Spatial cross-correlation among flood sequences impacts the accuracy of regional predictors. Our study investigates this impact for two regionalization procedures, generalized least squares (GLS) regression and top-kriging (TK), which deal with cross-correlation in two fundamentally different ways and therefore might be associated with different accuracy and uncertainty of predicted flood quantiles. We perform a Monte Carlo experiment based on a dataset of annual maximum flood series for 20 catchments in a hydrologically homogeneous region. Based on a log-Pearson type III parent distribution, we generate 3000 realizations of the region with different degrees of cross-correlation. For each realization, GLS and TK are applied in leave-one-out cross-validation to predict at-site flood quantiles. Our study shows that (a) TK outperforms GLS when catchment area is the only catchment descriptor used for predicting “true” population (theoretical) flood quantiles, regardless of the level of cross-correlation, and (b) GLS and TK perform similarly when multiple catchment descriptors are used.
A comparison between generalized least squares regression and top-kriging for homogeneous cross-correlated flood regions / Persiano, S.; Salinas, J. L.; Stedinger, J. R.; Farmer, W. H.; Lun, D.; Viglione, A.; Bloschl, G.; Castellarin, A.. - In: HYDROLOGICAL SCIENCES JOURNAL. - ISSN 0262-6667. - 66:4(2021), pp. 565-579. [10.1080/02626667.2021.1879389]
A comparison between generalized least squares regression and top-kriging for homogeneous cross-correlated flood regions
Viglione A.;Castellarin A.
2021
Abstract
Spatial cross-correlation among flood sequences impacts the accuracy of regional predictors. Our study investigates this impact for two regionalization procedures, generalized least squares (GLS) regression and top-kriging (TK), which deal with cross-correlation in two fundamentally different ways and therefore might be associated with different accuracy and uncertainty of predicted flood quantiles. We perform a Monte Carlo experiment based on a dataset of annual maximum flood series for 20 catchments in a hydrologically homogeneous region. Based on a log-Pearson type III parent distribution, we generate 3000 realizations of the region with different degrees of cross-correlation. For each realization, GLS and TK are applied in leave-one-out cross-validation to predict at-site flood quantiles. Our study shows that (a) TK outperforms GLS when catchment area is the only catchment descriptor used for predicting “true” population (theoretical) flood quantiles, regardless of the level of cross-correlation, and (b) GLS and TK perform similarly when multiple catchment descriptors are used.File | Dimensione | Formato | |
---|---|---|---|
Persianoetal2021HSJ.pdf
accesso aperto
Tipologia:
2a Post-print versione editoriale / Version of Record
Licenza:
Creative commons
Dimensione
4.74 MB
Formato
Adobe PDF
|
4.74 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11583/2987701