasing, whitespace

trimming, and punctuation removal. These prepro-

cessing steps reduce lexical variance and ensure uni-

formity, enabling robust matching against both key-

word and semantic indices.

Lexical and Syntactic Analysis: The system per-

forms tokenization (splitting input into words or

meaningful subunits) and applies part-of-speech tag-

ging to understand grammatical structure. Optional

named entity recognition (NER) identiﬁes geograph-

ical areas (e.g., “City Center”), common tree names

(e.g., “Jacaranda”), or well-being-related expressions

(e.g., “calm”, “shady”).

TISAS 2025 - Special Session on Trustworthy and Intelligent Smart Agriculture Systems: AI, Blockchain, and IoT Convergence

614

(a) Rating component (b) Comment submission section

Figure 4: User interface elements displaying the rating component (left) and the comment submission section (right).

Figure 5: Hybrid architecture of the AI-driven recommen-

dation chatbot.

Intent Detection and Slot Filling: The NLP engine

determines the user’s intent (e.g., search for shade,

locate a speciﬁc tree, get a recommendation) using

a lightweight rule-based or statistical classiﬁer. From

this intent, semantic slots are ﬁlled with extracted val-

ues such as tree features (e.g., shade, beauty), user

goals (e.g., relaxation, walking), and locations.

Context Tracking: For multi-turn conversations, a

memory buffer can be maintained to carry forward

unmentioned context. For instance, if a user ﬁrst asks

about ”shady trees in Boujaafar” and later asks ”and

in Sousse?”, the module infers that the user is still in-

terested in shady trees.

This module ensures that linguistic diversity in

user queries is faithfully mapped to the chatbot’s in-

ternal data schema, enabling accurate and consistent

processing.

3.2.2 Fuzzy Keyword Matching

This module addresses one of the practical challenges

in real-world chatbot systems: dealing with noisy,

imprecise, or misspelled user inputs. Unlike ex-

act keyword matching, which fails when there are

spelling errors or lexical variations, fuzzy matching

algorithms tolerate such inconsistencies by allowing

approximate matches.

Dataset Preparation: The system deﬁnes a struc-

tured dataset comprising frequently asked questions,

common tree attributes, and predeﬁned recommenda-

tion intents. Each item in the dataset is represented as

a JSON object with searchable ﬁelds, such as title,

description, or species.

Fuse.js Initialization: Fuse.js

is a popular

JavaScript library for efﬁcient fuzzy searching over a

predeﬁned corpus. A Fuse.js instance is instantiated

using the prepared dataset, with conﬁguration options

https://github.com/krisk/Fuse

Smart Urban Tree Valorization: An AI-Blockchain-Based Application for the Preservation of Remarkable Trees

615

tailored to the domain. The keys parameter speciﬁes

which ﬁelds to search, while the threshold value

controls the sensitivity of the fuzzy matching. A

lower threshold yields more precise matches, while a

higher value allows greater tolerance for variation.

const fuse = new Fuse(treeDataset,

keys: [’title’, ’description’],

threshold: 0.4 );

User Query Matching: Once the user’s input is

processed by the NLP module, the remaining query

string is passed to the Fuse.js search function. Fuse.js

computes similarity scores between the input and

dataset entries using approximate string matching al-

gorithms (e.g., Levenshtein distance).

Candidate Selection and Ranking: The output of

the search is a ranked list of candidate items with as-

sociated conﬁdence scores. The chatbot selects the

top match (or top-k) if the score falls below a prede-

ﬁned relevance threshold, ensuring that only mean-

ingful results are presented to the user.

Response Generator: The best-matched item is

mapped to a corresponding response template or ac-

tion, such as displaying detailed information about a

tree species or suggesting trees based on environmen-

tal ﬁlters. If no sufﬁciently relevant match is found,

the system gracefully prompts the user to rephrase the

query.

3.2.3 Semantic Search and RAG Engine

When user queries are abstract, personalized, or emo-

tionally expressive, the chatbot escalates to a se-

mantic retrieval pipeline underpinned by Retrieval-

Augmented Generation (RAG) techniques (Lewis

et al., 2020).

