Details

Prediction of Design Discharge Values in Ungauged Catchments in Slovenia : master thesis
ID Al Amin, Rafsan (Author), ID Bezak, Nejc (Mentor) More about this mentor... This link opens in a new window, ID Šraj, Mojca (Comentor)

.pdfPDF - Presentation file, Download (25,11 MB)
MD5: F9A04E6A9ECCFD34A132EFAD13307A36

Abstract
Reliable design discharge estimates are essential for flood risk management and hydraulic design, yet their estimation remains challenging in ungauged catchments where long-term discharge observations are unavailable. This study compared regression-based, artificial intelligence, and geostatistical regionalization methods for predicting design discharge in Slovenia using morphometric, terrain, land-cover, climatic, and rainfall-related catchment characteristics. Different target formulations were evaluated using leave-one-out cross validation, and the better-performing approaches were subsequently applied to an additional set of catchments treated as ungauged. Catchment size emerged as the dominant control variable on design discharge, with forest area repeatedly identified as an informative predictor because of its close relationship with catchment scale. Regression methods provided useful and interpretable predictions, while artificial intelligence and geostatistical methods generally offered greater flexibility in representing complex relationships. TabPFN showed particularly consistent performance across different target formulations and predictor combinations, while Top-Kriging demonstrated that strong predictions could also be obtained using relatively small external drift sets. Increasing the number of predictors did not consistently improve performance, showing the importance of parsimonious model development. Application to the ungauged catchments showed that no single tested method can be considered universally preferable. Model suitability depends on the characteristics of the target catchment, predictor availability, target formulation, model complexity, uncertainty, and the degree of extrapolation beyond the training data. The results therefore support a method selection approach in which several regionalization techniques are evaluated and the most appropriate model is chosen according to the specific ungauged catchment and available information.

Language:English
Keywords:master thesis, flood risk management, design discharge prediction, multi-variate regression, artificial intelligence, geo-statistics, topological kriging, tabPFN, ungauged catchments
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:[R. Al Amin ]
Year:2026
Number of pages:1 spletni vir (1 datoteka PDF ([XXII], 134 str., [126] str. pril.))
PID:20.500.12556/RUL-186404 This link opens in a new window
UDC:556.51(497.4)(043.2)
COBISS.SI-ID:290062595 This link opens in a new window
Publication date in RUL:01.09.2026
Views:296
Downloads:90
Metadata:XML DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:Slovenian
Title:Napoved projektnih pretokov v nemerjenih porečjih v Sloveniji : magistrsko delo
Abstract:
Zanesljive ocene projektnih pretokov so bistvenega pomena za obvladovanje poplavne ogroženosti in hidrotehnično načrtovanje, vendar je njihovo določanje v nemerjenih porečjih zahtevno zaradi pomanjkanja merjenih podatkov o pretokih. V tej nalogi so bile primerjane regresijske metode, metode umetne inteligence in geostatistične metode za napovedovanje projektnih pretokov v Sloveniji z uporabo morfometričnih, reliefnih, pokrovnostnih, podnebnih in padavinskih značilnosti porečij. Različne formulacije ciljnih spremenljivk so bile ovrednotene z metodo navzkrižnega preverjanja leave-one-out, uspešnejši pristopi pa so bili nato uporabljeni na dodatnem naboru porečij. Velikost porečja se je izkazala kot najpomembnejši dejavnik pri napovedovanju projektnih pretokov, pri čemer je površina gozda pogosto nastopala kot pomemben napovedni dejavnik zaradi tesne povezanosti z velikostjo porečja. Regresijske metode so omogočile uporabne in smiselne napovedi, medtem ko so metode umetne inteligence in geostatistični pristopi na splošno omogočili večjo prilagodljivost pri opisovanju kompleksnih povezav. TabPFN je pokazal posebno uspešnost pri različnih ciljnih formulacijah in kombinacijah napovednih spremenljivk, Top-Kriging pa je pokazal, da je mogoče dobre rezultate doseči tudi z razmeroma majhnim številom zunanjih spremenljivk. Povečevanje števila napovednih spremenljivk ni vedno izboljšalo uspešnosti, kar poudarja pomen razvoja modelov s poudarkom na smiselnem izboru vhodnih spremenljivk. Uporaba na nemerjenih porečjih je pokazala, da ni mogoče ene same metode obravnavati kot splošno najboljše. Primernost modela je odvisna od značilnosti ciljnega porečja, razpoložljivosti vhodnih podatkov, formulacije ciljne spremenljivke, kompleksnosti modela, negotovosti ter stopnje ekstrapolacije zunaj območja podatkov za učenje. Rezultati zato podpirajo pristop, pri katerem se primerja več metod in se najprimernejši model izbere glede na posamezno nemerjeno porečje in razpoložljive informacije.

Keywords:magistrska dela, upravljanje tveganja poplav, napoved projektnih pretokov, multivariatna regresija, umetna inteligenca, geostatistika, kriging, TabPFN, nemerjena porečja

Similar documents

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

Back