Classifying Global Low-Cloud Morphology with a Deep Learning Model: Results and Potential Use

2020 ◽  
Author(s):  
Tianle Yuan
Keyword(s):  
Cumulus Convection ◽  
Training Set ◽  
Team Members ◽  
Low Clouds ◽  
Low Cloud ◽  
One Step ◽  
Funded Project ◽  
Deep Learning Model ◽  
Observational Record ◽  
Potential Use

<p>Marine low clouds display rich mesoscale morphological types, distinct spatial patterns of cloud fields. Being able to differentiate low cloud morphology offers a tool for the research community to go one step beyond bulk cloud statistics such as cloud fraction and advance the understanding of low clouds. Here we report the progress of a NASA funded project that aims to create an observational record of low cloud mesoscale morphology at a near-global (60S-60N) scale. First, a training set is created by our team members manually labeling thousands of mesoscale (128x128) MODIS scenes into six different categories: stratus, closed cellular convection, disorganized convection, open cellular convection, clustered cumulus convection, and suppressed cumulus convection. Then we train a deep convolutional neural network model using this training set to classify individual MODIS scenes at 128x128 resolution, and test it on a test set. The trained model achieves a cross-type average precision of about 93%. We apply the trained model to 16 years of data over the Southeast Pacific. The resulting climatological distribution of low cloud morphology types show both expected and unexpected features and suggest promising potential for low cloud studies as a data product.</p>

scholarly journals Applying Deep Learning to NASA MODIS Data to Create a Community Record of Marine Low Cloud Mesoscale Morphology

2020 ◽  
Author(s):  
Tianle Yuan ◽  
Hua Song ◽  
Robert Wood ◽  
Johannes Mohrmann ◽  
Kerry Meyer ◽  
...  
Keyword(s):  
Cumulus Convection ◽  
Training Set ◽  
Average Precision ◽  
Team Members ◽  
Low Clouds ◽  
Low Cloud ◽  
One Step ◽  
Funded Project ◽  
Observational Record ◽  
Cross Type

Abstract. Marine low clouds display rich mesoscale morphological types, distinct spatial patterns of cloud fields. Being able to differentiate low cloud morphology offers a tool for the research community to go one step beyond bulk cloud statistics such as cloud fraction and advance the understanding of low clouds. Here we report the progress of a NASA funded project that aims to create an observational record of low cloud mesoscale morphology at a near-global (60S–60N) scale. First, a training set is created by our team members manually labeling thousands of mesoscale (128 x 128) MODIS scenes into six different categories: stratus, closed cellular convection, disorganized convection, open cellular convection, clustered cumulus convection, and suppressed cumulus convection. Then we train a deep convolutional neural network model using this training set to classify individual MODIS scenes at 128 x 128 resolution, and test it on a test set. The trained model achieves a cross-type average precision of about 93 %. We apply the trained model to 16 years of data over the Southeast Pacific. The resulting climatological distribution of low cloud morphology types show both expected and unexpected features and suggest promising potential for low cloud studies as a data product.

scholarly journals Applying deep learning to NASA MODIS data to create a community record of marine low-cloud mesoscale morphology

2020 ◽  
Vol 13 (12) ◽  
pp. 6989-6997
Author(s):  
Tianle Yuan ◽  
Hua Song ◽  
Robert Wood ◽  
Johannes Mohrmann ◽  
Kerry Meyer ◽  
...  
Keyword(s):  
Cumulus Convection ◽  
Training Set ◽  
Average Precision ◽  
Team Members ◽  
Southeastern Pacific ◽  
Low Clouds ◽  
Low Cloud ◽  
One Step ◽  
Observational Record ◽  
Cross Type

Abstract. Marine low clouds display rich mesoscale morphological types and distinct spatial patterns of cloud fields. Being able to differentiate low-cloud morphology offers a tool for the research community to go one step beyond bulk cloud statistics such as cloud fraction and advance the understanding of low clouds. Here we report the progress of our project that aims to create an observational record of low-cloud mesoscale morphology at a near-global (60∘ S–60∘ N) scale. First, a training set is created by our team members manually labeling thousands of mesoscale (128×128) MODIS scenes into six different categories: stratus, closed cellular convection, disorganized convection, open cellular convection, clustered cumulus convection, and suppressed cumulus convection. Then we train a deep convolutional neural network model using this training set to classify individual MODIS scenes at 128×128 resolution and test it on a test set. The trained model achieves a cross-type average precision of about 93 %. We apply the trained model to 16 years of data over the southeastern Pacific. The resulting climatological distribution of low-cloud morphology types shows both expected and unexpected features and suggests promising potential for low-cloud studies as a data product.

