| Title | Linear and evolutionary polynomial regression models to forecast coastal dynamics: Comparison and reliability assessment |
| Abstract | In this paper, the Evolutionary Polynomial Regression data modelling strategy has been applied to study small
scale, short-term coastal morphodynamics, given its capability for treating a wide database of known information,
non-linearly. Simple linear and multilinear regression models were also applied to achieve a balance between
the computational load and reliability of estimations of the three models. In fact, even though it is easy
to imagine that the more complex the model, the more the prediction improves, sometimes a "slight" worsening
of estimations can be accepted in exchange for the time saved in data organization and computational load. The
models' outcomes were validated through a detailed statistical, error analysis, which revealed a slightly better
estimation of the polynomial model with respect to the multilinear model, as expected. On the other hand,
even though the data organization was identical for the two models, the multilinear one required a simpler
simulation setting and a faster run time. Finally, the most reliable evolutionary polynomial regression model
was used in order to make some conjecture about the uncertainty increase with the extension of extrapolation
time of the estimation. The overlapping rate between the confidence band of the mean of the known coast position
and the prediction band of the estimated position can be a good index of the weakness in producing reliable
estimations when the extrapolation time increases too much. The proposed models and tests have been applied
to a coastal sector located nearby Torre Colimena in the Apulia region, south Italy. |
| Source | Geomorphology (Amst.) 300 (300), pp. 128–140 |
| Keywords | Evolutionary polynomial regressionMultilinear regressionCoastal morphodynamicsCoastal erosion |
| Journal | Geomorphology (Amst.) |
| Editor | Elsevier, Oxford ;, Paesi Bassi |
| Year | 2018 |
| Type | Articolo in rivista |
| DOI | 10.1016/j.geomorph.2017.10.012 |
| Authors | Delia Evelina Bruno, Emanuele Barca, Rodrigo Mikosz Goncalves, heithor Alexandre de Araujo Queiroz, Luigi Berardi, Giuseppe Passarella |
| Text | 377735 2018 10.1016/j.geomorph.2017.10.012 ISI Web of Science WOS WOS 000418981700009 Scopus 2 s2.0 85033606790 Science direct Elsevier S0169555X16309229 Evolutionary polynomial regression Multilinear regression Coastal morphodynamics Coastal erosion Linear and evolutionary polynomial regression models to forecast coastal dynamics Comparison and reliability assessment Delia Evelina Bruno, Emanuele Barca, Rodrigo Mikosz Goncalves, heithor Alexandre de Araujo Queiroz, Luigi Berardi, Giuseppe Passarella Water Research Institute, National Research Council, IRSA, CNR, Viale F. De Blasio, n. 5, 70125 Bari, Italy; Department of Cartography Engineering, Federal University of Pernambuco UFPE , Geodetic Science and Technology of Geoinformation Post Graduation Program, Recife 50670 901, PE, Brazil; Polytechnic University of Bari, Civil Engineering and Architecture Department, Via Edoardo Orabona, 4, 70125 Bari, Italy In this paper, the Evolutionary Polynomial Regression data modelling strategy has been applied to study small scale, short term coastal morphodynamics, given its capability for treating a wide database of known information, non linearly. Simple linear and multilinear regression models were also applied to achieve a balance between the computational load and reliability of estimations of the three models. In fact, even though it is easy to imagine that the more complex the model, the more the prediction improves, sometimes a slight worsening of estimations can be accepted in exchange for the time saved in data organization and computational load. The models outcomes were validated through a detailed statistical, error analysis, which revealed a slightly better estimation of the polynomial model with respect to the multilinear model, as expected. On the other hand, even though the data organization was identical for the two models, the multilinear one required a simpler simulation setting and a faster run time. Finally, the most reliable evolutionary polynomial regression model was used in order to make some conjecture about the uncertainty increase with the extension of extrapolation time of the estimation. The overlapping rate between the confidence band of the mean of the known coast position and the prediction band of the estimated position can be a good index of the weakness in producing reliable estimations when the extrapolation time increases too much. The proposed models and tests have been applied to a coastal sector located nearby Torre Colimena in the Apulia region, south Italy. 300 Published version https //www.sciencedirect.com/science/article/abs/pii/S0169555X16309229 15/10/2017 Bruno et al., Geomorphology 2018, Linear and evolutionary polynomial regression models to forecast coastal dynamics Comparison and reliability ... 1 s2.0 S0169555X16309229 main.pdf Articolo in rivista Elsevier 0169 555X Geomorphology Amst. Geomorphology Amst. Geomorphology Amst. Geomorphology. Amst. giuseppe.passarella PASSARELLA GIUSEPPE emanuele.barca BARCA EMANUELE deliaevelina.bruno BRUNO DELIA EVELINA TA.P04.005.008 Integrazione di metodologie per il monitoraggio e la modellizzazione per la gestione delle risorse idriche |
