A Simulation Model of Seawater Vertical Temperature by Using Back-Propagation Neural Network

Abstract

This study proposed a neural-network-based model to

An Artificial Neural Network to Infer the Mediterranean 3D Chlorophyll-a and Temperature Fields from Remote Sensing Observations

“…Along the water column, a maximum RMSE of 1.2 • C for temperature and 0.4 mgm −3 for chlorophyll was found within the range of about 10-60 m depth, namely where the highest variability is observed in both variables. The results of the temperature model validation agreed with most of similar approaches applied on other ocean/seas [20,75]. The statistics and general fit of chlorophyll validation on test set data demonstrated an improvement with respect to that already obtained in [48], with an r 2 = 0.71 and MeanAPE = 48% against the r 2 = 0.69 and MeanAPE = 57% of the network trained on in situ surface values and showed in [48].…”

“…Comparing the statistical results obtained from the MLP temperature validation with those shown in other works is not trivial, as most of them refer to global reconstructions or different specific regional oceans (e.g., North Atlantic, Southern Ocean), only few of which are focused on the Mediterranean basin, and also in those cases limited to specific sub-basins and fixed locations (e.g., [20,21]). Overall, most reconstructions show a RMSE close to our (0.7 • C) with values spanning in the range between 0.45 and 1.2 • C [35,[73][74][75]. Along the vertical, our MLP errors show similar shape and values of the same order of magnitude of those obtained for different seasons in [75].…”

“…Overall, most reconstructions show a RMSE close to our (0.7 • C) with values spanning in the range between 0.45 and 1.2 • C [35,[73][74][75]. Along the vertical, our MLP errors show similar shape and values of the same order of magnitude of those obtained for different seasons in [75].…”