Instant 99mTc Compounds

1971 ◽  
Vol 10 (03) ◽  
pp. 245-251 ◽  
Author(s):  
P. Richards ◽  
W. C. Eckelman
Keyword(s):  
Physical Properties ◽  
High Purity ◽  
Analytical Techniques ◽  
Full Potential ◽  
Labeled Compounds ◽  
Ph Adjustment ◽  
One Step ◽  
Stannous Ion ◽  
99Mtc Radiopharmaceuticals ◽  
Potential Use

SummaryThe full potential use of technetium has not been achieved despite its ideal physical properties, dosimetry and availability because of the complex preparations required for 99mTc radiopharmaceuticals. One of the goals of our work is to develop techniques for the preparation of high-purity 99mTc compounds which can be easily prepared, ideally by adding pertechnetate to a prepared solution.The use of stannous ion as reducing agent for technetium makes it possible to obtain such one-step, high-purity products. All non-radioactive components can be premixed in a single vial before addition of the radioactive pertechnetate. No final pH adjustment, further chemical manipulation or purification is required.Procedures for two instantly labeled compounds have been developed to date: 99mTc DTPA and 99mTc HSA. The 99mTc DTPA is prepared by adding pertechnetate to a previously prepared solution of stannous ion and CaNa3 DTPA which has been stored at pH 4. The 99mTc HSA is prepared by adding pertechnetate to a solution of stannous ion and HSA. The parametric variations and analytical techniques involved in formulating these procedures are described. It appears that development of kits for other biologically interesting compounds may be possible using similar procedures.

scholarly journals Climatology of stratocumulus cloud morphologies: microphysical properties and radiative effects

2014 ◽  
Vol 14 (13) ◽  
pp. 6695-6716 ◽  
Author(s):  
A. Muhlbauer ◽  
I. L. McCoy ◽  
R. Wood
Keyword(s):  
Physical Properties ◽  
Global Scale ◽  
Cloud Fraction ◽  
Cloud Base ◽  
Trade Wind ◽  
Radiative Effects ◽  
Regional Variability ◽  
Microphysical Properties ◽  
Low Clouds ◽  
Low Cloud

Abstract. An artificial neural network cloud classification scheme is combined with A-train observations to characterize the physical properties and radiative effects of marine low clouds based on their morphology and type of mesoscale cellular convection (MCC) on a global scale. The cloud morphological categories are (i) organized closed MCC, (ii) organized open MCC and (iii) cellular but disorganized MCC. Global distributions of the frequency of occurrence of MCC types show clear regional signatures. Organized closed and open MCCs are most frequently found in subtropical regions and in midlatitude storm tracks of both hemispheres. Cellular but disorganized MCC are the predominant type of marine low clouds in regions with warmer sea surface temperature such as in the tropics and trade wind zones. All MCC types exhibit a pronounced seasonal cycle. The physical properties of MCCs such as cloud fraction, radar reflectivity, drizzle rates and cloud top heights as well as the radiative effects of MCCs are found highly variable and a function of the type of MCC. On a global scale, the cloud fraction is largest for closed MCC with mean cloud fractions of about 90%, whereas cloud fractions of open and cellular but disorganized MCC are only about 51% and 40%, respectively. Probability density functions (PDFs) of cloud fractions are heavily skewed and exhibit modest regional variability. PDFs of column maximum radar reflectivities and inferred cloud base drizzle rates indicate fundamental differences in the cloud and precipitation characteristics of different MCC types. Similarly, the radiative effects of MCCs differ substantially from each other in terms of shortwave reflectance and transmissivity. These differences highlight the importance of low-cloud morphologies and their associated cloudiness on the shortwave cloud forcing.

scholarly journals Characterisation and surface radiative impact of Arctic low clouds from the IAOOS field experiment

2020 ◽  
Author(s):  
Julia Maillard ◽  
François Ravetta ◽  
Jean-Christophe Raut ◽  
Vincent Mariage ◽  
Jacques Pelon
Keyword(s):  
Field Experiment ◽  
Arctic Ocean ◽  
Cloud Cover ◽  
Great Majority ◽  
The North ◽  
Low Clouds ◽  
Radiative Impact ◽  
Low Cloud ◽  
Ocean Observing ◽  
Layer Cloud

Abstract. The Ice, Atmosphere, Arctic Ocean Observing System (IAOOS) field experiment took place from 2014 to 2019. Over this period, more than 20 instrumented buoys were deployed at the North Pole. Once locked into the ice, the buoys drifted for periods of a month to more than a year. Some of these buoys were equipped with 808 nm wavelength lidars which acquired a total of 1805 profiles over the course of the campaign. This IAOOS lidar dataset is exploited to establish a novel statistic of cloud cover and of the geometrical and optical characteristics of the lowest cloud layer. Cloud frequency is globally at 75 %, and above 85 % from May to October. Single layers are thickest in October/November and thinnest in the summer. Meanwhile, their optical depth is maximum in October. On the whole, the cloud cover is very low, with the great majority of first layer bases beneath 120 m. In the shoulder seasons, surface temperatures are markedly warmer when the IAOOS profile contains at least one low cloud than when it does not. This temperature difference is statistically insignificant in the summer months. Indeed, summer clouds have a shortwave cooling effect which can reach −60 W m−2 and balance out their longwave warming effect.

