FOWD: A Free Ocean Wave Dataset for Data Mining and Machine Learning

The occurrence of extreme (rogue) waves in the ocean is for the most part still shrouded in mystery, as the rare nature of these events makes them difficult to analyze with traditional methods. Modern data mining and machine learning methods provide a promising way out, but they typically rely on the availability of massive amounts of well-cleaned data.To facilitate the application of such data-hungry methods to surface ocean waves, we developed FOWD, a freely available wave dataset and processing framework. FOWD describes the conversion of raw observations into a catalogue that maps characteristic sea state parameters to observed wave quantities. Specifically, we employ a running window approach that respects the non-stationary nature of the oceans, and extensive quality control to reduce bias in the resulting dataset.We also supply a reference Python implementation of the FOWD processing toolkit, which we use to process the entire CDIP buoy data catalogue containing over 4 billion waves. In a first experiment, we find that, when the full elevation time series is available, surface elevation kurtosis and maximum wave height are the strongest univariate predictors for rogue wave activity. When just a spectrum is given, crest-trough correlation, spectral bandwidth, and mean period fill this role.

Historical Estimates of Surface Marine Temperatures

Surface temperature documents our changing climate, and the marine record represents one of the longest widely distributed, observation-based estimates. Measurements of near-surface marine air temperature and sea-surface temperature have been recorded on platforms ranging from sailing ships to autonomous drifting buoys. The raw observations show an imprint of differing measurement methods and are sparse in certain periods and regions. This review describes how the real signal of global climate change can be determined from these sparse and noisy observations, including the quantification of measurement method–dependent biases and the reduction of spurious signals. Recent progress has come from analysis of the observations at increasing levels of granularity and from accounting for artifacts in the data that depend on platform types, measurement methods, and environmental conditions. Cutting across these effects are others caused by how the data were recorded, transcribed, and archived. These insights will be integrated into the next generation of global products quantified with validated estimates of uncertainty and the dependencies of its correlation structure. Further analysis of these records using improved data, metadata, and methods will certainly uncover more idiosyncrasies and new ways to improve the record.

Assessment of Variability in Irregularly Sampled Time Series: Applications to Mental Healthcare

Variability is defined as the propensity at which a given signal is likely to change. There are many choices for measuring variability, and it is not generally known which ones offer better properties. This paper compares different variability metrics applied to irregularly (nonuniformly) sampled time series, which have important clinical applications, particularly in mental healthcare. Using both synthetic and real patient data, we identify the most robust and interpretable variability measures out of a set 21 candidates. Some of these candidates are also proposed in this work based on the absolute slopes of the time series. An additional synthetic data experiment shows that when the complete time series is unknown, as it happens with real data, a non-negligible bias that favors normalized and/or metrics based on the raw observations of the series appears. Therefore, only the results of the synthetic experiments, which have access to the full series, should be used to draw conclusions. Accordingly, the median absolute deviation of the absolute value of the successive slopes of the data is the best way of measuring variability for this kind of time series.

Precise astrometry and diameters of asteroids from occultations – a data set of observations and their interpretation

ABSTRACT Occultations of stars by asteroids have been observed since 1961, increasing from a very small number to now over 500 annually. We have created and regularly maintain a growing data set of more than 5000 observed asteroidal occultations. The data set includes the raw observations, astrometry at the 1 mas level based on centre of mass or figure (not illumination), where possible the asteroid’s diameter to 5 km or better, and fits to shape models, the separation and diameters of asteroidal satellites, and double star discoveries with typical separations being in the tens of mas or less. The data set is published at NASA’s Planetary Data System and is regularly updated. We provide here an overview of the data set, discuss the issues associated with determining the astrometry and diameters, and give examples of what can be derived from the data set. We also compare the occultation diameters of asteroids with the diameters measured by the satellites NEOWISE, AKARI AcuA, and IRAS, and show that the best satellite-determined diameter is a combination of the diameters from all three satellites.

GPS/BDS-2/Galileo Precise Point Positioning Ambiguity Resolution Based on the Uncombined Model

In this study, an uncombined precise point positioning (PPP) model was established and was used for estimating fractional cycle bias (FCB) products and for achieving ambiguity resolution (AR), using GPS, BDS-2, and Galileo raw observations. The uncombined PPP model is flexible and efficient for positioning services and generating FCB. The FCBs for GPS, BDS-2, and Galileo were estimated using the uncombined PPP model with observations from the Multi-GNSS Experiment (MGEX) stations. The root mean squares (RMSs) of the float ambiguity a posteriori residuals associated with all of the three GNSS constellations, i.e., GPS, BDS-2, and Galileo, are less than 0.1 cycles for both narrow-lane (NL) and wide-lane (WL) combinations. The standard deviation (STD) of the WL combination FCB series is 0.015, 0.013, and 0.006 cycles for GPS, BDS-2, and Galileo, respectively, and the counterpart for the NL combination FCB series is 0.030 and 0.0184 cycles for GPS and Galileo, respectively. For the BDS-2 NL combination FCB series, the STD of the inclined geosynchronous orbit (IGSO) satellites is 0.0156 cycles, while the value for the medium Earth orbit (MEO) satellites is 0.073 cycles. The AR solutions produced by the uncombined multi-GNSS PPP model were evaluated from the positioning biases and the success fixing rate of ambiguity. The experimental results demonstrate that the growth of the amount of available satellites significantly improves the PPP performance. The three-dimensional (3D) positioning accuracies associated with the PPP ambiguity-fixed solutions for the respective only-GPS, GPS/BDS-2, GPS/Galileo, and GPS/BDS-2/Galileo models are 1.34, 1.19, 1.21, and 1.14 cm, respectively, and more than a 30% improvement is achieved when compared to the results related to the ambiguity-float solutions. Additionally, the convergence time based on the GPS/BDS-2/Galileo observations is only 7.5 min for the ambiguity-fixed solutions, and the results exhibit a 53% improvement in comparison to the ambiguity-float solutions. The values of convergence time based on the only-GPS observations are estimated as 22 and 10.5 min for the ambiguity-float and ambiguity-fixed solutions, respectively. Lastly, the success fixing rate of ambiguity is also dramatically raised for the multi-GNSS PPP AR. For example, the percentage is approximately 99% for the GPS/BDS-2/Galileo solution over a 10 min processing period. In addition, the inter-system bias (ISB) between GPS, BDS-2, and Galileo, which is carefully considered in the uncombined multi-GNSS PPP method, is modeled as a white noise process. The differences of the ISB series between BDS-2 and Galileo indicate that the clock datum bias of the satellite clock offset estimation accounts for the variation of the ISB series.

