Metagenomic studies involve collecting soil samples, particularly from the rhizosphere, where soil microbes and root secretions interact (Weaver, 1994; Brooks, 2015). Total DNA is extracted from the samples using kits like Genejet Soil DNA Kit (Thermo Fisher) or Fast DNA SPIN Kit (MP Biochemicals). DNA is then enzymatically fragmented using library preparation kits such as Nextera Tagmentation (Illumina) or Fragmentase (New England Biolabs), with alternative methods including acoustic shearing, sonication, and others (Sabale et al., 2020). DNA concentration and purity are measured using Qubit and Nanodrop, while integrity is assessed via agarose gel electrophoresis or Agilent TapeStation. The DNA fragments are cloned into bacterial plasmids, featuring elements like an origin of replication, restriction sites, selective markers, and cloning sites (Granjou and Phillips, 2019). Fragments are analyzed using a fragment analyzer for quality and quantity. Sequencing is conducted on platforms like Illumina, Pyrosequencing, Nanopore, and PacBio. Post-sequencing, data are de-multiplexed and analyzed (Martin, 2011; Oulas et al., 2015; Mahmoud et al., 2019; Zhang et al., 2021).

Pre-processing of soil metagenomics data begins with quality control to filter out low-quality reads and remove adapter sequences. Tools such as UCHIME, MG-RAST, RDP tools (Bolger et al., 2014), KTrim, Trim Galore, and Trimmomatic (Sun, 2020) are utilized for these tasks, ensuring sequences are trimmed for uniform length and quality. Following trimming, reads are further filtered to exclude sequences below specified length thresholds, and errors are corrected while polymerase chain reaction (PCR) duplicates are removed to enhance accuracy. Denoising of metagenomic data is achieved using platforms like MOTHUR and QIIME 2, with UCHIME used for detecting and eliminating chimeric sequences (Santamaria et al., 2018). Post-processing involves grouping reads by unique barcodes, removing primers, and employing tools such as Taxator-tk (Dröge et al., 2015) and MEGAHIT (Liu et al., 2015) for further taxonomic and functional classification. Recent advancements in de novo assemblers like Meta-IDBA, metaSPAdes, Ray Meta, and Contig Extender allow for assembly of metagenomic reads into contigs, particularly beneficial for sequencing novel microbial genomes without prior reference sequences (Peng et al., 2011; Boisvert et al., 2012; Kumar et al., 2018; Deng and Delwart, 2021). Subsequently, reads are aligned to reference databases or assembled into contigs for comprehensive taxonomic and functional analysis, facilitating robust interpretation of soil microbial community data.

Metagenomic data processing forms the basis for taxonomic and functional profiling, essential for understanding microbial communities in soil. Tools like Krona, MEGAN, and phyloseq in R visualize taxonomic diversity and abundance (Huson et al., 2007; Ondov et al., 2011; McMurdie and Holmes, 2013). Functional analysis begins with gene prediction using tools such as Prodigal and MetaGeneMark, followed by annotation via KEGG, COG, and Pfam databases using eggNOG-mapper and InterProScan (Hyatt et al., 2010; Zhu et al., 2010). Pathway reconstruction is achieved using KEGG and MetaCyc databases with HUMAnN2 (Franzosa et al., 2018) and PathoScope (Hong et al., 2014), while functional capabilities are predicted by tools like PICRUSt, and Tax4Fun (Langille et al., 2013; Sun et al., 2020). Statistical analyses, including DESeq2, edgeR, PCA, and NMDS, identify and visualize differential taxa or functions. Integration of taxonomic and functional data, along with network and ecological models, further explores microbial interactions and their ecological roles (Robinson et al., 2010).

Common microbial species isolated from rhizosphere soil using metagenomic approaches include Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, and others (Babalola, 2010; Santos et al., 2019; Prasad and Zhang, 2022). Table 1 highlights microbial species and their impact on soil quality. By promoting beneficial microbes that fix nitrogen or solubilize phosphorus, farmers can reduce reliance on synthetic fertilizers, enhancing crop productivity, profits, and sustainability (Mendes et al., 2013). Enzymes like sulfatases, dehydrogenases, and phosphatases improve soil fertility, crop growth, and yield, reducing pesticide use (Peng et al., 2018). Metagenomics also supports the development of biofertilizers and microbial inoculants for agriculture and offers cost-effective alternatives to traditional soil remediation methods (Philippot et al., 2013, 2024). Several studies have demonstrated the economic advantages of microbial inoculants in agriculture. For instance, recurrent pre-sowing applications of Pseudomonas fluorescens have significantly boosted maize growth, reducing reliance on costly chemical fertilizers (Papin et al., 2024). Investigations into microbiota responses to nitric oxide regulation in Arabidopsis thaliana highlight the potential for enhancing crop productivity through optimized plant-microbe interactions (Berger et al., 2024). Moreover, trials using affordable microbial nutrient solutions have underscored their role in improving food security while offering cost-effective alternatives to conventional agricultural practices (van der Velde et al., 2013).

