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

Abstract

Remote sensing data provide a huge number of sea surface observations, but cannot give direct information on deeper ocean layers, which can only be provided by sparse in situ data. The combination of measurements collected by satellite and in situ sensors represents one of the most effective strategies to improve our knowledge of the interior structure of the ocean ecosystems. In this work, we describe a Multi-Layer-Perceptron (MLP) network designed to reconstruct the 3D fields of ocean temperature and chlorophyll-a concentration, two variables of primary importance for many upper-ocean bio-physical processes. Artificial neural networks can efficiently model eventual non-linear relationships among input variables, and the choice of the predictors is thus crucial to build an accurate model. Here, concurrent temperature and chlorophyll-a in situ profiles and several different combinations of satellite-derived surface predictors are used to identify the optimal model configuration, focusing on the Mediterranean Sea. The lowest errors are obtained when taking in input surface chlorophyll-a, temperature, and altimeter-derived absolute dynamic topography and surface geostrophic velocity components. Network training and test validations give comparable results, significantly improving with respect to Mediterranean climatological data (MEDATLAS). 3D fields are then also reconstructed from full basin 2D satellite monthly climatologies (1998–2015) and resulting 3D seasonal patterns are analyzed. The method accurately infers the vertical shape of temperature and chlorophyll-a profiles and their spatial and temporal variability. It thus represents an effective tool to overcome the in-situ data sparseness and the limits of satellite observations, also potentially suitable for the initialization and validation of bio-geophysical models.

Inversion of Phytoplankton Pigment Vertical Profiles from Satellite Data Using Machine Learning

“…Works attempting to predict vertical pigment profiles from surface data targeted the Chla and were based on the surface Chla and/or assigned with other physical factors such as SST and currents ( [13,14,[17][18][19][20]). However, during the optimization process of Sat2Profile, we showed that the problem is more complex when dealing with different pigments at the same time, each with their own particular variability.…”

The Mediterranean Sea we want

“…To tackle this challenge, the relevant gaps reported in Table 2 need to be addressed at first. A required step forward is the integration of satellite observations with the sparse in situ offshore and coastal observations into AI and/or numerical models to bring out the synoptic 4D description of Mediterranean (Sammartino et al, 2020) and its marine coastal areas (Melet et al, 2020). Several scientific challenges still exist related to the need of improving the observing capabilities at small spatial and temporal scales, to capture their variability, to improve model-data integration and uncertainty estimation (Tintoré et al, 2019), and to address the biological and ecological dimensions, which are essential components to meet the society demand.…”

Process-Oriented Estimation of Chlorophyll-a Vertical Profile in the Mediterranean Sea Using MODIS and Oceanographic Float Products

“…A deeper understanding of the dynamics and evolution of marine phytoplankton requires three dimensions (3D) observations of the algal abundance at different temporal and spatial scales and much wider and regular coverage than currently achievable (Sammartino et al, 2020). However, the traditional measurement is mainly based on in situ sampling either through coastal monitoring programs or time-limited oceanographic cruises, or fixed platforms such as moored buoys, which can accurately describe local conditions along the water column, but are clearly inadequate to describe processes occurring within the wide range of temporal and spatial scales impacted by undergoing changes (von Schuckmann et al, 2018).…”

“…Result-oriented approach is to infer Chl-a concentration at different depths directly from other measurable ocean variables. Due to the complexity of the marine ecosystem, this kind of method usually requires tools as artificial neural network (Sammartino et al, 2020) and its variants (Puissant et al, 2021), which are capable of revealing non-linear relations. Although the performance of such methods has been validated on regional and even global scale, the huge demand for input variables and computational resources greatly limits their utilization potentiality (Erickson et al, 2019;Lee et al, 2015).…”