Semantic Vectorization: The user query is en-

coded into a dense vector representation using a

pre-trained transformer-based model (e.g., Sentence-

BERT or OpenAI embeddings) that captures the se-

mantic meaning of the query (Reimers and Gurevych,

2019)).

FAISS-Based Similarity Search: The chatbot

maintains a FAISS (Facebook AI Similarity Search)

index populated with vector embeddings of curated

knowledge chunks from environmental reports, tree

guides, and urban ecology texts (Johnson et al., 2019).

The semantic vector of the user query is compared

against this index using approximate nearest neighbor

search.

Context Prompt Construction: If relevant docu-

ments are retrieved, they are concatenated with the

original query to form a prompt for the language

model. This grounding ensures that generated re-

sponses are factual and context-speciﬁc, reducing the

likelihood of hallucinated or vague output (Shuster

et al., 2021).

Fallback Mechanism: In cases where no relevant

chunks are retrieved (e.g., ambiguous or out-of-scope

queries), the system invokes a fallback mode, allow-

ing the LLM to generate responses from its internal

pretrained knowledge without retrieved context.

Local LLM Response Generation: The structured

prompt is fed into a local language model (e.g., us-

ing the Ollama API with LLaMA or GPT models) for

ﬁnal response generation. The model produces ﬂu-

ent and semantically relevant responses that reﬂect the

user’s intent and contextual data.

Response Generator: The output is reformatted for

display and sent back to the user interface. Optional

modules adjust tone, verbosity, and terminology for

clarity and appropriateness.

3.3 Blockchain-Based NFT

Certiﬁcation Engine

To enhance the credibility, traceability, and long-term

preservation of remarkable trees in urban environ-

ments, the proposed system integrates a blockchain-

based certiﬁcation engine. This module employs the

Ethereum blockchain and the ERC-721 token stan-

dard to issue a non-fungible token (NFT) for each

tree ofﬁcially validated by authorized entities (e.g.,

municipal or environmental authorities). Each NFT

serves as a veriﬁable, immutable digital certiﬁcate

that encapsulates the tree’s scientiﬁc, cultural, and

spatial attributes, thereby promoting transparent eco-

logical governance and public engagement (Taher-

doost, 2022).

3.3.1 NFT Smart Contract Design

The smart contract is written in Solidity and built

upon the OpenZeppelin implementation of the ERC-

721 standard. It deﬁnes a minting function restricted

to the contract owner (i.e., the certifying authority)

and includes URI management to associate NFTs

with their corresponding off-chain metadata. Figure 6

illustrates the smart contract used to mint tree certiﬁ-

cation NFTs.

TISAS 2025 - Special Session on Trustworthy and Intelligent Smart Agriculture Systems: AI, Blockchain, and IoT Convergence

616

Figure 6: Solidity Smart Contract for Minting Tree Certiﬁ-

cation NFTs Using the ERC-721 Standard.

Each time a tree is validated, a transaction is

submitted by the authority, triggering the mintNFT()

function. This function mints a new NFT and links it

to a metadata URI hosted on IPFS. The smart contract

is deployed on Sepolia using test ETH.

The certiﬁcation engine is deployed on the Sepolia

testnet, a publicly accessible Ethereum test network,

and supports interaction via MetaMask

, a widely

adopted browser-based cryptocurrency wallet. Meta-

data describing the tree is stored off-chain using the

InterPlanetary File System (IPFS), while the corre-

sponding URI is permanently linked on-chain within

the NFT contract.

3.3.2 Metadata Structure and Off-Chain

Storage

The NFT metadata follows a JSON schema and in-

cludes both botanical and contextual information,

such as location, species, dimensions, historical sig-

niﬁcance, and issuer identity. This data is uploaded

to IPFS using decentralized pinning services (e.g.,

Pinata), and the resulting content identiﬁer (CID) is

used to construct the metadata URI. A typical NFT

metadata structure is shown in the following JSON

Figure 7.

Figure 7: Example of NFT Metadata Structure Represent-

ing a Certiﬁed Tree (in JSON Format).

