Low Cloud Detection in Multilayer Scenes using Satellite Imagery with Machine Learning Methods

2021 ◽  
Keyword(s):  
Machine Learning ◽  
Infrared Imaging ◽  
Weather Forecasting ◽  
Auxiliary Information ◽  
Probability Of Detection ◽  
Cloud Detection ◽  
Brightness Temperatures ◽  
Low Clouds ◽  
Low Cloud ◽  
Satellite Sensors

Abstract The detection of multilayer clouds in the atmosphere can be particularly challenging from passive visible and infrared imaging radiometers since cloud boundary information is limited primarily to the topmost cloud layer. Yet detection of low clouds in the atmosphere is important for a number of applications, including aviation nowcasting and general weather forecasting. In this work, we develop pixel-based machine learning-based methods of detecting low clouds, with a focus on improving detection in multilayer cloud situations and specific attention given to improving the Cloud Cover Layers (CCL) product, which assigns cloudiness in a scene into vertical bins. The Random Forest (RF) and Neural Network (NN) implementations use inputs from a variety of sources, including GOES Advanced Baseline Imager (ABI) visible radiances, infrared brightness temperatures, auxiliary information about the underlying surface, and relative humidity (which holds some utility as a cloud proxy). Training and independent validation enlists near-global, actively-sensed cloud boundaries from the radar and lidar systems onboard the CloudSat and CALIPSO satellites. We find that the RF and NN models have similar performances. The probability of detection (PoD) of low cloud increases from 0.685 to 0.815 when using the RF technique instead of the CCL methodology, while the false alarm ratio decreases. The improved PoD of low cloud is particularly notable for scenes that appear to be cirrus from an ABI perspective, increasing from 0.183 to 0.686. Various extensions of the model are discussed, including a nighttime-only algorithm and expansion to other satellite sensors.

PATMOS-x v6.0: Improvements to AVHRR Cloud Climate Record and Analysis of the Updated Data

2020 ◽  
Author(s):  
Coda Phillips ◽  
Michael Foster ◽  
Andrew Heidinger
Keyword(s):  
High Resolution ◽  
Detection Accuracy ◽  
Cloud Detection ◽  
Edge Preserving ◽  
Brightness Temperatures ◽  
Cloud Properties ◽  
Satellite Sensors ◽  
The Difference ◽  
Climate Analysis ◽  
Spectral Channels

<p>Since 1978, an Advanced Very-High-Resolution Radiometer (AVHRR) has flown onboard 17 polar-orbiting satellites. Together, they are the longest global record from a homogeneous set of satellite sensors. The Pathfinder Atmosphere’s Extended (PATMOS-x) dataset is a long-term cloud record derived from the AVHRR radiances, and suitable for climate analysis. It has demonstrated intersensor stability and has been rigorously compared with other cloud datasets.</p><p>However, the AVHRR alone has only limited spectral information, so cloud detection during nighttime or over ice is challenging. Therefore, performance degrades over regions with extreme diurnal patterns or low temperatures such as the poles, despite our interest.</p><p>The next production version of PATMOS-x will include numerous algorithmic changes as well as the use of High-resolution Infrared Radiation Sounder (HIRS) spectral channels to improve detection accuracy in previously difficult conditions. The low-resolution HIRS soundings are upsampled to match the AVHRR pixels through an edge-preserving process called “fusion”. The higher-resolution AVHRR imagery guides the upsampling and the resulting combination is spectrally consistent with the AVHRR and has a high spatial resolution.</p><p>For cloud detection, the difference between the AVHRR and HIRS 11μm and HIRS 6.7μm brightness temperatures has been added as a feature in the naive Bayesian cloud detector. The effect on cloud precision is seen especially in the Antarctic where false-positive cloud detections have decreased dramatically.</p><p>Other cloud properties can be improved with the new spectral channels. For example, the new cloud phase algorithm uses the HIRS 6.7μm to determine cloud phase and the AVHRR and HIRS 11μm-13.3μm beta ratio identifies overlapping clouds. Also, the 11μm, 12μm, and HIRS 13.3μm are used in the new cloud height algorithm.</p><p>We report on the development of this new version of the PATMOS-x cloud climate dataset, and the methods used to calibrate and homogenize the participating sensors. Finally, observed trends in the improved dataset will be examined and related to the old dataset. In particular, attention will be given to whether high-latitude analysis of climatic trends is finally possible on the new dataset.</p>

scholarly journals First fully diurnal fog and low cloud satellite detection reveals life cycle in the Namib

