📄 earth_science

Geospatial Methods for SDG-Aligned Land and Water Resource Management: A Bibliometric and Systematic Review of Research Trends (2021–2026)

This study combines bibliometric and systematic reviews of 2021–2026 research to map global trends in geospatial methods for land and water management, revealing a shift toward cloud-based platforms and machine learning while identifying critical gaps in real-time monitoring and policy integration to better support SDG 6 and SDG 15.

Original authors: Ajay Prakash M., Ragunath K. P., Pazhanivelan S., Muthumanickam D., Sivamurugan A. P., Vanitha G., Fawaz Parapurath, Kamalesh Kanna S., Sivakumar D., Kumar V.

Published 2026-08-06

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Ajay Prakash M., Ragunath K. P., Pazhanivelan S., Muthumanickam D., Sivamurugan A. P., Vanitha G., Fawaz Parapurath, Kamalesh Kanna S., Sivakumar D., Kumar V.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Earth as a giant, living puzzle where the land and the water are two pieces that constantly press against each other. If you push too hard on the land (like building a city or farming too much), the water gets squeezed out, dries up, or gets dirty. If the water moves wrong, the land can wash away. For a long time, scientists tried to solve this puzzle by sending people into the field with clipboards and buckets, but that's slow, expensive, and sometimes impossible when the weather is bad or the area is too huge. Enter the "Geospatial" detectives. These are scientists who use satellites orbiting high above like giant, super-powered eyes to take pictures of the Earth, and powerful computers to turn those pictures into maps. They use tools like Remote Sensing (taking photos from space), GIS (a digital map layer system that lets you stack information like a sandwich), and Machine Learning (computers that learn to spot patterns better than humans can). We care about this because the world is getting hotter, cities are growing, and water is becoming scarce. If we can't figure out where the water is hiding or how the land is changing, we can't feed people or keep them safe from floods.

Now, picture a massive library containing every research paper written between 2021 and 2026 about these space-satellite maps. The authors of this paper decided to be the ultimate librarians. They didn't just read a few books; they scanned 335 peer-reviewed articles to see what the whole world was doing, and then they dug deep into 59 specific studies to see if those studies were actually doing a good job. Think of it like reviewing a season of a TV show: they looked at the ratings (how many people watched) and then critiqued the plot (did the story make sense?).

Here is what they found, and it's a bit of a plot twist.

First, the library is getting crowded fast. The number of papers published jumped by 59.2% between 2021 and 2025, peaking at 1,735 articles in 2025. It's like a sudden explosion of interest, driven by the pandemic (which made it hard to go outside, so people used satellites instead) and the global push to meet the UN's "Sustainable Development Goals" (a checklist for saving the planet). While big players like China and the US wrote the most papers, the most "cited" (most talked about) work came from places like Kenya, Ecuador, and South Africa. It seems that when you are in a place where water is really scarce, your research gets a lot of attention because it solves a real, urgent problem.

But here is the big surprise, the "gotcha" moment of the story. For years, many scientists have been following a recipe: "To make a better map, just add more ingredients." They thought that if you stacked more layers of data (like adding more toppings to a pizza) or checked your map against more ground points (like tasting the pizza more times), your map would automatically be perfect.

The authors tested this idea with math, and it didn't work.

They found that adding more "thematic input layers" (the data toppings) barely improved the accuracy of the maps. In fact, the math showed that the number of layers explained only 9.4% of why a map was good or bad. It's like realizing that adding a tenth slice of pepperoni to your pizza doesn't make it taste significantly better; the quality of the dough and the sauce matters way more. Similarly, just having more ground-check points helped a little, but it wasn't the magic bullet everyone thought it was.

So, what does make a map good? The paper found that the method you use is the real secret sauce.

For a long time, the most popular method was called AHP (Analytical Hierarchy Process). Think of AHP as a "voting system" where experts decide how important each map layer is. It's easy to use and very common, especially in places like South Asia and Africa where computers might be slower. But the paper found that AHP is like a reliable, old-school calculator: it gets the job done, but it's not the fastest or smartest tool available.