The metadata is publicly accessible through the

https://metamask.io/

generated IPFS gateway link and linked permanently

to the on-chain token via its tokenURI.

3.3.3 Certiﬁcation Workﬂow

The NFT certiﬁcation process follows a multi-stage

workﬂow:

Tree Submission: A tree is submitted via the sys-

tem interface by a veriﬁed user, such as a munici-

pal agent or environmental researcher. The submis-

sion includes geolocation, physical attributes, pho-

tographs, and cultural or historical context. The tree

information is stored in a MongoDB database, ensur-

ing persistence and efﬁcient querying. Upon success-

ful submission, a certiﬁcation notiﬁcation is automat-

ically sent to the authorized administrators.

Expert Validation: Authorized administrators

evaluate certiﬁcation submissions based on prede-

ﬁned criteria (e.g., age, rarity, aesthetic or historical

value). Upon validation, a structured metadata JSON

ﬁle is generated and securely uploaded to the IPFS

for decentralized storage.

NFT Minting: Using MetaMask, the administrator

connects to the Ethereum Sepolia network and signs a

transaction to mint the NFT using the smart contract.

The IPFS metadata URI is passed to the mintNFT()

function, and the token is issued to the municipality’s

digital wallet.

Figure 8 displays the detailed record of the mint-

ing transaction on Etherscan, illustrating the success-

ful execution of the mintNFT() function, including the

transaction hash, block number, sender and recipient

addresses, gas fees, and the status of the operation,

providing transparent and veriﬁable proof of NFT cre-

ation on the blockchain.

Figure 8: Transaction Details of Tree NFT Minting on Se-

polia Etherscan.

Smart Urban Tree Valorization: An AI-Blockchain-Based Application for the Preservation of Remarkable Trees

617

Public Certiﬁcation Access: The certiﬁed tree is

displayed on the application’s map interface with a

“Veriﬁed” badge. The blockchain certiﬁcate is acces-

sible through a link to Sepolia Etherscan and may also

be viewed in MetaMask or compatible NFT explorers.

4 IMPLEMENTATION AND

EVALUATION

To assess the feasibility and effectiveness of our ap-

proach, we conducted preliminary experiments evalu-

ating both the AI chatbot and blockchain certiﬁcation

components.

4.1 Implementation Insights

The proposed system is implemented using a mod-

ern web technology stack designed for modularity,

scalability, and user interactivity. The frontend in-

terface is developed with React.js, a widely adopted

JavaScript framework that facilitates dynamic user in-

terfaces and seamless chatbot interactions. The AI

chatbot module combines Python with the Transform-

ers library (Hugging Face) and Ollama for local de-

ployment of large language models, enabling natu-

ral language understanding, intent classiﬁcation, and

response generation. To support ﬂexible keyword-

based retrieval, it incorporates Fuse.js, allowing fuzzy

matching between user input and the indexed urban

tree dataset. The backend API is built using Node.js

with Express.js, handling business logic, database

interactions, and coordination with blockchain ser-

vices. For data persistence, MongoDB is used to

store non-sensitive tree information and user interac-

tion logs. The blockchain layer is developed using

Solidity smart contracts deployed on the Ethereum

Sepolia test network, with IPFS employed for decen-

tralized storage of tree metadata linked to NFTs. De-

velopment and testing workﬂows utilize Visual Studio

Code as the primary IDE, with Remix IDE

used for

smart contract coding and veriﬁcation. Figure 9 illus-

trates the overall software system architecture, high-

lighting the interactions between the user interface,

AI chatbot, backend services, blockchain network,

and data storage components.

https://remix.ethereum.org/

Figure 9: Overall software system architecture.

