Create abstract_flocco.md

assets/abstract_flocco.md

@@ -0,0 +1,20 @@
+---
+layout: minimal
+---
+
+## A double speed road: melt pond modelling and observations
+### *Daniela Flocco*, Ellen Buckley, Chris Horvat, Daniel Feltham, David Schroeder, Lucia Hosekova
+
+
+Melt ponds forming on the surface of Arctic sea ice during the melt season are known to play a critical role in the ice-albedo feedback mechanism and significantly impact sea ice dynamics and energy exchanges in polar regions. Modelling melt ponds is crucial for improving predictions of sea ice state and understanding broader climate feedbacks in a changing Arctic.
+
+They affect the sea ice -atmosphere balance through their low albedo and heat capacity and delay the basal ice growth at the end of summer. They impact the sea ice beneath them changing its temperature and permeability, and affect the upper layer of the ocean under the ice with their discharge of freshwater. Melt ponds also present a high variability in time and space and their parameterization in Global Climate Models is not trivial due to their size which is sub-grid scale compared to GCMs.
+
+Observations of melt ponds are extremely useful to study melt ponds characteristics and find emergent links with available sea ice data in order to refine melt pond parameterization that can be introduced in GCMs to better represent the spatial distribution of ponded sea ice.
+Machine Learning techniques have been used to detect and classify melt ponds from remote sensing data, focusing on deep learning approaches such as convolutional neural networks (CNNs), as well as unsupervised methods like clustering algorithms. These models, trained on multispectral, hyperspectral, and synthetic aperture radar (SAR) datasets, enable high-accuracy melt pond detection, even in challenging environmental conditions like cloud cover. These approaches are of fundamental importance to find new relationships that can help forecast the presence/absence of melt ponds. 
+
+In fact, the integration of melt pond dynamics into coupled ocean-atmosphere models has enhanced the accuracy of sea ice projections, though challenges remain in capturing the fine-scale heterogeneity of pond structures and their interactions with sea ice topography, permeability and snow cover. This presentation reviews the development and application of melt pond models, focusing on the representation of pond formation, evolution, and drainage processes. Future model improvements are discussed, with particular attention to the incorporation of remote sensing data and machine learning techniques, highlighted as a promising direction for future research.
+
+Understanding and refining melt pond models is essential for predicting future Arctic sea ice loss and its implications for global climate systems.
+
+[back to the Workshop page](https://nansencenter.github.io/superice-nersc/workshop/)