When the authors compared AHP to Machine Learning (where the computer learns from data on its own), the computer won every time. In the studies that compared the two side-by-side, Machine Learning achieved an accuracy score (called ROC-AUC) of 0.860, while AHP only reached 0.748. That's a 15% improvement. It's the difference between a GPS that gets you to the right street and one that gets you to the exact front door.

However, there is a sad twist. Even though Machine Learning is clearly better, the places that need it the most (South Asia and Africa, where groundwater is running dry) are still mostly using the old AHP method. It's like driving a slow, reliable car when a fast electric one is sitting in the garage, but you're too scared to switch. The paper suggests that if these regions could switch to the newer, smarter methods, they could find water much more accurately.

The paper also points out some other missing pieces in the puzzle.

The "Black Box" Problem: Many studies use smart computers but don't explain why the computer made a decision. It's like a teacher giving you an "A" but not telling you what you did right. The authors say we need "Explainable AI" so policymakers can trust the results.
The "Static" Trap: Most maps are just a snapshot of one moment in time. But the world is always moving. The paper argues we need "Real-Time" monitoring, like a live video feed instead of a photograph, to catch floods or droughts as they happen.
The "Supply-Only" Blind Spot: Almost all the water studies only looked at where the water is (supply), but none of them looked at how much people are using it (demand). It's like measuring how much water is in a bucket but ignoring the hole in the bottom. Without knowing how much is being used, we can't really manage the resource.

In the end, this paper is a call to action. It tells us that the tools we have are getting better, but we are using the old, slower ones in the places that need speed the most. The future isn't just about taking more pictures from space; it's about using smarter computers to interpret those pictures, explaining our answers clearly, and connecting the dots between where the water is and how we use it. If we can do that, we might just solve the biggest puzzle of all: keeping our land and water healthy for everyone.

Technical Summary: Geospatial Methods for SDG-Aligned Land and Water Resource Management (2021–2026)

Problem Statement

The interconnection between land and water resources is fundamental to sustainable development, yet effective management is increasingly challenged by climate change, urbanization, and water scarcity. While geospatial technologies (Remote Sensing, GIS, GNSS) offer robust frameworks for observation and analysis, previous literature reviews have been limited by domain-specific scopes (focusing solely on groundwater or land cover) or single-method frameworks lacking cross-domain synthesis. Furthermore, there is a lack of quantitative evidence regarding how specific study design choices—such as the number of validation points or input layers—actually influence model accuracy. This study addresses the absence of a comprehensive bibliometric characterization of the 2021–2026 research landscape, the lack of quantitative analysis linking study design to model performance, and the absence of a geographic mapping of method–performance disparities.

Methodology

The study employs a dual-method approach combining bibliometric analysis with a PRISMA-based systematic review and meta-regression.

1. Bibliometric Analysis:

Data Source: The primary dataset was retrieved from the Dimensions API, selected for its broad multidisciplinary coverage, open-access metadata, and programmatic API capabilities.
Search Strategy: A Boolean search string ("Geospatial Methods AND Land AND Water Resource Management") was applied to titles and abstracts.
Filtering: From an initial 7,104 records, the dataset was filtered to include only peer-reviewed articles published in English between 2021 and 2026. This resulted in a final analytical corpus of 335 articles.
Validation: Cross-validation against Scopus and Web of Science confirmed high concordance (Spearman $\rho$ = 0.92 and 0.89, respectively) in country-level rankings.
Tools: Microsoft Excel for trend analysis and VOSviewer for keyword co-occurrence and co-authorship network mapping.

2. Systematic Review:

Search: Conducted across Scopus, PubMed, and Dimensions AI, yielding 583 initial records.
Screening: Following duplicate removal and title/abstract screening, 136 full-text articles were assessed.
Inclusion: 59 studies met the eligibility criteria, requiring empirical geospatial applications with quantifiable validation metrics.
Quality Assessment: A novel five-dimension quality appraisal tool (adapted from clinical frameworks like GRADE and ROBINS-I) was developed to assess spatial resolution bias, temporal consistency, model validation integrity, ground truth reliability, and selection bias.

3. Meta-Regression:

A subset of studies reporting both ROC-AUC scores and specific design parameters (validation point count and thematic input layer count) was selected for regression analysis to determine predictors of model accuracy.