4.2 Evaluation

4.2.1 Evaluation of Chatbot Response Relevance

The AI module is evaluated based on its ability to ac-

curately understand and respond to user queries. The

evaluation used 250 test queries spanning informa-

tional and action-oriented requests, reﬂecting ecolog-

ical, recreational, and emotional user intents to sim-

ulate realistic chatbot interactions. To measure the

chatbot’s effectiveness in matching user intents and

retrieving relevant tree information, we use standard

metrics including Precision (the proportion of rele-

vant trees correctly returned among all retrieved re-

sults), Recall (the proportion of relevant trees cor-

rectly identiﬁed among all actual relevant items), and

F1-Score (the harmonic mean of precision and recall,

reﬂecting overall performance). Table 1 compares

the performance of our AI-driven tree recommenda-

tion system against a keyword-only retrieval baseline,

highlighting the added value of our approach.

• Precision =

T P

T P+FP

• Recall =

T P

T P+FN

• F1 − Score = 2 ×

Precision×Recall

Precision+Recall

Where: TP = True Positives, FP = False Positives, FN

= False Negatives

Table 1: Comparison between AI-driven Tree Recommen-

dation System and Keyword-only Retrieval Baseline.

Metric AI-driven System Keyword-only Retrieval

Precision 0.91 0.58

Recall 0.87 0.62

F1-Score 0.89 0.60

4.2.2 Evaluation of Personalization and User

Satisfaction

To evaluate the chatbot’s personalization capacity and

user experience, we conducted a user study with

30 participants. The results are derived from post-

interaction surveys and behavior logs, using metrics

TISAS 2025 - Special Session on Trustworthy and Intelligent Smart Agriculture Systems: AI, Blockchain, and IoT Convergence

618

such as Click-Through Rate (CTR), which indicates

how often users interacted with tree suggestions, and

the User Satisfaction Score (USS), which reﬂects the

perceived usefulness and relevance of the responses.

• C T R =

Numbero fC licks

Numbero f Recommendations

• USS =

∑

(UserRatings)

TotalUsers

(scale : 1to5)

Table 2: User Interaction and Satisfaction Metrics.

Metric Value

CTR 0.68

USS 4.3 / 5

4.2.3 Evaluation of Blockchain Certiﬁcation

Integrity

The blockchain mechanism is evaluated based on its

ability to ensure data authenticity, traceability, and

resistance to tampering. We employ metrics such

as Data Immutability Rate (the proportion of certiﬁ-

cates that remain unaltered after issuance), Transac-

tion Latency (the average time between submission

and successful on-chain registration), and Certiﬁca-

tion Uniqueness (ensuring that each remarkable tree

is uniquely represented on-chain through a NFT).

• Data Immutability Rate =



ImmutableRecords

TotalRecords



× 100

• Transaction Latency =

Average time taken to validate and store a certiﬁcate

• Certiﬁcation Uniqueness =

#o fUniqueNFT s

#o fC erti f iedTrees

Table 3: Blockchain-Based Certiﬁcation Evaluation Met-

rics.

Metric Value

Data Immutability Rate 100%

Transaction Latency ∼ 12 seconds

Certiﬁcation Uniqueness 1.00

4.2.4 Discussion

The evaluation shows that the proposed system pro-

vides highly relevant and personalized responses

through its chatbot, achieving strong performance

metrics (Precision 0.91, F1-score 0.89), which in-

dicate the reliability of AI-generated recommenda-

tions. Compared to a keyword-only retrieval base-

line (Precision 0.58, F1-score 0.60), the AI-driven ap-

proach demonstrates a clear improvement, highlight-

ing the value of personalized AI recommendations

over simpler query-matching methods. The relatively

high CTR and satisfaction scores suggest that users

found the interaction both intuitive and valuable. On

the blockchain side, the system demonstrates full im-

mutability and uniqueness of tree certiﬁcates, ensur-

ing data trust and accountability. The average latency

(∼ 12 seconds) is within acceptable thresholds for

public blockchain networks. These results validate

the integration of AI and blockchain as an effective

and trustworthy solution for urban tree valorization.

