Details

Prednostna območja za umeščanje modro-zelene infrastrukture za blaženje urbanega pregrevanja : magistrsko delo
ID Potočnik Buhvald, Ana (Author), ID Zavodnik Lamovšek, Alma (Mentor) More about this mentor... This link opens in a new window, ID Oštir, Krištof (Comentor)

.pdfPDF - Presentation file, Download (43,22 MB)
MD5: 48B560E2AA7AB76DD53D9EB9B8EE5811

Abstract
Urbano pregrevanje je prostorsko heterogen pojav, zato najtoplejše površine niso nujno prednostna območja za umeščanje modro-zelene infrastrukture (MZI). Namen naloge je bil na primeru Mestne občine Ljubljana (MOL) razviti pregleden in ponovljiv postopek za prepoznavanje območij, kjer se ponavljajoča površinska toplotna obremenitev prekriva z veliko socialno-prostorsko potrebo in izrazitim biofizikalnim primanjkljajem MZI. Analiza je temeljila na 174 poletnih satelitskih posnetkih Landsat in ECOSTRESS iz obdobja 2018–2025 ter na mreži celic velikosti 100 × 100 m. Modularni pristop je združil veččasovno analizo temperature zemeljskega površja (angl. Land Surface Temperature, LST), modeliranje z generaliziranim aditivnim modelom (GAM) in algoritmom LightGBM ter vrednotenje socialno-prostorske potrebe in biofizikalnega primanjkljaja MZI. Napovedna uspešnost je bila ocenjena s prostorskim blokovnim navzkrižnim preverjanjem. Ožje analitično območje je bilo v večini opazovanj toplejše od širšega območja MOL, predvsem podnevi; ponoči so bile temperaturne razlike manjše. Med pomembnejšimi napovedovalci LST sta izstopala delež neprepustnih površin in drevesni pokrov. LightGBM je na širšem območju v vseh petih senzorskih in časovnih sklopih dosegel boljšo napovedno uspešnost kot GAM, vendar je v rezidualih obeh modelov ostala prostorska avtokorelacija. Velika socialno-prostorska potreba in velik biofizikalni primanjkljaj sta se prekrivala na 40,78 % ožjega območja, strožji izbor pa je določil 179 prednostnih celic (1,59 % območja). Rezultati potrjujejo smiselnost usmerjanja nadaljnje prostorske presoje na območja, kjer se sočasno pojavljajo toplotna obremenitev, socialno-prostorska potreba in pomanjkanje hladilnih elementov.

Language:Slovenian
Keywords:magistrska dela, urbano pregrevanje, temperatura zemeljskega površja, modro-zelena infrastruktura, daljinsko zaznavanje, strojno učenje, večkriterijsko vrednotenje, Ljubljana
Work type:Master's thesis/paper
Typology:2.09 - Master's Thesis
Organization:FGG - Faculty of Civil and Geodetic Engineering
Place of publishing:Ljubljana
Publisher:[A. Potočnik Buhvald]
Year:2026
Number of pages:1 spletni vir (1 datoteka PDF (XI, 72 str.))
PID:20.500.12556/RUL-190977 This link opens in a new window
UDC:711.4:502.12(497.451.1)(043.2)
COBISS.SI-ID:304574979 This link opens in a new window
Publication date in RUL:09.11.2026
Views:123
Downloads:48
Metadata:XML DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:English
Title:Priority areas for the implementation of blue-green infrastructure to mitigate urban overheating
Abstract:
Urban overheating is spatially heterogeneous, meaning that the warmest surfaces are not necessarily the highest-priority areas for siting blue-green infrastructure (BGI). The aim of this study was to develop a transparent and reproducible procedure for identifying areas where recurring surface thermal load coincides with high socio-spatial need and a pronounced biophysical deficit of BGI, using the City Municipality of Ljubljana (MOL) as a case study. The analysis was based on 174 summer satellite scenes acquired by Landsat and ECOSTRESS between 2018 and 2025, and on a grid of 100 × 100 m cells. The modular approach combined multi-temporal analysis of land surface temperature (LST), modelling with a generalised additive model (GAM) and the LightGBM algorithm, and assessments of socio-spatial need and the biophysical deficit of BGI. Predictive performance was assessed using spatial block cross-validation. The narrower analytical area was warmer than the wider MOL area in most observations, particularly during the day, while temperature differences were smaller at night. The proportions of impervious surfaces and tree cover emerged as important predictors of LST. LightGBM achieved better predictive performance than GAM across the wider analytical area in all five sensor–time combinations, although spatial autocorrelation remained in the residuals of both models. High socio-spatial need and high biophysical deficit overlapped across 40.78 % of the narrower analytical area, while a stricter selection identified 179 priority cells (1.59 % of the area). These findings support directing further spatial assessment towards areas where surface thermal load, socio-spatial need and a lack of cooling elements coincide.

Keywords:master thesis, urban overheating, land surface temperature, blue-green infrastructure, remote sensing, machine learning, multi-criteria assessment, Ljubljana

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back