scholarly journals Characterisation and surface radiative impact of Arctic low clouds from the IAOOS field experiment

2021 ◽  
Vol 21 (5) ◽  
pp. 4079-4101
Author(s):  
Julia Maillard ◽  
François Ravetta ◽  
Jean-Christophe Raut ◽  
Vincent Mariage ◽  
Jacques Pelon
Keyword(s):  
Field Experiment ◽  
Cloud Cover ◽  
Great Majority ◽  
Cloud Base ◽  
The North ◽  
Low Clouds ◽  
Radiative Impact ◽  
Low Cloud ◽  
Ocean Observing ◽  
Cloud Base Height

Abstract. The Ice, Atmosphere, Arctic Ocean Observing System (IAOOS) field experiment took place from 2014 to 2019. Over this period, more than 20 instrumented buoys were deployed at the North Pole. Once locked into the ice, the buoys drifted for periods of a month to more than a year. Some of these buoys were equipped with 808 nm wavelength lidars which acquired a total of 1777 profiles over the course of the campaign. This IAOOS lidar dataset is exploited to establish a novel statistic of cloud cover and of the geometrical and optical characteristics of the lowest cloud layer. The average cloud frequency from April to December over the course of the campaign was 75 %. Cloud occurrence frequencies were above 85 % from May to October. Single layers are thickest in October/November and thinnest in the summer. Meanwhile, their optical depth is maximum in October. On the whole, the cloud base height is very low, with the great majority of first layer bases beneath 120 m. In April and October, surface temperatures are markedly warmer when the IAOOS profile contains at least one low cloud than when it does not. This temperature difference is statistically insignificant in the summer months. Indeed, summer clouds have a shortwave cooling effect which can reach −60 W m−2 and balance out their longwave warming effect.

Improved Site-Specific Numerical Prediction of Fog and Low Clouds: A Feasibility Study

2005 ◽  
Vol 20 (4) ◽  
pp. 627-646 ◽  
Author(s):  
Thierry Bergot ◽  
Dominique Carrer ◽  
Joël Noilhan ◽  
Philippe Bougeault
Keyword(s):  
Winter Season ◽  
Integrated Approach ◽  
Atmospheric Measurements ◽  
Forecast System ◽  
Short Term ◽  
Low Clouds ◽  
Atmospheric Profiles ◽  
Low Cloud ◽  
Accurate Forecast ◽  
1D Model

Abstract Accurate short-term forecasts of low ceiling and visibility are vital to air traffic operation, in order to maximize the use of an airport. The research presented here uses specific local observations and a detailed numerical 1D model in an integrated approach. The goal of the proposed methodology is to improve the local prediction of poor visibility and low clouds at Paris’s Charles de Gaulle International Airport. In addition to the development of an integrated observations and model-based forecasting system, this study will try to assess whether or not the increased local observing network yields improvements in short-term forecasts of low ceiling and poor visibility. Tests have been performed in a systematic manner during 5 months (the 2002/03 winter season). Encouraging results show that the inclusion of dedicated observations into the local 1D forecast system provides significant improvement to the forecast. Inspection of events indicates that the improvement in very short-term forecasts is a consequence of the ability of the forecast system to more accurately characterize the boundary layer processes, especially during night. Accurate forecast of low cloud seems more difficult since it strongly depends on the 3D mesoscale flow. This study also demonstrates that the use of a 1D model to forecast fogs and low clouds could only be beneficial if it is associated with local measurements and with a local assimilation scheme. The assimilation procedure used in this study is based on different steps: in the first step the atmospheric profiles are estimated in a one-dimensional variational data assimilation (1DVAR) framework, in the second step these atmospheric profiles are modified when fog and/or low clouds are detected, and in the third step the soil profiles are estimated in order to keep the consistency between the soil state and atmospheric measurements.