Efficient multi-GNSS processing based on raw observations from large global station networks

<p>The year 2020 is going to mark the first time of four global navigation satellite systems (i.e., GPS, GLONASS, Galileo, and BeiDou) in full operational capability. Utilizing the various available observation types together in global multi-GNSS processing offers new opportunities, but also poses many challenges. The raw observation approach facilitates the incorporation of any undifferenced and uncombined code and phase observation on any frequency into a combined least squares adjustment. Due to the increased number of observation equations and unknown parameters, using raw observations directly is more computationally demanding than using, for example, ionosphere-free double-differenced observations. This is especially relevant for our contribution to the third reprocessing campaign of the International GNSS Service, where we process observations from up to 800 stations per day to three GNSS constellations at a 30-second sampling. For a single day, this results in more than 200 million raw observations, from which we estimate almost 5 million parameters.</p><p>Processing such a large number of raw observations together is computationally challenging and requires a highly optimized processing chain. In this contribution, we detail the key steps that make such a processing feasible in the context of a distributed computing environment (i.e., large computer clusters). Some of these steps are the efficient setup of observation equations, a suitable normal equation structure, a sophisticated integer ambiguity resolution scheme, automatic outlier downweighting based on variance component estimation, and considerations regarding the estimability of certain parameter groups.</p>

Multi-GNSS PPP instantaneous ambiguity resolution with precise atmospheric corrections augmentation

<p>The performance of precise point positioning (PPP) can be significantly improved with multi-GNSS observations, but it still needs more than ten minutes to obtain positioning results at centimeter-level accuracy. In order to shorten the initialization time and improve the positioning accuracy, we develop a multi-GNSS (GPS + GLONASS + Galileo + BDS) PPP method augmented by precise atmospheric corrections to achieve instantaneous ambiguity resolution (IAR). In the proposed method, regional augmentation corrections including precise atmospheric corrections and satellite uncalibrated phase delays (UPDs) are derived from PPP fixed solutions at reference network and provided to user stations for correcting the dual-frequency raw observations. Then the regional augmentation corrections from nearby reference stations are interpolated on the client through a modified linear combination method (MLCM). With the corrected observations, IAR can be achieved with centimeter-level accuracy. This method is validated experimentally with Hong Kong CORS network, and the results indicate that multi-GNSS fusion can improve the performance in terms of both positioning accuracy and reliability of AR. The percentage of IAR for multi-GNSS solutions is up to 99.7%, while the percentage of GPS-only solutions is 88.7% when the cut-off elevation angle is 10°. The benefit of multi-GNSS fusion is more significant with high cut-off elevation angle. The percentage of IAR can be still above 98.4% for multi-GNSS solutions while the result of GPS-only solutions is below 43.5% when the cut-off elevation angle reaches 30°.  The positioning accuracy of multi-GNSS solutions is improved by 30.0% on the horizontal direction (0.7 cm) and 17.1% on the vertical direction (2.9 cm) compared to GPS-only solutions.</p>

Improving the usability of the MISR L1B2 Georectified Radiance Product (2000–present) in land surface applications

The Multi-angle Imaging SpectroRadiometer (MISR) instrument on NASA's Terra platform has been acquiring global measurements of the spectro-directional reflectance of the Earth since 24 February 2000 and is still operational as of this writing. The primary radiometric data product generated by this instrument is known as the Level 1B2 Georectified Radiance Product (GRP): it contains the 36 radiometric measurements acquired by the instrument's 9 cameras, each observing the planet in 4 spectral bands. The product version described here is projected on a digital elevation model and is available from the NASA Langley Atmospheric Science Data Center (ASDC) (https://doi.org/10.5067/Terra/MISR/MI1B2T_L1.003 (Jovanovic et al., 1999). The MISR instrument is highly reliable. Nevertheless, its on-board computer occasionally becomes overwhelmed by the amount of raw observations coming from the cameras' focal planes, especially when switching into or out of Local Mode acquisitions that are often requested in conjunction with field campaigns. Whenever this occurs, one or more lines of data are dropped while the computer resets and readies itself for accepting new data. Although this type of data loss is minuscule compared to the total amount of measurements acquired, and is marginal for atmospheric studies dealing with large areas and long periods of time, this outcome can be crippling for land surface studies that focus on the detailed analysis of particular scenes at specific times. This paper describes the problem, reports on the prevalence of missing data, proposes a practical solution to optimally estimate the values of the missing data and provides evidence of the performance of the algorithm through specific examples in South Africa. The software to process MISR L1B2 GRP data products as described here is openly available to the community from the GitHub web site https://github.com/mmverstraete or https://doi.org/10.5281/zenodo.3519989.