Aerial platforms such as UAVs or drones generally provide higher spatial resolution (< 5 meters) images compared to satellites (Bochtis et al., 2023). Thus, UAVs and other ground-based platforms offer greater flexibility in providing images at fine spatial and temporal resolutions (more frequently) or as needed. The hydrologic and climatic parameters—such as soil organic carbon, soil moisture, soil characteristics, the normalized Difference Vegetation Index (NDVI), leaf area index (LAI), groundwater, and rainfall—as well as the health of the vegetation and soil are monitored using unmanned aerial vehicles (UAVs) (Zhang et al., 2022).

Equipped with various sensors, UAVs perform specialized tasks: multispectral sensors capture plant health data in specific light wavelengths, thermal cameras detect temperature changes to identify irrigation or pest issues, and LIDAR creates detailed topographic maps for land and water management (Maddikunta et al., 2021; Tahir et al., 2023). However, global adoption of drone technology varies due to differing legal, financial, and physical conditions across countries. Supplementary Table S1 lists the various drones that are recognized for their capabilities in precision farming and improved crop management, ensuring compliance with safety and operational standards of respective aviation authorities.

Data collection and preprocessing in precision agriculture (PA) involve drones and IoT sensors, creating a comprehensive dataset. IoT sensors, such as Decagon EC-5 for soil, Davis Vantage Pro2 for weather, and GreenSeeker for crops, gather key data on moisture, temperature, pH, nutrients, climate, and crop health (García et al., 2020; Fuentes and Chang, 2022). Supplementary Table S2 highlights IoT sensors employed in precision agriculture, emphasizing their importance in enabling real-time, data-driven farming solutions. Connectivity networks like LoRaWAN, NB-IoT, and 5G transmit this data through gateways like Kerlink Wirnet Station. Flight planning software like DJI Ground Station Pro and drones like DJI Phantom 4 RTK capture aerial images (KriŽanović et al., 2023), followed by georeferencing and image stitching using tools like Agisoft Metashape. Noise reduction in software like Pix4Dmapper enhances image quality, while cloud computing supports secure data storage and real-time analysis (Debauche et al., 2022).

Precision farming is widely adopted due to the power of AI, driven by machine learning (ML) and deep learning (DL) (Khan et al., 2022; Ojo and Zahid, 2022; Hashmi and Kesakr, 2023). ML models analyze UAV-captured images, and AI-enabled farm management systems use sensor data to provide real-time recommendations for farmers. Machine learning (ML) approaches like Support Vector Machines (SVM) and Random Forests were applied for soil management to analyze soil temperature, moisture, and drying patterns (Liakos et al., 2018; Sharma et al., 2021). Models utilizing Decision Trees and Neural Networks were created to predict soil pH and fertility (Suchithra and Pai, 2020), while Multiple Linear Regression (MLR) and Support Vector Regression (SVR) were used to estimate pH and Soil Organic Matter (SOM) in paddy soils (Yang et al., 2019). Partial Least Squares Regression (PLSR) was used to predict moisture content (MC), total nitrogen (TN), and soil organic carbon (SOC) (Morellos et al., 2016). Soil moisture was estimated by combining the Auto-Regressive Error Function (AREF) with Gradient Boosting and k-Nearest Neighbors (k-NN) (Johann et al., 2016). Extreme Learning Machine (ELM) integrated with a Self-Adaptive Evolutionary agent (SaE) was used to assess soil temperature (Nahvi et al., 2016), while ELM was employed to forecast surface humidity (Acar et al., 2019). Lastly, Random Forests and SVM were used to estimate SOC and TN in Moroccan soils (Reda et al., 2019). Several studies also advanced ML and image recognition techniques for seed sorting and counting (Li et al., 2021; Nehoshtan et al., 2021; Laudari et al., 2022; Ekramirad et al., 2024). Deep Learning, especially Convolutional Neural Networks (CNNs), revolutionized plant disease and pest detection, allowing rapid and accurate diagnosis, which is crucial for minimizing crop loss (Mohanty et al., 2016; Ramcharan et al., 2017; Too et al., 2019; Argüeso et al., 2020). These methods optimize agricultural processes like planting, irrigation, and fertilization, enhancing productivity. Table 2 highlights key soil parameters detected using UAV technologies across various crops, showcasing advancements in PA research. AI models further refine resource use by recommending precise amounts of water, fertilizer, and pesticides, reducing waste and minimizing environmental impact. These models also simulate different farming scenarios, helping farmers assess and select the most effective strategies (Marvuglia et al., 2022).

Complex data is simplified into actionable insights for farmers, delivered to farmers via mobile apps and dashboards, providing real-time updates, visualizations, and alerts. For instance, farmers receive notifications about pest outbreaks with suggested treatments or alerts about sudden soil moisture drops with actionable irrigation advice. Farmers implement these recommendations and provide feedback, which helps refine the system's accuracy and relevance over time. This feedback loop, combined with adaptive learning, enhances the system's predictive capabilities for future growing seasons. The benefits include optimized operations through precise irrigation, fertilization, and pest control, increased yields from timely interventions, and sustainable practices by reducing chemical use and promoting efficient resource allocation. This practice empowers farmers to make informed decisions, boosting productivity, profitability, and sustainability in agriculture.