Case Study — 2020–2024

Greece
Wildfire Analysis

A 5-year data analysis of wildfire incidents across Greece, combining Greek Fire Service records, hourly meteorological data, and geographic coordinates to uncover patterns, risk factors, and resource effectiveness.

3.87M
Acres Burned
27K
Incidents Recorded
5 Yrs
Data Coverage
3
Data Sources Merged

What & Why

Greece faces a growing wildfire crisis. With climate change intensifying drought conditions and heat waves, understanding the patterns behind wildfire ignition and spread is critical for resource allocation and prevention policy.

This project integrates three independent datasets — fire service incident logs, weather station readings, and geographic coordinates — into a unified Power BI model, enabling analysis that no single source could support alone.

The goal was to move beyond simple incident counts and answer harder questions: Do weather conditions at ignition time predict final burned area? Which regions are chronically under-resourced? When is the risk truly "extreme"?

ΠΥ
Greek Fire Service (ΠΥ)
999+ incidents · 43 columns · 2020–2024
Personnel, vehicles, aircraft, timestamps, coordinates
🌡
Meteorological Stations
Hourly readings · Per prefecture · 5 years
Temp, humidity, wind speed & gusts, precipitation, pressure
📍
Geographic Coordinates
Greek prefectures · Lat/Lon centroids
Used for map visuals and spatial join
Methodology

How It Was Built

01
Data Modeling

Built a star schema in Power BI with Bridge tables for Calendar and Prefecture, enabling many-to-many relationship resolution between the fire and weather datasets.

Bridge_CalendarBridge_ΝομοίStar Schema
02
Weather Join at Ignition

Used DAX LOOKUPVALUE to match each fire incident to the precise hourly weather reading at ignition time and prefecture — not daily averages — capturing the actual conditions at the moment of ignition.

LOOKUPVALUEDateTime JoinHour Precision
03
Fire Weather Index

Developed a custom simplified FWI scoring model using temperature, relative humidity, wind speed, and precipitation — each scored 0–5 and combined into a 0–17 risk index with four danger categories.

Custom DAXSWITCH LogicRisk Scoring
Key Findings

What The Data Reveals

1.41%
Extreme Conditions Drive Catastrophic Outcomes

Just 1.41% of fires ignite under "Extreme" FWI conditions — yet they produce an average burned area of ~450 acres per incident, compared to ~5 acres under Low conditions. The vast majority of incidents (44.7%) occur under Moderate conditions, yet cause a fraction of the total damage. This confirms that early intervention in high-FWI scenarios is the highest-leverage policy intervention available.

2023
Worst Year on Record

2023 saw 1.74M acres burned — a +520% YoY surge — following a relative calm in 2022 (0.24M acres). The most destructive year in the 5-year dataset, with total completion hours peaking at 2M.

July
Peak Risk Month

July records the highest incident count (6K), followed by August (5.2K). The FWI index peaks in these months, and the line chart confirms that average burned area closely follows the FWI curve — validating the index as a reliable seasonal predictor.

+525% / +520%
Biennial Cycle 2021/2023

A clear 2-year catastrophic cycle emerged: 2021 (+525% YoY) and 2023 (+520% YoY) were both devastating years, while 2022 and 2024 saw sharp contractions (−72% each). This pattern suggests a compounding fuel load dynamic — vegetation recovers in the calm year, then burns catastrophically the next.

Evros & Euboea
Structural Gap

These two prefectures are consistent outliers across all metrics: highest burned area, longest completion times, and highest aerial asset deployment — yet outcomes remain disproportionately severe. High resources, poor outcomes: a structural resourcing and response deficit that single-year interventions cannot resolve.

770 acres
Burned Area per Aircraft Deployed

Each aerial asset is associated with 770 acres of burned area — the highest of all resource types (38.8 acres/vehicle, 15.3 acres/person). This suggests aircraft are deployed reactively in the largest fires, raising questions about pre-emptive vs. reactive deployment strategy.

Technical Stack

Tools & Techniques

📊
Power BI Desktop
Visualization & Reporting
⚙️
DAX
Measures & Calculated Columns
🔄
Power Query (M)
Data Transformation
🗺️
Geospatial Join
Lat/Lon Coordinate Mapping
DAX Showcase

Notable Measures

FWI_ΈναρξηςCalculated Column
-- Custom Fire Weather Index (0–17)
VAR _temp = VALUE('Φύλλο1'[Temp_Έναρξης])
VAR _rhum = VALUE('Φύλλο1'[Rhum_Έναρξης])
VAR _wspd = VALUE('Φύλλο1'[Wspd_Έναρξης])
VAR _temp_score =
  SWITCH(TRUE(),
    _temp >= 40, 5, _temp >= 35, 4,
    _temp >= 30, 3, 0)
RETURN _temp_score + _rhum_score + _wspd_score
Hydration YoY%Measure
-- Year-over-year humidity change
VAR __PREV =
  CALCULATE(
    AVERAGE('Καιρός'[rhum]),
    DATEADD(
      'Bridge_Calendar'[Date],
      -1, YEAR)
  )
RETURN
  DIVIDE(AVERAGE('Καιρός'[rhum]) - __PREV, __PREV)
Get In Touch

Interested in this work?

Available for data analysis roles, freelance projects, and collaboration opportunities.

↗ View Live Report ↗ GitHub Repository