Differences

This shows you the differences between two versions of the page.

--- working_groups:computational_earth_sciences:ai4es:applications [2018/01/10 09:53]
asanche2
+++ working_groups:computational_earth_sciences:ai4es:applications [2020/08/07 15:44] (current)
cgome1
@@ Line 1: / Line 1: @@
-==== Machine Learning Applications on Earth Sciences Dept. ====
+===== BSC-ES projects where the AI4ES research line plays an important or leading role =====
-These activities benefit from the disciplinary structure of the Earth Sciences department as well as the collaboration with other BSC departments like the Computer Sciences and external institutions.
-Some of the potential topics of interest for the department are listed below.
+For a more complete list of ongoing and potential use cases, see the strategy document.
-  * Classification Methods
-  * Multivariate Methods
-  * Deep Learning : Convolutional Neuronal Networks
-  * Causal Effect Networks for Time-series Analysis
-=== Atmospheric Composition ===
+  * AI for extreme extreme weather events and climate services (ASTERISKS H2020). Deep Neural Networks from the field of computer Vision can be used for pattern recognition in a supervised way, exploiting reanalysis data and historical event databases (e.g., IBTrACS for tropical cyclones) or other impact data.
+  * Learning to simulate precipitation with Deep Neural Networks ({{https://www.esiwace.eu/events/6th-hpc-workshop/presentations/learning-to-simulate-precipitation-with-deep-neural-networks|slides}}). Talk presented at the 6th ENES HPC workshop. We implemented encoder-decoder networks, such as those used for image translation in Computer Vision, for the task of learning precipitation fields from other predictor variables. The  transfer function aims to capture information from predictor variables to infer precipitation, a variable highly dependent on parameterizations schemes.
-Currently working on the application of classification methods
+  * Simulating precipitation with supervised and generative learning models. Abstract submitted to the ECMWF-ESA Workshop on Machine Learning for Earth System Observation and Prediction. We compare data-driven models based on supervised convolutional neural networks and conditional generative adversarial networks for simulating precipitation fields.
-for the effect evaluation of the variability of meteorological variables
-on the concentration of specific pollutants. Another topic of study is the improvement
+  * May Deep Learning contribute to improve Downscaling techniques? ({{:library:external:pen_a-izquierdo_downscalingwithdl_isms2020.pdf |poster}}). Presented at the ISMSM 2020 conference. We implemented encoder-decoder networks for the task of downscaling, that is taking climate variables from a low resolution simulation (predictors) for training an deep neural network to produce an observational gridded field at higher spatial resolution.
-of the emission model outputs by combining classification/Analogues methods with existing Kalman Filters (KF) algorithms.
-For more details, look at [[https://earth.bsc.es/wiki/doku.php?id=working_groups:ac:machine_learning_in_ac&#meetings|AC Machine Learning]]
-=== Climate and Weather Prediction ===
+  * Deep neural networks as surrogate models for creating or improving data-driven parameterizations. Collaboration with Leonie Esters (Uppsala University) focusing on sea-air gas exchange and ocean turbulence.
-== Hurricanes Prediction ==
-The use of Deep Learning methods, in particular convolutional neuronal networks, can be applied to
-Hurricanes and cyclones observational databases to predict the number of such extreme events for the subsequent years.
-== Bias Correction ==
-The use of the Analogues technique utilized to improve the CALIOPE forecast might also be implemented within the MEDSCOPE project
-to improve the Bias correction and forecast calibration. Mutivariate-Analysis will also play an important role in the development of multivariate scores using EOF approach.
-== Teleconnections and Multivariate analysis==
-Causal interships between different predictants like geopotential, sea level pressure, seat surface temperature, precipitation level, ozone concentration, etc .. might be characterized and identified using causal effect networks. These studies help scientist to evaluate to what extent variability modes occur and the impact of their influence into different components of the earth dynamics system(Atmosphere, Land, Ocean, Sea Ice).
-=== Earth Sciences Services ===
-This particular area benefits from the application of machine learning techniques like multivariate analysis among others,
-applied to the different research fields of the department to extract valuable information for decision making in a wide spectrum of applications such as society, agriculture, energy, health etc.
-=== Computational Earth Sciences ===
-== Current Developments ==
-  * Improvement in the characterization of uncertainty in multi-ensembles models like EC-Earth by using an adaptive methodology.[[https://earth.bsc.es/wiki/lib/exe/fetch.php?media=library:internal:adaptiveenssim_asanchez_2017.pptx_1_.pdf | adaptive]]
-  * Extending and adapting existing Classification /Clustering algorithms to the department tools.
-  * Deep Learning and GPUs applications to Online Diagnostics and Hurricanes forecast.
+  * Inference of surface pollutant concentrations from satellite and global scale meteorological and chemical simulations (AQ-WATCH project). Application to NO2 (TROPOMI sensor) and aerosols.
+  * Identification of Wastewater CH4 Emission Sources Using Deep Learning Applied to Sentinel-2 Observations. Abstract submitted to the ESA phi-week 2020 conference. Here we train object detection models from the field of computer vision for the task of localizing wastewater treatment plants and inferring the amount of processed water (as a proxy for CH4 emissions).

User Tools

Site Tools

Differences

Page Tools