2018 ◽  
Vol 11 (10) ◽  
pp. 5461-5470 ◽  
Author(s):  
Hendrik Andersen ◽  
Jan Cermak
Keyword(s):  
Typical Feature ◽  
Probability Of Detection ◽  
Data Sets ◽  
Infrared Observations ◽  
Low Clouds ◽  
Low Cloud ◽  
Radiation Measurements ◽  
Term Data

Abstract. Fog and low clouds (FLCs) are a typical feature along the southwestern African coast, especially in the central Namib, where fog constitutes a valuable resource of water for many ecosystems. In this study, a novel algorithm is presented to detect FLCs over land from geostationary satellite data using only infrared observations. The algorithm is the first of its kind as it is stationary in time and thus able to reveal a detailed view of the diurnal and spatial patterns of FLCs in the Namib region. A validation against net radiation measurements from a station network in the central Namib reveals a high overall accuracy with a probability of detection of 94 %, a false-alarm rate of 12 % and an overall correctness of classification of 97 %. The average timing and persistence of FLCs seem to depend on the distance to the coast, suggesting that the region is dominated by advection-driven FLCs. While the algorithm is applied to study Namib-region fog and low clouds, it is designed to be transferable to other regions and can be used to retrieve long-term data sets.

Accounting for Non-detects in Satellite Retrievals: Application Using CrIS Ammonia Observations

2021 ◽  
Author(s):  
Evan White ◽  
Mark Shephard ◽  
Karen Cady-Periera ◽  
Shailesh Kharol ◽  
Enrico Dammers ◽  
...  
Keyword(s):  
Detection Limit ◽  
Survival Data ◽  
Infrared Imaging ◽  
Ammonia Concentration ◽  
Detection Algorithm ◽  
Atmospheric Conditions ◽  
Cloud Detection ◽  
Brightness Temperatures ◽  
Source Regions ◽  
Radiometric Signal

<p>For measurements from any instrument there is a minimum detection limit below which the sensor cannot measure (i.e., non-detects). Measurements of trace gases from satellite instruments can also suffer from a significant number of non-detects, especially for species with very low atmospheric concentrations  and that have a very weak or absent signals (signal-to-noise<1) in the spectral region used to detect the species (e.g., ammonia).  For ammonia, these non-signal conditions generally occur when thick clouds obscure the ammonia signal, or atmospheric conditions generates too weak of a radiometric signal to detect (e.g., very low concentrations). Presented is a robust approach to explicitly identify and account for cloud-free satellite observations that are below the detection limit of the sensor (which occur principally in  non-source regions) for the Cross-Track Infrared Sounder (CrIS) Fast Physical Retrieval (CFPR) ammonia (NH<sub>3</sub>) product. This approach uses the newly developed CrIS Ammonia Cloud Detection Algorithm (CACDA) to compute a cloud flag based on the CrIS IMG (CIMG) product . The CIMG product uses coincident Visible Infrared Imaging Radiometer Suite (VIIRS) brightness temperatures and cloud fractions mapped onto CrIS Field of Views (FOV). This cloud flag is used to separate CrIS FOVs without signal due to clouds from FOVs that are below the detection limit due to the atmospheric state (referred to as non-detects).  Survival data is generated from in-situ surface observations from non-emission source regions to produce ammonia concentration values under CrIS non-detect conditions. Accounting for these non-detects can be significant in reducing bias of averaged values (i.e., Level 3 products) in regions or conditions with low concentration amounts (e.g. wintertime, non-agriculture regions, etc.), with little impact on concentrations in emission regions. This presentation will provide examples and evaluations of the CACDA and the inclusion of non-detects in the CFPR generated ammonia product. This will include comparisons of annual and seasonal averages of surface level ammonia concentrations with and without survival data to demonstrate the reduction in bias.</p>

scholarly journals First fully-diurnal fog and low cloud satellite detection reveals life cycle in the Namib

2018 ◽  
Author(s):  
Hendrik Andersen ◽  
Jan Cermak
Keyword(s):  
Typical Feature ◽  
Probability Of Detection ◽  
Data Sets ◽  
Infrared Observations ◽  
Low Clouds ◽  
Low Cloud ◽  
Radiation Measurements ◽  
Term Data

