Collins Edem Hlordzie | Geospatial Data Analyst & Earth Observation Scientist

About

Bridging environmental data and decision-making

I'm a geo-information and Earth Observation scientist who holds an Erasmus+ Joint Master's in Geo-information Science and Earth Observation for Environmental Modelling and Management (GEM) from Lund University, Sweden, and the University of Twente (ITC), the Netherlands. My work sits at the intersection of GIS, satellite and UAV remote sensing, and machine learning.

My experience spans high-resolution UAV crop mapping with deep learning, two-decade satellite time-series analysis of vegetation phenology, and — most recently — contributing to AI-driven agricultural advisory systems and Earth-Observation-based carbon MRV methodologies as a Geospatial Intern at RAMANI B.V.

I'm particularly interested in building spatial frameworks that make environmental data usable for the people who have to act on it: smallholder farmers, planners, and researchers working on climate-resilient agriculture.

Data Storytelling

Turning raster/vector outputs into narratives non-specialists can act on

Remote Sensing

UAV, Sentinel, MODIS & Landsat imagery analysis end-to-end

GeoAI & Spatial Modelling

Deep learning segmentation, classification, and trend/uncertainty analysis

UAV Operations

EASA-certified remote pilot (A1/A3) with field survey design experience

Tools & Technologies

Python

ArcGIS Pro / Desktop

QGIS

Google Earth Engine

ESA SNAP

PIX4D / UAV Piloting

IDRISI TerrSet

MATLAB

Deep Learning / ML

Excel / Office

Spatial Databases

Selected Work

Projects & case studies

Seven projects spanning UAV deep learning, satellite time-series analysis, spatial modelling, and applied GeoAI — each framed as problem → approach → insight.

MSc Thesis Featured

UAV Crop Mapping & Vegetation Estimation

Mozambique

Problem

Smallholder farms in Mozambique feature small, intercropped fields that coarse satellite data can't resolve — a blind spot for crop-area and food-security monitoring.

Approach

Pre-processed high-resolution UAV imagery, applied threshold-based cropland detection, then classified crop types and intercrop compositions with a Segformer deep-learning model; estimated fractional vegetation cover (Fcover) and validated against enumerator ground-truth.

UAV ImageryPythonSegformer / Deep LearningQGIS

Field-level crop-type & Fcover maps validated against ground survey data

Thesis graded 8.5/10 — supervised by Dr. F. Ellsäßer & Dr. F. Osei

Read Full Thesis

Internship Featured

EO-Driven Carbon MRV & Agri-Advisory AI

RAMANI B.V., NL

Problem

Nature-based carbon-removal projects need EO-based methodologies that stay consistent and interoperable across protocols, while agri-advisory systems need EO indicators fused with agronomic decision rules.

Approach

Validated and cross-compared EO-based implementations of Verra's VM0047 methodology for nature-based carbon removal; built semantic-uplift mappings for vocabulary and data interoperability across carbon protocols (Gold Standard, Open Forest); contributed EO indicators to an AI-driven agricultural advisory pipeline.

Earth ObservationGeoAISemantic Data ModellingCarbon MRV

Strengthened cross-protocol interoperability for carbon MRV data

Fed EO indicators directly into a production AI advisory system

Launch Live App

Remote Sensing

Land Cover Classification, Lund Municipality

Sweden

Problem

Municipal planners need reliable land-cover maps, but classifier choice and class granularity strongly affect what "reliable" means in practice.

Approach

Benchmarked Maximum Likelihood, Random Forest, and SVM classifiers on Sentinel-2 imagery against field-collected training data (85/15 split), then ran full accuracy assessment — error matrix, kappa, user's/producer's accuracy — across 8-class and 6-class schemes.

Sentinel-2ArcGIS / TerrSetR (RF, SVM)Field GPS Survey

Reducing 8→6 classes lifted MLC accuracy from 25% to 79%

SVM and RF consistently outperformed MLC on spectrally overlapping classes

Time Series

Vegetation Phenology, 2001–2021

N. Scandinavia

Problem

Understanding how climate change is reshaping growing seasons across the Scandinavian Peninsula requires two decades of consistent, noise-resistant vegetation monitoring.

Approach

Derived the Plant Phenology Index from 21 years of MODIS reflectance data, smoothed the time series in TIMESAT, and applied Theil-Sen and Mann-Kendall trend tests to extract SOS/EOS/LOS and productivity metrics.

MODISTIMESAT (MATLAB)Trend AnalysisArcGIS Pro

Growing seasons lengthening across central/northern Scandinavia

High-altitude/latitude vegetation trending opposite — flags climate sensitivity hotspots

Carbon Mapping

Carbon Storage Mapping, UT Campus

Netherlands

Problem

Quantifying how much carbon a campus forest stores requires linking sparse field measurements to full spatial coverage.

Approach

Ran a stratified field inventory (DBH measurements across coniferous/broadleaf/mixed plots), calculated above-ground biomass with allometric equations, then regressed plot-level carbon against Sentinel-2 vegetation indices to extrapolate across the full forest extent.