5 CONCLUSION

This paper presented an innovative AI- and

blockchain-based system designed to enhance

the preservation, visibility, and citizen engagement

surrounding remarkable urban trees. By combin-

ing an intelligent conversational agent that guides

users through personalized interactions with a

secure blockchain-powered certiﬁcation process,

the proposed application bridges the gap between

technological innovation and ecological heritage

valorization. The integration of non-fungible tokens

(NFTs) ensures traceability and digital permanence

of each tree’s unique identity, while the AI mod-

ule fosters awareness and user participation by

recommending trees based on speciﬁc ecological

or emotional beneﬁts. Implementation insights

and evaluation conﬁrmed its feasibility and user

acceptance as a participatory urban ecology tool.

Future research will extend this study with a larger

participant pool to enhance the generalizability of the

ﬁndings and provide more robust evidence for the ob-

served effects. Additionally, integrating supplemen-

tary environmental data sources—such as IoT sensors

for real-time monitoring of tree health or microcli-

matic conditions—could further enrich the system’s

insights. Enhancements to the AI module, including

multimodal interactions (e.g., voice or image input),

could improve accessibility and foster greater user en-

gagement.

REFERENCES

Abdelghani, S. B., Nabli, H., Djemaa, R. B., and Sliman,

L. (2023). Blockchain-integrated technologies to ad-

dress counterfeit drugs in the pharmaceutical supply

chain. In International Conference on Hybrid Intelli-

gent Systems, pages 292–301. Springer.

Abdelhamid, M., Sliman, L., Djemaa, R. B., and Perboli, G.

(2024). A review on blockchain technology, current

challenges, and ai-driven solutions. ACM Computing

Surveys, 57(3):1–39.

Arora, B., Chaudhary, D., Satsangi, M., Yadav, M., Singh,

L., and Sudhish, P. (2020). Agribot: a natural lan-

guage generative neural networks engine for agricul-

Smart Urban Tree Valorization: An AI-Blockchain-Based Application for the Preservation of Remarkable Trees

619

tural applications. In International Conference on

Contemporary Computing and Applications (IC3A),

pages 28–33. IEEE.

Chathurya, C., Sachdeva, D., and Arora, M. (2023). Agri-

culture chatbot (agribot) using natural language pro-

cessing. In 2023 14th International Conference on

Computing Communication and Networking Tech-

nologies (ICCCNT), pages 1–5. IEEE.

Hasan, H., Musamih, A., Salah, K., Jayaraman, R., Omar,

M., Arshad, J., and Boscovic, D. (2024). Smart agri-

culture assurance: Iot and blockchain for trusted sus-

tainable produce. Computers and Electronics in Agri-

culture, 224:109184.

Hosseinalibeiki, H. and Zaree, M. (2023). A blockchain

based solution to improve loyalty program with nft

in agribusiness. Journal of Smart Environments and

Green Computing, 3(4):127–146.

Johnson, J., Douze, M., and J

egou, H. (2019). Billion-scale

similarity search with gpus. IEEE Transactions on Big

Data, 7(3):535–547.

Kamble, S., Gunasekaran, A., and Sharma, R. (2020). Mod-

eling the blockchain enabled traceability in agriculture

supply chain. International Journal of Information

Management, 52:101967.

Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin,

V., Goyal, N., K

uttler, H., Lewis, M., Yih, W.,

Rockt

aschel, T., et al. (2020). Retrieval-augmented

generation for knowledge-intensive nlp tasks. Ad-

vances in neural information processing systems,

33:9459–9474.

Li, X. and Wu, J. (2022). Public engagement in urban green

infrastructure planning: A critical review. Landscape

and Urban Planning, 225:104412.

Marla, A., Paul, R., Saha, A., Basha, N., and Anandhakr-

ishnan, B. (2023). An agrobot: Natural language pro-

cessing based chatbot for farmers. In 4th International

Conference on Smart Electronics and Communication

(ICOSEC), pages 1235–1241. IEEE.

Mostaco, G., Souza, I. D., Campos, L., and Cugnasca, C.

(2018). Agronomobot: a smart answering chatbot ap-

plied to agricultural sensor networks. In 14th interna-

tional conference on precision agriculture, volume 24,

pages 1–13.

Mullaney, J., Lucke, T., and Trueman, S. (2015). The ben-

eﬁts of large species urban trees: A review of ecosys-

tem services. Urban Forestry & Urban Greening,

14(4):607–614.