Key Contributions

The paper claims three primary methodological and empirical contributions:

First Integrated Review: It is the first systematic review to combine bibliometric mapping with PRISMA-based meta-regression specifically for geospatial methods in land and water management during the 2021–2026 period.
Quantitative Design Framework: It introduces a meta-regression framework demonstrating that commonly reported study design parameters (layer count, validation point count) are weak predictors of model accuracy, challenging the assumption that "more data" equates to "better models."
Standardized Quality Appraisal: It proposes a five-dimension quality appraisal tool adapted for geospatial systematic reviews, providing the first standardized scoring rubric for this field.

Key Results

1. Publication Trends and Collaboration:

Publication output grew by 59.2% between 2021 and 2025, peaking at 1,735 articles in 2025.
China and the United States dominate collaborative output and link strength. However, countries with acute resource stress (Kenya, Ecuador, South Africa) demonstrate disproportionately high citation impact per document.
Groundwater potential mapping is the most studied application (39% of the corpus), followed by Land Use/Land Cover (LULC) change detection.

2. Methodological Performance (Meta-Regression):

Validation Points: There is a positive correlation between the number of field validation points and ROC-AUC ( $R^2 = 0.394$ ), but this explains less than 40% of the variance.
Input Layers: The number of thematic input layers has a negligible impact on model accuracy ( $R^2 = 0.094$ ). Studies with fewer layers often outperformed those with more, indicating that layer quality and relevance are more critical than quantity.
Method Family: Machine Learning (ML) and ensemble methods consistently outperformed Analytical Hierarchy Process (AHP) frameworks.
- AHP-only: Mean ROC-AUC = 0.748.
- AHP-hybrid: Mean ROC-AUC = 0.815.
- ML/Ensemble: Mean ROC-AUC = 0.860.
Geographic Disparity: Despite ML's superior performance, AHP remains the dominant method in South Asia and Africa (regions with the highest groundwater stress), while ML is concentrated in North America and parts of South Asia with better computational infrastructure.

3. Thematic Findings:

Land Management: 92% of comparative studies confirmed that urban and agricultural expansion measurably reduces groundwater levels and water body extent. However, none of the 15 land management studies explicitly linked findings to SDG 15 quantitative targets.
Water Management: Groundwater potential zone (GPZ) mapping accounted for 56% of water studies. A critical gap identified is the supply-side bias: 100% of water management studies focused on supply mapping without integrating demand-side indicators (pricing, consumption efficiency), rendering them insufficient for SDG 6.4 monitoring.
Technology Gap: Despite bibliometric prominence, advanced tools like SAR data (8% usage), Google Earth Engine (7% usage), and hyperspectral sensors (0% usage) are underutilized in operational research compared to optical imagery and desktop GIS.

4. Quality and Reproducibility:

Ground Truth Reliability: The most frequent methodological weakness (flagged in 57% of assessed studies) was inadequate ground truth, often relying on Google Earth interpretation or fewer than 20 field points.
Temporal Consistency: Many studies used static, single-epoch datasets in contexts requiring multi-temporal analysis, limiting the ability to detect trends.

Significance and Claims

The paper positions itself as a critical intervention in the field of geospatial resource management by moving beyond descriptive mapping to evidence-based methodological critique.

Policy Relevance: The study argues that current geospatial research is often disconnected from policy implementation. It highlights that without integrating demand-side data and governance indicators, geospatial outputs cannot effectively support SDG 6 (Clean Water) and SDG 15 (Life on Land) monitoring.
Methodological Shift: It claims that the field is shifting from data acquisition (Remote Sensing) to data interpretation and decision integration. However, it identifies a "translation gap" where superior methods (ML) are not being adopted in the regions that need them most due to data and computational constraints.
Future Priorities: The authors identify three critical priorities for advancing the field:
1. Real-time monitoring: Moving from static snapshots to continuous, IoT-integrated frameworks.
2. Explainable AI (XAI): Requiring SHAP or LIME outputs to ensure model interpretability for policymakers.
3. Climate-resilient frameworks: Integrating dynamic climate projections (CMIP6) into geospatial models to ensure long-term relevance.

The study concludes that while geospatial technologies are indispensable, the field must evolve from descriptive, static, and supply-side focused studies to predictive, policy-integrated, and reproducible decision tools to meet the 2030 SDG agenda.