Sentinel-2ESA SNAPRField DBH Survey

SAVI gave the strongest fit (R² = 0.7) among four vegetation indices tested

Estimated 590,737 tons of carbon stored across 43.67 ha of forest

Spatial Modelling

Nutrient Leaching & Eutrophication

Rönne å, Sweden

Problem

The Rönne å basin is a major nutrient source feeding coastal eutrophication in Skälderviken Bay — authorities need to know which land uses drive the pollution, and at what cost to fix.

Approach

Modelled phosphorus and nitrogen loads by land-use type in ArcGIS Model Builder, then designed protective buffer zones along cultivated-land waterways and costed farmer compensation plus construction.

ArcGIS Model BuilderExcelLand-use Data

Cultivated land identified as the dominant phosphorus & nitrogen source

Proposed buffer zones (46,345 ha) could cut P leaching by 11,586 kg/year

Fuzzy Modelling

Land Suitability Modelling

Hambantota, Sri Lanka

Problem

Binary suitable/unsuitable land classifications hide the nuance land-use planners actually need when soil and terrain factors trade off against each other.

Approach

Built a fuzzy-set land suitability model using two aggregation methods — Weighted Linear Combination and Fuzzy AND — and compared how each represents suitability thresholds and trade-offs.

Fuzzy Set TheoryArcGISMulti-criteria Analysis

WLC produced broader, trade-off-tolerant suitability zones (max score 0.95)

Fuzzy AND isolated smaller, uniformly high-suitability areas for stricter decisions

Global Footprint

Where the fieldwork happened

From UAV flights over Mozambican smallholder fields to phenology modelling across the Scandinavian Peninsula — click a marker to see the project.

Career Path

Experience

Feb 2026 – Jun 2026

Geospatial Intern

RAMANI B.V., The Netherlands

Contributed to AI-driven agricultural advisory systems integrating Earth Observation indicators with agronomic decision rules
Validated and compared EO-based implementations of Verra's VM0047 methodology for nature-based carbon removal
Developed semantic-uplift mappings for vocabulary and data interoperability across carbon-sequestration protocols (Gold Standard, Open Forest)

Dec 2025 – Jul 2026

MSc Thesis Researcher

University of Twente / Lund University

UAV-based crop mapping and fractional vegetation estimation in smallholder-dominated landscapes of Mozambique, using a Segformer deep-learning model

Sep 2020 – Aug 2021

Undergraduate Teaching & Research Assistant

University of Cape Coast, Ghana

Tutored over 50 students in Human Geography and Resource Perception
Mentored 10 students with low academic standing, leading to measurable performance gains
Assisted with class setup and assignment grading

Jun 2019 – Aug 2019

Intern

Environmental Protection Agency, Ho, Ghana

Assisted in environmental control and monitoring
Wrote Environmental Impact Assessment reports
Developed materials for a climate-change awareness & education project

Education

MSc, Geo-information Science & Earth Observation (GEM)

Lund University, Sweden & University of Twente (ITC), Netherlands · 2024 – 2026 · Erasmus+ Scholarship

Thesis: UAV-based Crop Mapping and Fractional Vegetation Estimation in Smallholder-Dominated Landscapes of Mozambique — Grade 8.5/10

BA, Geography

University of Cape Coast, Ghana · 2016 – 2020 · CGPA 3.5/4.0 · GNPC Scholarship

Thesis: Climate Change and Subsistence Farming in Northern Ghana

Technical Proficiency

Remote Sensing & GIS (ArcGIS, QGIS, GEE)95%

UAV Data Processing & Piloting90%

Python / R / MATLAB85%

Deep Learning / GeoAI80%

Spatial Statistics & Trend Analysis85%

Analytical & Soft Skills

Data Storytelling Stakeholder Engagement Scientific Writing Field Survey Design Cross-cultural Collaboration Science Communication

Certifications

EASA UAS Remote Pilot (A1/A3), 2025 UAV Segmentation Techniques, Geoversity 2026 UAVs in Precision Agriculture, ITC 2025 Copernicus & Sentinel Data, ITC 2025 Python, U. Michigan 2025 Spatial Data Science, ESRI 2020

Honors & Awards

Erasmus+ Scholarship, 2024–2026 GNPC Scholarship, 2016–2020

Collins Edem
Hlordzie

Bridging environmental data and decision-making

Data Storytelling

Remote Sensing

GeoAI & Spatial Modelling

UAV Operations

Tools & Technologies

Featured Insight

Projects & case studies

Problem

Approach

Problem

Approach

Problem

Approach

Problem

Approach

Problem

Approach

Problem

Approach

Problem

Approach

Where the fieldwork happened

Experience

Education

Technical Proficiency

Analytical & Soft Skills

Certifications

Honors & Awards

Curriculum Vitae

Get the complete picture

Get in touch

Email

LinkedIn

Full e-Portfolio

GitHub