Nabli, H., Djemaa, R. B., and Amor, I. B. (2024). Im-

proved clustering-based hybrid recommendation sys-

tem to offer personalized cloud services. Cluster Com-

puting, 27(3):2845–2874.

Nabli, H., Ghannem, A., Djemaa, R. B., and Sliman, L.

(2025). How innovative technologies shape the future

of pharmaceutical supply chains. Computers & Indus-

trial Engineering, 199:110745.

Nandhini, J., Anuratha, K., Sangeetha, K., and Jaswant,

K. (2021). Smart tree management with biodiversity

preservation using image processing and blockchain

technology. In 2021 International Conference on Sys-

tem, Computation, Automation and Networking (IC-

SCAN), pages 1–6. IEEE.

Niranjan, P., Rajpurohit, V., and Malgi, R. (2019). A survey

on chat-bot system for agriculture domain. In 2019 1st

International Conference on Advances in Information

Technology (ICAIT), pages 99–103. IEEE.

Ong, R., Raof, R., Sudin, S., and Choong, K. (2021). A re-

view of chatbot development for dynamic web-based

knowledge management system (kms) in small scale

agriculture. In Journal of Physics: Conference Series,

volume 1755, page 012051. IOP Publishing.

Pravinkrishnan, K., Balasundaram, P., and Kalinathan, L.

(2022). An overview of chatbots using ml algorithms

in agricultural domain. International Journal of Com-

puter Applications, 975(11):15–22.

Radhakrishnan, R., Tripathi, A., and Singh, V. (2022).

Chatbot-based citizen engagement in smart cities: A

framework for personalized urban services. Journal

of Urban Technology, 29(1):3–22.

Reimers, N. and Gurevych, I. (2019). Sentence-bert: Sen-

tence embeddings using siamese bert-networks. arXiv

preprint arXiv:1908.10084.

Santos, R. D., Pantoni, R., and Torrisi, N. (2023).

Blockchain tokens for agri-food supply chain. Journal

of Engineering Research and Sciences, 2(2):15–23.

Shuster, K., Poff, S., Chen, M., Kiela, D., and Weston, J.

(2021). Retrieval augmentation reduces hallucination

in conversation. arXiv preprint arXiv:2104.07567.

Slama, W. B., Charroux, B., Sliman, L., and Djemaa, R. B.

(2024). Towards blockchain like soa. In International

Conference on Web Information Systems Engineering,

pages 301–311. Springer.

Taherdoost, H. (2022). Non-fungible tokens (nft): A sys-

tematic review. Information, 14(1):26.

Tegeltija, S., Dejanovi

c, S., Feng, H., Stankovski, S., Os-

toji

c, G., Ku

cevi

c, D., and Marjanovi

c, J. (2022).

Blockchain framework for certiﬁcation of organic

agriculture production. Sustainability, 14(19):11823.

Torres, K. and McDonald, R. (2021). Urban trees and hu-

man health: A scoping review. International Jour-

nal of Environmental Research and Public Health,

18(9):4645.

Treiblmaier, H. (2018). The impact of blockchain on the

supply chain: A theory-based research framework and

a call for action. Supply Chain Management: An In-

ternational Journal, 23(6):545–559.

Usip, P., Udo, E., Asuquo, D., and James, O. (2022). A

machine learning-based mobile chatbot for crop farm-

ers. In International Conference on Electronic Gov-

ernance with Emerging Technologies, pages 192–211.

Springer.

Varveris, D., Styliadis, A., Xoﬁs, P., and Dimen, L. (2023).

Distributed and collaborative tree architecture: A low-

cost experimental approach for smart forest monitor-

ing. Baltic Journal of Modern Computing, 11(4):653–

685.

Yoon, J. and Fischer, J. (2023). The importance of cultur-

ally signiﬁcant trees in urban biodiversity strategies.

Urban Forestry & Urban Greening, 80:127828.

TISAS 2025 - Special Session on Trustworthy and Intelligent Smart Agriculture Systems: AI, Blockchain, and IoT Convergence

620