Abstract. Fog and low clouds (FLC) are a typical feature along the southwestern African coast, especially in the central Namib, where fog constitutes a valuable resource of water for many ecosystems. In this study, a novel algorithm to detect FLC over land from geostationary satellite data using only infrared observations is presented. The algorithm is the first of its kind as it is stationary in time and thus able to reveal a detailed view into the diurnal and spatial patterns of FLC in the Namib region. A validation against net radiation measurements from a station network in the central Namib reveals a high overall accuracy with a probability of detection of 94 %, a false alarm rate of 12 % and an overall correctness of classification of 97 %. The average timing and persistence of FLC seem to depend on the distance to the coast, suggesting that the region is dominated by advection-driven FLC. While the algorithm is applied to study Namib-region fog and low clouds, it is designed to be transferable to other regions and can be used to retrieve long-term data sets.

scholarly journals Investigation of Low-Cloud Characteristics Using Mesoscale Numerical Model Data for Improvement of Fog-Detection Performance by Satellite Remote Sensing

2014 ◽  
Vol 53 (10) ◽  
pp. 2246-2263 ◽  
Author(s):  
Haruma Ishida ◽  
Kentaro Miura ◽  
Teruaki Matsuda ◽  
Kakuji Ogawara ◽  
Azumi Goto ◽  
...  
Keyword(s):  
Mesoscale Model ◽  
Potential Temperature ◽  
Vertical Gradient ◽  
Meteorological Conditions ◽  
Lapse Rate ◽  
Cloud Detection ◽  
Ancillary Data ◽  
Low Clouds ◽  
Lower Cloud ◽  
Low Cloud

AbstractThe comprehensive relationship between meteorological conditions and whether low water cloud touches the surface, particularly at sea, is examined with the goal of improving low-cloud detection by satellite. Gridpoint-value data provided by an operational mesoscale model with integration of Multifunction Transport Satellite-2 data can provide sufficient data for statistical analyses to find general parameters that can discern whether low clouds touch the surface, compensating for uncertainty due to the scarcity of observation sites at sea and the infrequent incidence of fog. The analyses reveal that surface-touching low clouds tend to have lower cloud-top heights than those not touching the surface, although the frequency distribution of cloud-top height differs by season. The bottom of the Γ > Γm layer (where Γ and Γm are the vertical gradient and the moist-adiabatic lapse rate of the potential temperature, respectively) with surface-touching low-cloud layers tends to be very low or almost attached to the surface. In contrast, the tops of low-cloud layers not touching the surface tend to occur near the bottom of the Γ > Γm layer. Mechanisms to correlate these meteorological conditions with whether low clouds touch the surface are inferred from investigations into the vertical structure of equivalent potential temperature. These results indicate that the temperature difference between cloud-top height and the surface can be an appropriate parameter to infer whether low clouds touch the surface. It is also suggested that only a little addition of meteorological ancillary data, such as the forecast sea surface temperature, to satellite data allows successful performance of the discrimination.

scholarly journals A Rail-Temperature-Prediction Model Based on Machine Learning: Warning of Train-Speed Restrictions Using Weather Forecasting

Sensors ◽  
2021 ◽  
Vol 21 (13) ◽  
pp. 4606
Author(s):  
Sunguk Hong ◽  
Cheoljeong Park ◽  
Seongjin Cho
Keyword(s):  
Machine Learning ◽  
Prediction Model ◽  
Weather Forecasting ◽  
Thermal Environment ◽  
Management Plan ◽  
Temperature Prediction ◽  
Railway Systems ◽  
Train Speed ◽  
Speed Restrictions ◽  
Entire Network

Predicting the rail temperature of a railway system is important for establishing a rail management plan against railway derailment caused by orbital buckling. The rail temperature, which is directly responsible for track buckling, is closely related to air temperature, which continuously increases due to global warming effects. Moreover, railway systems are increasingly installed with continuous welded rails (CWRs) to reduce train vibration and noise. Unfortunately, CWRs are prone to buckling. This study develops a reliable and highly accurate novel model that can predict rail temperature using a machine learning method. To predict rail temperature over the entire network with high-prediction performance, the weather effect and solar effect features are used. These features originate from the analysis of the thermal environment around the rail. Precisely, the presented model has a higher performance for predicting high rail temperature than other models. As a convenient structural health-monitoring application, the train-speed-limit alarm-map (TSLAM) was also proposed, which visually maps the predicted rail-temperature deviations over the entire network for railway safety officers. Combined with TSLAM, our rail-temperature prediction model is expected to improve track safety and train timeliness.

