Windhoek, Namibia

Honest estimates from sparse data.

All models are wrong, but some are useful.

— George Box

Missing decades. Failing sensors. Sparse records. I build data pipelines and statistical models that remain correct anyway - auditable, versioned and honest enough to say "we do not know" when that is the answer.

Wilka Igulu — quantitative analyst, Windhoek Namibia

I am a quantitative analyst based in Windhoek, Namibia.

My Honours research — published in Wind (MDPI) — found that spurious sensor data had inflated Namibia's extreme wind estimates by 35%. That project taught me my real interest: not climate science as a cause, but the craft of building methods that remain correct when data is sparse, sensors fail, and the client needs a number by Friday.

80%
Honours thesis score. EVT applied to Namibian wind data.
64%
Of Namibia lies beyond >100 km of the nearest long-record station.
3 / 10
Namibian stations in the global archive with usable records to the present.
825k
km² covered by ~20 long-record stations. The gap this work addresses.
Data infrastructure · NOAA ISD · GitHub Actions Live — updates monthly

Namibia Station Monitor

A living data-quality audit of every Namibian weather station in the NOAA global archive. Recomputes a Data Quality Score (DQS v1.0) for all 56 stations on the 1st of each month via automated pipeline. Confirms Lüderitz present in ISD metadata since 1949, returns HTTP 404 in GHCND. Citable dataset: Igulu (2026), DOI 10.5281/zenodo.21229782.

Python NOAA ISD GitHub Actions DQS v1.0 Zenodo Open data
Data investigation · June 2026 Complete

The Namibian station audit: what the global climate archive actually contains

I pulled and quality-scored every Namibian record in NOAA's global daily archive. Three of ten stations reach the present day. Ondangwa is missing the 1990s. Lüderitz is not in the archive at all. The audit behind the shadow map and the article below.

NOAA GHCND Data quality ArcGIS Namibia
Peer-reviewed research · Wind (MDPI), August 2026 Published

Modelling Wind Speed Extremes Using Extreme Value Theory: A Case Study for Namibia

Identified that quality control - not model choice - was the dominant source of error in Namibia's extreme wind estimates. Removed 70 spurious AWS records (0.24% of observations) that were inflating 100-year return levels by 35-45%. Produced the first station-level GPD return-level estimates with profile-likelihood uncertainty bounds for six Namibian stations.

doi.org/10.3390/wind6030040 ↗

EVT / GPD R, ArcGIS Pro Wind energy Namibia
Research project · In progress

Namibian Climate Risk Sandbox

An open, reproducible pipeline from raw NOAA station records and CMIP6 projections to boardroom-ready physical risk assessments for a real Windhoek asset, the Goreangab Water Reclamation Plant. Station audit and CMIP6 bias-correction phases complete; extreme value and financial layers in build on a published schedule.

Python NOAA GHCND CMIP6 EVT PyMC Open source
Research program · v0.2 Active development

SACRF - Sparse-Adaptive Climate Risk Framework

Built on peer-reviewed extreme value research (Wind/MDPI, 2026). A Bayesian engine for honest climate-risk estimation in data-scarce African markets, where prior precision adapts to station-level data quality so that short records and long records are treated with appropriate uncertainty. Financial translation layer and pre-registered calibration experiment in design.

Bayesian EVT GPD / POT Uncertainty quantification Climate finance
Map of Namibia showing dark shadow zones more than 100 km from the nearest long-record weather station, with 100 km coverage circles around each station and DFI infrastructure sites marked.
Namibia Weather Station Shadow Map: zones more than 100 km from the nearest WMO long-record station (dark wash) cover roughly 64% of the country. Coverage circles show each station's 100 km observation footprint; DFI infrastructure sites marked by category. Built in ArcGIS. Sources: WMO OSCAR/Surface, Natural Earth, NASA SRTM, Strohbach (2014).  Read the article →
Live dashboard · Updated 1st of each month Live

Namibia Station Monitor - live DQS feed

56 Namibian weather stations scored monthly by Data Quality Score. Filter by usability tier or GHCND status. Lüderitz confirmed absent from the global daily archive despite ISD registration since 1949.

DQS v1.0 56 stations Live data

Production

KoboToolbox / XLSForm
Power BI / DAX
ETL pipeline design
Python (pandas, numpy, APIs)
SQL

Research

Applied Statistics
R (statistical computing)
Extreme Value Theory
Spatial analysis (ArcGIS / QGIS)
Bayesian methods / PyMC

Building

CMIP6 + QDM bias correction
Climate risk disclosure (IFRS S2 / TCFD)
DCF & scenario-adjusted NPV
Project finance fundamentals

Currently reading

Knaflic (2015) — Storytelling with Data
June 2026

Why Namibia's weather stations are a development finance problem

Twenty long-record stations, 825,000 km², a shadow map covering 64% of the country. What that means for infrastructure lending in southern Africa.

I take on selective consulting work in data pipeline architecture, statistical sampling, and risk quantification. For project inquiries or professional opportunities, reach out directly.