Geospatial Workflow

A Python-based geospatial analysis workflow developed for spatial data processing, land parcel analysis, road-network analysis, and multispectral remote sensing analysis of Tarkwa.

Project Overview

This project demonstrates the application of Python-based geospatial and remote sensing techniques to analyse spatial datasets within Tarkwa.

The workflow is divided into two main parts:

• Part A: Vector-based geospatial analysis using land parcels, roads, buildings, and water bodies.
• Part B: Multispectral raster analysis and land-cover classification using spectral indices.

Objectives

The workflow was developed to:

• Inspect and process geospatial datasets using Python.
• Examine and transform coordinate reference systems.
• Analyse land parcels based on land use, area, value, and registration status.
• Perform spatial analysis involving roads, buildings, and land parcels.
• Generate road buffers and determine affected land parcels.
• Visualise land-use and land-value distributions.
• Calculate multispectral vegetation, water, built-up, soil, and burn-related indices.
• Produce true-colour and false-colour composites.
• Perform threshold-based land-cover classification.
• Evaluate classification performance using a confusion matrix and overall accuracy.
• Investigate potential mining/degraded land areas in Tarkwa.

Technologies and Libraries

The project uses:

• Python
• Jupyter Notebook
• GeoPandas
• Pandas
• NumPy
• Matplotlib
• Matplotlib-Scalebar
• Scikit-learn
• GeoJSON
• NumPy .npz raster data

Part A — Vector Geospatial Analysis

1. Land Parcel Analysis

The land parcel dataset is loaded and inspected to determine:

• Number of land parcels
• Attribute structure
• Geometry information
• Coordinate Reference System (CRS)
• Spatial extent

The parcel data is transformed to UTM Zone 30N (EPSG:32630) for projected spatial analysis.

2. Land Use Analysis

The workflow analyses land parcels based on:

• Land-use type
• Parcel area
• Land value
• Registration status
• Tenure type

The analysis includes filtering residential parcels greater than 500 m², calculating average land values, grouping parcels by land use, and identifying the most valuable parcels.

3. Road Buffer Analysis

Road data is transformed to UTM coordinates and a 50 m buffer is generated around the road network.

The buffer is dissolved into a single geometry and its total area is calculated in hectares.

The workflow also identifies land parcels intersecting the 50 m road corridor.

4. Building and Parcel Spatial Analysis

A spatial join is performed between land parcels and building features to determine:

• Parcels containing/intersecting buildings
• Parcels without associated buildings

5. Geospatial Visualisation

Maps are produced to visualise:

• Land-use distribution and road networks
• Continuous distribution of land parcel values
• Water bodies

The maps include cartographic elements such as:

• North arrow
• Scale bar
• Legends
• Map titles

Part B — Multispectral Remote Sensing Analysis

Multispectral data for Tarkwa is loaded from a NumPy .npz dataset.

The workflow uses:

• Blue
• Green
• Red
• Near Infrared (NIR)
• Short-Wave Infrared 1 (SWIR-1)
• Short-Wave Infrared 2 (SWIR-2)

Spectral Indices

The following spectral indices are calculated:

Index	Purpose
NDVI	Vegetation assessment
NDWI	Water-related feature assessment
NDBI	Built-up area assessment
SAVI	Vegetation analysis with soil adjustment
NBR	Burned/degraded area assessment

For each index, the workflow calculates minimum, maximum, and mean values.

Image Visualisation

The workflow produces:

• True Colour Composite (TCC)
• False Colour Composite (FCC)
• Spectral index visualisations

Histogram stretching using the 2nd and 98th percentiles is applied to improve image visualisation.

Land Cover Classification

A threshold-based classification is implemented using spectral indices.

The classification includes:

1. Dense Vegetation
2. Sparse Vegetation
3. Urban / Built-up
4. Bare Soil
5. Water

The number and percentage of pixels assigned to each class are calculated.

Accuracy Assessment

The classified land-cover image is compared with the available ground-truth land-cover mask.

The assessment includes:

• Overall classification accuracy
• Confusion matrix
• Visualisation of the confusion matrix

Mean NDVI values are also calculated for each land-cover class to assess whether the spectral behaviour is physically consistent with the assigned class.

Mining / Degraded Land Analysis

Potential mining or degraded areas within Tarkwa are identified using a combination of NDVI and NDBI thresholds.

The workflow calculates:

• Number of identified pixels
• Total area in square metres
• Total area in hectares

Project Structure

Geospatial-Workflow/
│
├── Geospatial workflow.ipynb
├── README.md
├── .gitignore
│
└── Data/
    ├── land_parcels.geojson
    ├── roads.geojson
    ├── buildings.geojson
    ├── water_bodies.geojson
    └── tarkwa_multispectral.npz## How to Use

1. Clone or download this repository.
2. Install the required Python libraries.
3. Ensure the required datasets are available in the appropriate directories.
4. Open `Geospatial workflow.ipynb` in Jupyter Notebook or JupyterLab.
5. Run the notebook cells sequentially.