Last time
Last week was Asteroid Day! A day that was introduced to remind us that cosmic threats are real … not only for the Dinosaurs, but also in recent decades like the Tunguska event in 1908 or the Meteor of Chelyabinsk in the year 2013.
Observation surveys, follow up measurements, simulations, and so on are required to catalogue and understand our very cosmic vicinity. Currently, there are no larger objects on a direct collision course with our blue planet. However, observational errors propagate through to data and do not allow us to determine the position of an object with 100 % accuracy. The error-bars depend on the number of observations, observational conditions, total observation time, the distance to the object, its movement and brightness and other factors. It is a multi-dimensional error that can only be faced with more and more and even more data.Luckily, 2020 JX1 is a “good” asteroid and the error-bars during the recent fly-by were quite small (considering cosmic scales).But X, Y and Z coordinates do not help us at all … we need to set ecliptic, equatorial or azimuthal coordinates for our telescopes. Further, we have a solution space of Cartesian coordinates. _How can we translate this solution space to a proper ecliptic coordinate system function? _Let’s find out!
Space Science with Python — Space maps
Multidimensional Kernel Density Estimators
scikit-learnis a great resource for data science and machine learning algorithms. The library covers classifications, dimension reduction, as well as feature engineering and also clustering methods. The sophisticated documentation provides examples for miscellaneous use cases: One example covers the application of Kernel Density Estimators (KDEs) in spherical coordinates. Instead of the Euclidean metric, this example uses the so-called Haversine metric that is applied on the longitude and latitude values in radians:
A KDE for ecliptic longitude and latitude coordinates appears suitable: Let’s go.For this tutorial, we use the already introduced libraries numpy, pandas, tqdm and maptlotlib. We then load the data that were created last time (the file is also part of the GitHub repository).
#data-science #science #space #space-science-with-python #python