🌿 Soil Monitoring Dashboard

📊 🪴🔆 1

Abstract

This study addresses global water scarcity by proposing an innovative IoT-driven precision irrigation system utilizing advanced machine learning models for sustainable agriculture. The aim is to enhance irrigation practices by deploying low-cost ESP32 IoT sensors to continuously collect environmental data, including soil moisture, temperature, humidity, and atmospheric pressure. This data feeds into a robust AWS cloud infrastructure involving IoT Core, DynamoDB, Lambda functions, and serverless Docker-based ML forecasting pipelines using LSTM and Prophet models. Results demonstrate significant improvements in irrigation scheduling accuracy, achieving approximately 95% prediction accuracy with a marked reduction in water consumption by up to 17%. These findings underscore the substantial potential of integrated IoT and machine learning systems for boosting agricultural productivity while conserving vital water resources.

Introduction

Global agriculture currently faces unprecedented challenges due to climate change and escalating water scarcity. Increasing variability in rainfall patterns, extended drought periods, and rising temperatures threaten crop productivity, necessitating more intelligent resource management strategies. Precision irrigation emerges as a crucial solution to these challenges, aiming to optimize water application by accurately responding to real-time environmental conditions. Effective precision irrigation can significantly conserve water resources, improve crop yields, and support agricultural sustainability. This research specifically targets precision irrigation by leveraging cutting-edge Internet of Things (IoT) and machine learning (ML) technologies. Our objectives include: Deployment of IoT Sensors: Implement low-cost ESP32-based IoT sensors to continuously gather critical environmental metrics such as soil moisture, ambient temperature, atmospheric humidity, and pressure. Real-time Predictive Irrigation Scheduling: Develop a robust forecasting model utilizing advanced ML algorithms, specifically Long Short-Term Memory (LSTM) and Prophet models, deployed via AWS Lambda and serverless computing frameworks. Impact Assessment: Evaluate the effectiveness of precise irrigation schedules on water-use efficiency and overall plant health, analyzing improvements compared to traditional fixed-schedule irrigation methods. Through this integrated approach, the study aims to demonstrate how precision agriculture supported by IoT-driven analytics and predictive modeling can serve as a scalable, efficient, and sustainable solution for global agricultural practices.

📊 🪴🔆 2

Literature Review

IoT in Precision Agriculture

Internet of Things (IoT) technologies have significantly transformed precision agriculture by enabling real-time data collection, analysis, and management of agricultural processes. For instance, IoT-based systems utilize sensors to monitor environmental parameters, including soil moisture, temperature, and humidity, providing valuable insights for resource optimization. However, existing systems frequently encounter limitations, notably high implementation costs, inadequate sensor accuracy, and scalability issues. Costly proprietary technologies restrict accessibility for small-scale farmers, whereas sensor accuracy directly affects the reliability of collected data and subsequent decision-making (Ray, 2017; Wolfert et al., 2017).

Soil Moisture Prediction Models

Accurate prediction of soil moisture is critical for optimal irrigation scheduling. Traditional models often use regression analysis and other statistical methods but tend to underperform under changing environmental conditions due to their inability to handle non-linear relationships and temporal dynamics effectively (Mittelbach & Seneviratne, 2012). Additionally, conventional models typically offer limited forecasting lead-times, hindering proactive management decisions. The substantial computational resources required by some advanced models further limit their applicability, especially for real-time, practical scenarios in resource-constrained environments (Vereecken et al., 2014).

ML Applications in Irrigation Scheduling

Machine learning (ML) has recently shown considerable promise in agricultural forecasting, with deep learning approaches such as Long Short-Term Memory (LSTM) and Prophet gaining attention for their forecasting capabilities. LSTM networks are particularly suited for sequential data and effectively capture complex temporal dependencies, making them ideal for environmental data forecasting. Prophet models excel in handling non-linear and seasonal data, improving prediction reliability in agricultural contexts. Recent studies demonstrate that ML-driven models significantly outperform traditional approaches in irrigation scheduling by enhancing prediction accuracy and resource efficiency (Adeyemi et al., 2018; Taylor & Letham, 2018).

Research Gap & Motivation

Despite these advancements, a notable research gap exists in the comprehensive integration of IoT sensors, cloud-based infrastructures, and sophisticated ML forecasting models into a unified system capable of delivering real-time actionable recommendations for irrigation management. Many existing solutions lack effective end-to-end integration, resulting in practical deployment challenges and limited efficiency gains. Addressing this gap, our study proposes an integrated IoT, cloud-computing, and ML framework aimed at delivering precise, real-time irrigation recommendations, thus overcoming the identified limitations of current methodologies and enhancing agricultural productivity and sustainability (Ferrández-Pastor et al., 2016; Pham & Stack, 2018).

📊 🪴🔆 3

Methodology

Study Site and Experimental Setup

The experimental study was conducted in a controlled garden-based environment representative of temperate climate conditions, with typical diurnal variation in temperature and humidity. The site primarily featured moisture-sensitive vegetation and shallow-rooted crops to evaluate the system’s responsiveness to changing environmental conditions. The microclimate included moderate daily fluctuations in temperature (ranging from 15 °C to 30 °C), humidity (40%–90%), and intermittent rainfall, making it ideal for testing adaptive irrigation strategies.

IoT Infrastructure

The core sensing infrastructure comprised ESP32-based microcontrollers interfaced with five environmental sensors:

Cloud Architecture

The system was built using a lightweight, scalable AWS-based cloud architecture: AWS IoT Core acted as the MQTT broker, facilitating secure ingestion of sensor data in real-time. AWS DynamoDB served as the primary NoSQL database to store time-stamped sensor readings with millisecond precision and device metadata. AWS Lambda functions, containerized using Docker and deployed through Amazon ECR, performed data cleaning, transformation, and time-series forecasting using ML models. Daily forecast plots and data snapshots were cached in an Amazon S3 bucket for frontend access and archival. A Flask web application hosted on an EC2 instance delivered the dashboard interface. It used Chart.js to render dynamic visualizations, including moisture trends, VPD calculations, and 24-hour soil forecast graphs. This modular architecture ensured separation of data ingestion, processing, storage, and visualization, enabling easy scalability and low-latency insights delivery.

📊 🪴🔆 4

Canopy Closure Difference (Ccd)

CCD deviations signal early stress and yield impacts; tracking guides targeted interventions to maintain crop health (Zhang et al., 2020).

📊 🪴🔆 5

Soil Deficit (soil_deficit_mm)

Soil deficit below field capacity indicates irrigation necessity; trend analysis helps predict drought risk and optimize water use.

📊 🪴🔆 6

Custom Insight

Insert your additional research or tips here for this cycle of four charts.

📊 🪴🔆 7

Custom Insight

Insert your additional research or tips here for this cycle of four charts.