scholarly journals The short life of the volcanic island New Late’iki (Tonga) analyzed by multi-sensor remote sensing data

2020 ◽  
Vol 10 (1) ◽  
Author(s):  
Simon Plank ◽  
Francesco Marchese ◽  
Nicola Genzano ◽  
Michael Nolde ◽  
Sandro Martinis
Keyword(s):  
Infrared Imaging ◽  
Rapid Change ◽  
Monitoring Network ◽  
Remote Sensing Data ◽  
Volcanic Island ◽  
Imaging Spectrometer ◽  
Remote Areas ◽  
Moderate Resolution ◽  
Satellite Sensors ◽  
Resolution Imaging

AbstractSatellite-based Earth observation plays a key role for monitoring volcanoes, especially those which are located in remote areas and which very often are not observed by a terrestrial monitoring network. In our study we jointly analyzed data from thermal (Moderate Resolution Imaging Spectrometer MODIS and Visible Infrared Imaging Radiometer Suite VIIRS), optical (Operational Land Imager and Multispectral Instrument) and synthetic aperture radar (SAR) (Sentinel-1 and TerraSAR-X) satellite sensors to investigate the mid-October 2019 surtseyan eruption at Late’iki Volcano, located on the Tonga Volcanic Arc. During the eruption, the remains of an older volcanic island formed in 1995 collapsed and a new volcanic island, called New Late’iki was formed. After the 12 days long lasting eruption, we observed a rapid change of the island’s shape and size, and an erosion of this newly formed volcanic island, which was reclaimed by the ocean two months after the eruption ceased. This fast erosion of New Late’iki Island is in strong contrast to the over 25 years long survival of the volcanic island formed in 1995.

A Machine Learning-based Cloud Detection Algorithm for the Himawari-8 Spectral Image

2021 ◽  
Author(s):  
Chao Liu ◽  
Shu Yang ◽  
Di Di ◽  
Yuanjian Yang ◽  
Chen Zhou ◽  
...  
Keyword(s):  
Machine Learning ◽  
Detection Algorithm ◽  
Cloud Detection ◽  
Spectral Image

scholarly journals Evaluation of WRF Model Forecasts and Their Use for Hydroclimate Monitoring over Southern South America

2016 ◽  
Vol 31 (3) ◽  
pp. 1001-1017 ◽  
Author(s):  
Omar V. Müller ◽  
Miguel A. Lovino ◽  
Ernesto H. Berbery
Keyword(s):  
South America ◽  
Lead Time ◽  
Weather Forecasting ◽  
Wrf Model ◽  
Probability Of Detection ◽  
Southern South America ◽  
Climate Conditions ◽  
Forecast Lead Time ◽  
Highly Correlated ◽  
Precipitation Season

Abstract Weather forecasting and monitoring systems based on regional models are becoming increasingly relevant for decision support in agriculture and water management. This work evaluates the predictive and monitoring capabilities of a system based on WRF Model simulations at 15-km grid spacing over the La Plata basin (LPB) in southern South America, where agriculture and water resources are essential. The model’s skill up to a lead time of 7 days is evaluated with daily precipitation and 2-m temperature in situ observations for the 2-yr period from 1 August 2012 to 31 July 2014. Results show high prediction performance with 7-day lead time throughout the domain and particularly over LPB, where about 70% of rain and no-rain days are correctly predicted. Also, the probability of detection of rain days is above 80% in humid regions. Temperature observations and forecasts are highly correlated (r > 0.80) while mean absolute errors, even at the maximum lead time, remain below 2.7°C for minimum and mean temperatures and below 3.7°C for maximum temperatures. The usefulness of WRF products for hydroclimate monitoring was tested for an unprecedented drought in southern Brazil and for a slightly above normal precipitation season in northeastern Argentina. In both cases the model products reproduce the observed precipitation conditions with consistent impacts on soil moisture, evapotranspiration, and runoff. This evaluation validates the model’s usefulness for forecasting weather up to 1 week in advance and for monitoring climate conditions in real time. The scores suggest that the forecast lead time can be extended into a second week, while bias correction methods can reduce some of the systematic errors.

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.

Sign in / Sign up

Export Citation Format

Share Document