Exploratory Spatial Data Analysis (ESDA) soil analysis dashboard

Domain: Spatial Data Science · Data Engineering · Business Intelligence · GIS · Soil Analytics
Geographic Scope: Maghreb Region (Algeria, Tunisia, Morocco, Libya, and Mauritania)

Overview

Developed as a research initiative, the solution combines Data Warehousing, ETL pipelines, spatial analytics, and interactive geospatial visualization to transform heterogeneous soil datasets into structured, analysis-ready information. The primary objective was to provide a unified environment for storing, processing, analyzing, and visualizing soil properties, enabling researchers and decision-makers to investigate their spatial distributions, identify geographic patterns, and understand spatial relationships.

The project introduced a data warehouse architecture specifically designed for spatial soil data analysis in the Maghreb region, an area particularly vulnerable to climate change, soil degradation, and food insecurity.

An interactive web-based dashboard was developed to explore soil properties across five countries, with filtering capabilities by country and soil property, and integrated Exploratory Spatial Data Analysis (ESDA) techniques.

Key Components & Technical Implementation

1. Spatial Soil Data Collection & Integration

• Collection of soil information from the FAO/UNESCO World Soil Map through the FAO Soil Portal.
• Extraction and integration of soil data covering Algeria, Tunisia, Morocco, Libya, and Mauritania.
• Processing of a global soil dataset containing approximately 34,111 records and 1,700 soil profiles.
• Extraction and transformation of relevant soil attributes, geographic information, and soil mapping units.
• Construction of a regional dataset containing approximately 800 records.
• Integration of geographical and soil-related attributes, including soil texture, pH, organic carbon, nitrogen, cation exchange capacity, and other physical and chemical properties.

2. Spatial Data Warehouse Architecture & ETL Pipeline

Design and implementation of a four-layer spatial data warehouse architecture:

• Data Acquisition Layer: Collection of soil information from external sources, including FAO soil datasets.
• Staging Area: Intermediate storage and processing of raw data, including data cleaning, validation, and transformation.
• Storage Area: Design and implementation of a spatial data warehouse for structured soil data storage and multidimensional analysis.
• Data Mining & Visualization Layer: Integration of analytical and visualization tools to support spatial exploration and decision-making.

An ETL (Extract, Transform, Load) pipeline was developed to ensure data consistency, quality, and interoperability.

3. Multidimensional Data Modeling & Spatial Database Design

• Design of a star-schema data warehouse using a fact table and two dimensions.
• Implementation of a Location Dimension to represent geographical information.
• Development of a Soil Units Dimension to describe FAO soil mapping units and their characteristics.
• Integration of soil property measurements into the fact table.
• Implementation of the spatial database using MySQL, leveraging its spatial data types and geographic analysis capabilities.
• Organization of soil attributes to support multidimensional queries and spatial analysis.

4. Exploratory Spatial Data Analysis (ESDA)

Implementation of a comprehensive ESDA workflow to investigate the spatial distribution, heterogeneity, and autocorrelation of soil properties.

Spatial Distribution & Outlier Detection

• Statistical exploration using histograms and boxplots.
• Development of choropleth maps to visualize geographical distributions.
• Implementation of multiple map classification techniques, including:
  • Quantile classification.
  • Boxmaps.
  • Jenks-Caspall classification.
  • Fisher-Jenks classification.
• Identification and spatial localization of soil property outliers.

Global Spatial Autocorrelation

• Implementation of Moran’s I statistic to quantify global spatial autocorrelation.
• Assessment of spatial clustering patterns and relationships between neighboring observations.
• Statistical significance evaluation using permutation-based simulations.

Local Spatial Autocorrelation & Cluster Detection

• Development of Moran scatter plots to investigate local spatial relationships.
• Implementation of Local Indicators of Spatial Association (LISA).
• Generation of local spatial autocorrelation maps.
• Identification of High-High (HH), Low-Low (LL), High-Low (HL), Low-High (LH), and non-significant (NS) spatial clusters.
• Visualization of spatial heterogeneity and statistically significant local patterns.

5. Interactive Web-Based Soil Analysis Dashboard

Development of a web-based Human-Machine Interface (HMI) to provide interactive access to the spatial data warehouse and analytical results.

• Interactive exploration of soil data across the five Maghreb countries.
• Dynamic filtering by country and soil property.
• Visualization of soil property distributions through choropleth maps.
• Display of statistical distributions and spatial outliers.
• Integration of global and local spatial autocorrelation analysis.
• Visualization of Moran scatter plots and LISA cluster maps.
• Presentation of spatial analysis results to support scientific exploration and informed decision-making.

Key Findings & Research Contributions

The experiments focused particularly on topsoil organic carbon (OC), an important soil property associated with soil health and climate regulation.

Key findings included:

• Identification of higher organic carbon values in Morocco and Tunisia.
• Observation of similar organic carbon spatial distributions in Algeria and Libya.
• Detection of significant positive global spatial autocorrelation, with Moran’s I approximately equal to 0.44 using eight nearest neighbors.
• Identification of spatial clustering patterns across the Maghreb region.
• Detection of local High-High (HH) clusters, particularly in northern Algeria, Tunisia, and large parts of Morocco.
• Identification of Low-Low (LL) clusters predominantly in the southern Saharan areas.
• Detection of localized negative spatial autocorrelation and non-significant regions, demonstrating the importance of local analysis beyond global statistical indicators.

The results demonstrated how integrating spatial data warehousing with ESDA techniques can provide valuable insights into regional soil characteristics and spatial relationships.