Ferran Hernández-Macià

Mr. Ferran Hernández-Macià

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
31 May 2024 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
FHFerran Hernández-Macià
Ferran Hernández-Macià
CGCarolina Gabarro
Carolina Gabarro
GGGemma Sanjuan Gomez
Gemma Sanjuan Gomez
MEMaria Jose Escorihuela
Maria Jose Escorihuela
This study proposes a machine learning based methodology for estimating Arctic thin sea ice thickness (up to 1 m) from brightness temperature measurements of SMOS. The approach involves employing the so-called Burke model for sea ice emission modeling, integrating a suitable permittivity model and a radiative transfer equation. The training dataset is generated through a model-based simulation, and is then used to train and evaluate two machine learning regression algorithms: Random Forest and Gradient Boosting. Overall, this machine learning methodology results in great agreement with the ESA's official sea ice thickness product. Additionally, a validation performed by using data from mooring measurements shows a subtle improvement by the machine learning algorithms with respect to the ESA's official product. These results indicate their potential to surpass the performance of the current SMOS thin sea ice thickness retrievals.

21 October 2024 in Annals of Glaciology
Verified
Journal Article
Open access
PDF file
FHFerran Hernández-Macià
Ferran Hernández-Macià
CGCarolina Gabarró
Carolina Gabarró
MHMarcus Huntemann
Marcus Huntemann
RNReza Naderpour
Reza Naderpour
JJJoel T. Johnson
Joel T. Johnson
KJKenneth C. Jezek
Kenneth C. Jezek
The retrieval of sea ice thickness using L-band passive remote sensing requires robust models for emission from sea ice. In this work, measurements obtained from surface-based radiometers during the MOSAiC expedition are assessed with the Burke, Wilheit and SMRT radiative transfer models. These models encompass distinct methodologies: radiative transfer with/without wave coherence effects, and with/without scattering. Before running these emission models, the sea ice growth is simulated using the Cumulative Freezing Degree Days (CFDD) model to further compute the evolution of the ice structure during each period. Ice coring profiles done near the instruments are used to obtain the initial state of the computation, along with Digital Thermistor Chain (DTC) data to derive the sea ice temperature during the analyzed periods. The results suggest that the coherent approach used in the Wilheit model results in a better agreement with the horizontal polarization of the in situ measured brightness temperature. The Burke and SMRT incoherent models offer a more robust fit for the vertical component. These models are almost equivalent since the scattering considered in SMRT can be safely neglected at this low frequency, but the Burke model misses an important contribution from the snow layer above sea ice. The results also suggest that a more realistic permittivity falls between the spheres and random needles formulations, with potential for refinement, particularly for L-band applications, through future field measurements.

13 January 2026 in EGUsphere
Verified
Preprint
Open access
PDF file
Valentin LudwigValentin Ludwig
Valentin Ludwig
CRCaroline Ribere
Caroline Ribere
SFSara Fleury
Sara Fleury
CHChristian Haas
Christian Haas
MTMichel Tsamados
Michel Tsamados
ABAlessandro Di Bella
Alessandro Di Bella
Sea-ice thickness is a key component of the Arctic climate system, but yet a comprehensive assessment across observations and numerical models is still missing. Previous studies have either compared only a small ensemble of sea-ice thickness products, focused on a short time scale, or both. We use an ensemble of 23 harmonised large-scale satellite, model, reanalysis and multi-product data ranging partly from 1995 to 2024, but mostly from 2010 to 2023. The products are compared against reference data derived from upward-looking sonar measurements in the Beaufort Sea. We find that biases are typically in the range of −0.2 m to 0.3 m, root mean square deviations are usually between 0.25 m and 0.5 m, and correlation coefficients mostly fall between 0.7 and 0.85, although larger deviations occur in some cases. Satellite and multi-product data mostly have lower biases and lower root mean square deviations (RMSD), but similar correlation coefficients compared to models and reanalyses. We examine the reliability of the uncertainties stated by the providers of twelve products and find that, while individual products tend to state uncertainties that are either too small or too large relative to their actual difference towards reference data, the ensemble as a whole shows comparable magnitudes of uncertainties and difference towards reference data. Subsequently, we do a pairwise comparison between decadal averages of large-scale products. We find biases largely between 0.2 and 0.4 m, RMSDs largely between 0.4 and 0.9 m and correlation coefficients largely between 0.5 and 0.8. Our study concludes with a time-series analysis of sea-ice thickness for each category (model/reanalysis, satellite, multi-product) in November and March for 2010–2023 and 1995–2023, with the second period being limited to the region south of 81.5° N. For the first period, we find no significant trend in any category for both months. For the second period, we find that sea-ice thickness has declined by roughly 0.5–0.6 m in November and 0.3–0.4 m in March, with stronger trends for models/reanalyses and multi-product data than for satellite products.

28 July 2025 in Computers
Verified
Journal Article
Open access
PDF file
FHFerran Hernández-Macià
Ferran Hernández-Macià
GGGemma Sanjuan Gomez
Gemma Sanjuan Gomez
CGCarolina Gabarro
Carolina Gabarro
MEMaria Jose Escorihuela
Maria Jose Escorihuela
This study evaluates machine learning-based methods for retrieving thin Arctic sea ice thickness (SIT) from L-band radiometry, using data from the European Space Agency’s (ESA) Soil Moisture and Ocean Salinity (SMOS) satellite. In addition to the operational ESA product, three alternative approaches are assessed: a Random Forest (RF) algorithm, a Convolutional Neural Network (CNN) that incorporates spatial coherence, and a Long Short-Term Memory (LSTM) neural network designed to capture temporal coherence. Validation against in situ data from the Beaufort Gyre Exploration Project (BGEP) moorings and the ESA SMOSice campaign demonstrates that the RF algorithm achieves robust performance comparable to the ESA product, despite its simplicity and lack of explicit spatial or temporal modeling. The CNN exhibits a tendency to overestimate SIT and shows higher dispersion, suggesting limited added value when spatial coherence is already present in the input data. The LSTM approach does not improve retrieval accuracy, likely due to the mismatch between satellite resolution and the temporal variability of sea ice conditions. These results highlight the importance of L-band sea ice emission modeling over increasing algorithm complexity and suggest that simpler, adaptable methods such as RF offer a promising foundation for future SIT retrieval efforts. The findings are relevant for refining current methods used with SMOS and for developing upcoming satellite missions, such as ESA’s Copernicus Imaging Microwave Radiometer (CIMR).

