A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems
<p>A neural-network-based method, quantile regression neural networks (QRNNs), is proposed as a novel approach to estimating the a posteriori distribution of Bayesian remote sensing retrievals. The advantage of QRNNs over conventional neural network retrievals is that they learn to predict...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2018-08-01
|
| Series: | Atmospheric Measurement Techniques |
| Online Access: | https://www.atmos-meas-tech.net/11/4627/2018/amt-11-4627-2018.pdf |
| _version_ | 1811257702606700544 |
|---|---|
| author | S. Pfreundschuh P. Eriksson D. Duncan B. Rydberg N. Håkansson A. Thoss |
| author_facet | S. Pfreundschuh P. Eriksson D. Duncan B. Rydberg N. Håkansson A. Thoss |
| author_sort | S. Pfreundschuh |
| collection | DOAJ |
| description | <p>A neural-network-based method, quantile regression neural networks (QRNNs), is
proposed as a novel approach to estimating the a posteriori distribution of
Bayesian remote sensing retrievals. The advantage of QRNNs over conventional
neural network retrievals is that they learn to predict not only a single
retrieval value but also the associated, case-specific uncertainties. In this
study, the retrieval performance of QRNNs is characterized and compared to
that of other state-of-the-art retrieval methods. A synthetic retrieval
scenario is presented and used as a validation case for the application of
QRNNs to Bayesian retrieval problems. The QRNN retrieval performance is
evaluated against Markov chain Monte Carlo simulation and another Bayesian
method based on Monte Carlo integration over a retrieval database. The
scenario is also used to investigate how different hyperparameter
configurations and training set sizes affect the retrieval performance. In the
second part of the study, QRNNs are applied to the retrieval of cloud top
pressure from observations by the Moderate Resolution Imaging
Spectroradiometer (MODIS). It is shown that QRNNs are not only capable of
achieving similar accuracy to standard neural network retrievals but also
provide statistically consistent uncertainty estimates for non-Gaussian
retrieval errors. The results presented in this work show that QRNNs are able
to combine the flexibility and computational efficiency of the machine
learning approach with the theoretically sound handling of uncertainties of
the Bayesian framework. Together with this article, a Python implementation of
QRNNs is released through a public repository to make the method available to
the scientific community.</p> |
| first_indexed | 2024-04-12T18:01:34Z |
| format | Article |
| id | doaj.art-84ae848833404322a508b0524effe0d6 |
| institution | Directory Open Access Journal |
| issn | 1867-1381 1867-8548 |
| language | English |
| last_indexed | 2024-04-12T18:01:34Z |
| publishDate | 2018-08-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Atmospheric Measurement Techniques |
| spelling | doaj.art-84ae848833404322a508b0524effe0d62022-12-22T03:22:09ZengCopernicus PublicationsAtmospheric Measurement Techniques1867-13811867-85482018-08-01114627464310.5194/amt-11-4627-2018A neural network approach to estimating a posteriori distributions of Bayesian retrieval problemsS. Pfreundschuh0P. Eriksson1D. Duncan2B. Rydberg3N. Håkansson4A. Thoss5Department of Space, Earth and Environment, Chalmers University of Technology, Gothenburg, SwedenDepartment of Space, Earth and Environment, Chalmers University of Technology, Gothenburg, SwedenDepartment of Space, Earth and Environment, Chalmers University of Technology, Gothenburg, SwedenMöller Data Workflow Systems AB, Gothenburg, SwedenSwedish Meteorological and Hydrological Institute (SMHI), Norrköping, SwedenSwedish Meteorological and Hydrological Institute (SMHI), Norrköping, Sweden<p>A neural-network-based method, quantile regression neural networks (QRNNs), is proposed as a novel approach to estimating the a posteriori distribution of Bayesian remote sensing retrievals. The advantage of QRNNs over conventional neural network retrievals is that they learn to predict not only a single retrieval value but also the associated, case-specific uncertainties. In this study, the retrieval performance of QRNNs is characterized and compared to that of other state-of-the-art retrieval methods. A synthetic retrieval scenario is presented and used as a validation case for the application of QRNNs to Bayesian retrieval problems. The QRNN retrieval performance is evaluated against Markov chain Monte Carlo simulation and another Bayesian method based on Monte Carlo integration over a retrieval database. The scenario is also used to investigate how different hyperparameter configurations and training set sizes affect the retrieval performance. In the second part of the study, QRNNs are applied to the retrieval of cloud top pressure from observations by the Moderate Resolution Imaging Spectroradiometer (MODIS). It is shown that QRNNs are not only capable of achieving similar accuracy to standard neural network retrievals but also provide statistically consistent uncertainty estimates for non-Gaussian retrieval errors. The results presented in this work show that QRNNs are able to combine the flexibility and computational efficiency of the machine learning approach with the theoretically sound handling of uncertainties of the Bayesian framework. Together with this article, a Python implementation of QRNNs is released through a public repository to make the method available to the scientific community.</p>https://www.atmos-meas-tech.net/11/4627/2018/amt-11-4627-2018.pdf |
| spellingShingle | S. Pfreundschuh P. Eriksson D. Duncan B. Rydberg N. Håkansson A. Thoss A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems Atmospheric Measurement Techniques |
| title | A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems |
| title_full | A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems |
| title_fullStr | A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems |
| title_full_unstemmed | A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems |
| title_short | A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems |
| title_sort | neural network approach to estimating a posteriori distributions of bayesian retrieval problems |
| url | https://www.atmos-meas-tech.net/11/4627/2018/amt-11-4627-2018.pdf |
| work_keys_str_mv | AT spfreundschuh aneuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT periksson aneuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT dduncan aneuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT brydberg aneuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT nhakansson aneuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT athoss aneuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT spfreundschuh neuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT periksson neuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT dduncan neuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT brydberg neuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT nhakansson neuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems AT athoss neuralnetworkapproachtoestimatingaposterioridistributionsofbayesianretrievalproblems |