Research
Studies spanning spatial econometrics, composite territorial indicators and remote-sensing segmentation — with the methods and the limits stated in the same breath.
turkiye-housing-prices-pandemic
A reproducible, region-level analysis separating real inflation-adjusted house price growth from inflation itself, across 150 months of Türkiye's housing market.
Months, November 2013 to April 2026.
Real growth 2014–2019, against +84% nominal.
Real growth 2020–2025, against +1521% nominal.
Ankara's real growth, against İstanbul's 91%.
Method: real index construction from the TCMB residential property price index deflated by headline CPI, with CAGR, Fisher identity validation and alternative deflator comparison. Spatial structure tested with global Moran's I and LISA. Data from TCMB EVDS and TÜİK; province boundaries derived from OpenStreetMap.
Maps are simplified rather than cartometric. LISA has limited power at 19 spatial units. The analysis is descriptive, not causal, and no multiple-comparison correction is applied.
kutri-resilience-index
The Kaş Urban-Territorial Resilience Index: a reproducible five-pillar composite indicator prototype for the Kaş/Bayındır planning area in Antalya.
Distributed 7 / 7 / 8 / 10 / 8 across the five pillars.
Hazard, socio-demographic, economic, infrastructure, environmental-cultural.
Directional logic per indicator, positive or negative.
Arithmetic mean within pillars, weighted geometric mean across them.
Pillars: Natural Hazard & Physical Vulnerability; Socio-Demographic Adaptive Capacity; Economic Resilience; Infrastructure & Service Continuity; Environmental & Cultural Capital. Weights come from an Analytic Hierarchy Process eigenvector, with nominal policy weights tested as an alternative. MIT licence covers the source; raw data is not redistributed.
Explicitly not a universal or fully validated resilience model. It is a case-specific evidence base for planning decisions in one district.
agri-unet
Downstream agricultural pattern identification from satellite imagery, kept separate from the reusable data-preparation layer so that experiments stay auditable and the dataset work can be cited on its own.
Segmentation work sits downstream of sentinel-crop-pipeline, which supplies the spatially blocked patches and aligned label masks. The repository is not yet populated, so there are no published metrics or trained weights to report here.