Insights into deep learning for earthquake magnitude and location estimation

2021 ◽  
Author(s):  
Jannes Münchmeyer ◽  
Dino Bindi ◽  
Ulf Leser ◽  
Frederik Tilmann
Keyword(s):  
Deep Learning ◽  
Real Time ◽  
Early Warning ◽  
Source Parameters ◽  
Spectral Composition ◽  
Learning Model ◽  
Location Estimation ◽  
Training Set ◽  
Earthquake Source Parameters ◽  
Deep Learning Model

<p><span>The estimation of earthquake source parameters, in particular magnitude and location, in real time is one of the key tasks for earthquake early warning and rapid response. In recent years, several publications introduced deep learning approaches for these fast assessment tasks. Deep learning is well suited for these tasks, as it can work directly on waveforms and </span><span>can</span><span> learn features and their relation from data.</span></p><p><span>A drawback of deep learning models is their lack of interpretability, i.e., it is usually unknown what reasoning the network uses. Due to this issue, it is also hard to estimate how the model will handle new data whose properties differ in some aspects from the training set, for example earthquakes in previously seismically quite regions. The discussions of previous studies usually focused on the average performance of models and did not consider this point in any detail.</span></p><p><span>Here we analyze a deep learning model for real time magnitude and location estimation through targeted experiments and a qualitative error analysis. We conduct our analysis on three large scale regional data sets from regions with diverse seismotectonic settings and network properties: Italy and Japan with dense networks </span><span>(station spacing down to 10 km)</span><span> of strong motion sensors, and North Chile with a sparser network </span><span>(station spacing around 40 km) </span><span>of broadband stations. </span></p><p><span>We obtained several key insights. First, the deep learning model does not seem to follow the classical approaches for magnitude and location estimation. For magnitude, one would classically expect the model to estimate attenuation, but the network rather seems to focus its attention on the spectral composition of the waveforms. For location, one would expect a triangulation approach, but our experiments instead show indications of a fingerprinting approach. </span>Second, we can pinpoint the effect of training data size on model performance. For example, a four times larger training set reduces average errors for both magnitude and location prediction by more than half, and reduces the required time for real time assessment by a factor of four. <span>Third, the model fails for events with few similar training examples. For magnitude, this means that the largest event</span><span>s</span><span> are systematically underestimated. For location, events in regions with few events in the training set tend to get mislocated to regions with more training events. </span><span>These characteristics can have severe consequences in downstream tasks like early warning and need to be taken into account for future model development and evaluation.</span></p>

eTTCampus

2013 ◽  
pp. 528-540
Author(s):  
David E. Gray ◽  
Malcolm Ryan
Keyword(s):  
Decision Making ◽  
Project Management ◽  
Critical Analysis ◽  
Virtual Campus ◽  
Team Members ◽  
Management Processes ◽  
Funded Project ◽  
Effective Use ◽  
The Impact ◽  
Training Opportunities

This chapter critically examines innovative approaches to the evaluation of a European funded project involving nine countries in the development of a virtual campus to provide training opportunities in ICT for teachers and trainers across Europe. It explores project management processes and decision-making and the impact on outcomes as well as relationships between project team members. It concludes with recommendations for the more effective use of a range of these approaches, asserting that a critical analysis of the processes of engagement is as important as the outcomes.

scholarly journals Drivers of the Low-Cloud Response to Poleward Jet Shifts in the North Pacific in Observations and Models

2018 ◽  
Vol 31 (19) ◽  
pp. 7925-7947 ◽  
Author(s):  
Mark D. Zelinka ◽  
Kevin M. Grise ◽  
Stephen A. Klein ◽  
Chen Zhou ◽  
Anthony M. DeAngelis ◽  
...  
Keyword(s):  
Surface Temperature ◽  
North Pacific ◽  
Low Level ◽  
Cloud Coverage ◽  
The North ◽  
Low Clouds ◽  
Temperature Advection ◽  
Low Cloud ◽  
The North Pacific ◽  
Cloud Response

The long-standing expectation that poleward shifts of the midlatitude jet under global warming will lead to poleward shifts of clouds and a positive radiative feedback on the climate system has been shown to be misguided by several recent studies. On interannual time scales, free-tropospheric clouds are observed to shift along with the jet, but low clouds increase across a broad expanse of the North Pacific Ocean basin, resulting in negligible changes in total cloud fraction and top-of-atmosphere radiation. Here it is shown that this low-cloud response is consistent across eight independent satellite-derived cloud products. Using multiple linear regression, it is demonstrated that the spatial pattern and magnitude of the low-cloud-coverage response is primarily driven by anomalous surface temperature advection. In the eastern North Pacific, anomalous cold advection by anomalous northerly surface winds enhances sensible and latent heat fluxes from the ocean into the boundary layer, resulting in large increases in low-cloud coverage. Local increases in low-level stability make a smaller contribution to this low-cloud increase. Despite closely capturing the observed response of large-scale meteorology to jet shifts, global climate models largely fail to capture the observed response of clouds and radiation to interannual jet shifts because they systematically underestimate how sensitive low clouds are to surface temperature advection, and to a lesser extent, low-level stability. More realistic model simulations of cloud–radiation–jet interactions require that parameterizations more accurately capture the sensitivity of low clouds to surface temperature advection.

Sign in / Sign up

Export Citation Format

Share Document