17 June 2026 in EGUsphere
Verified
Preprint
Open access
PDF file
Marta UmbertMarta Umbert
Marta Umbert
NHNina Hoareau
Nina Hoareau
JCJúlia Crespin
Júlia Crespin
MSMaria Sánchez-Urrea
Maria Sánchez-Urrea
FHFerran Hernández-Macià
Ferran Hernández-Macià
EDEva De-Andrés
Eva De-Andrés
In situ hydrographic observations in Greenland fjords remain sparse, particularly during late winter and early spring, despite their importance for Arctic freshwater export, glacier–ocean interactions, and the seasonal development of stratification. A major limitation of many existing hydrographic datasets is the poor representation of the near-surface layer, where meltwater lenses, sharp salinity gradients, and strong density gradients are often concentrated within the upper few metres. Here we present a quality-controlled Level 2 conductivity, temperature, and depth dataset comprising 73 vertical hydrographic profiles collected across Southwest Greenland fjords (60.2–69.5° N, 43.6–53.9° W) during two expeditions aboard the icebreaker Le Commandant Charcot: the Sea Ice Measurements for Satellite thickness retrieval Validation campaign (32 profiles, 24 March–4 April 2025) and the Freshwater fluxes and Atlantic-Nordic Seas hydrography campaign (41 profiles, 18–28 April 2025). The dataset includes pressure, temperature, conductivity, practical salinity, and derived thermodynamic variables. Two RBR Concerto3 conductivity, temperature, and depth instruments sampling at 16 Hz and 2 Hz were processed through a reproducible eight-step workflow including atmospheric-pressure correction, automatic soak detection, signal conditioning, loop removal, thermodynamic calculations, and automated quality control following international ocean-data recommendations. Unlike many standard conductivity, temperature, and depth products, the processing strategy preserves the shallowest valid observations and the native vertical sampling. This allows users to investigate thin freshwater lenses and sharp near-surface gradients rather than systematically removing the upper few metres of the water column. The dataset therefore provides a detailed view of the onset of meltwater-driven stratification and offers the flexibility to apply alternative vertical averaging or filtering strategies depending on the objectives of future analyses. Thermohaline diagnostics identify four main hydrographic regimes: coastal and glacier meltwater-influenced waters, Polar Water, Winter Water, and modified Atlantic or Irminger Water. The comparison between the late-winter and early-spring campaigns documents the onset of surface restratification across Southwest Greenland fjords. The strongest freshening and shallow stratification are observed in fjords directly influenced by tidewater glaciers, while deeper water masses remain comparatively stable over the one-month period. The dataset is distributed in NetCDF-4 format compliant with Climate and Forecast Metadata Conventions version 1.8 and Attribute Convention for Data Discovery version 1.3, and is publicly available through Zenodo. The processing code is archived on GitHub and Zenodo. By preserving the upper-ocean structure during a critically undersampled season, this dataset provides a valuable resource for studies of Arctic freshwater dynamics, glacier–ocean interactions, fjord stratification, ocean model evaluation, and satellite product validation.

3 August 2026
Verified
Other
Open access
PDF file
ZYZiyu Yan
Ziyu Yan
YYYufang Ye
Yufang Ye
GHGeorg Heygster
Georg Heygster
BCBin Cheng
Bin Cheng
CGCarolina Gabarro
Carolina Gabarro
XCXiao Cheng
Xiao Cheng
The complexity of passive microwave (PM) retrieval of snow depth over Arctic sea ice stems from non-linear interactions between snow microstructure, wetness, and basal ice properties. These mechanisms remain insufficiently quantified, resulting in large uncertainties in PM-based snow products. We employ the snow microwave radiative transfer (SMRT) model together with a global sensitivity analysis, i.e. the Extended Fourier Amplitude Sensitivity Test, to decompose SMRT-simulated TB variance into contributions from individual parameters and their interactions. Averaging kernel analysis is then used to quantify snow depth retrievability across standard PM channels from 6 to 89 GHz under single- and multi-layer snowpack scenarios. 1) For single-layer dry snow, snow depth, density and grain radius are strongly coupled to each other, dominating the PM signals. When liquid water is present in the snow, the PM signals are primarily controlled by snow density and liquid water content. 2) In multi-layer dry snow, channels below 23 GHz are strongly influenced by the basal snow ice, while those above or equal to 23 GHz are dominated by depth hoar. At 6 GHz, retrievability is limited to dry snow with grain radius ≥ 0.5 mm and density ≤ 250 kg m-³, expanding toward finer grains and higher densities with increasing frequency. Regarding gradient ratio (GR), GR(18/6) provides limited retrievability for grain radius < 0.5 mm, whereas GR(36/18) remains effective for grain radius >0.2 mm. Notably, incorporating 89 GHz in GR improves the retrievability for new snow. Furthermore, sea ice type exerts a significant constraint on GR retrievability of snow depth and becomes increasingly pronounced under fine-grained snow conditions.