Snow height measurement has changed considerably in the last few years from the classic measurement with an observer to the automatic mode, where more and more stations use laser based instruments to measure snow height. However, the disadvantage of these instruments, is that they do not directly measure the height of newly fallen snow. So far, this problem has been replaced by measuring the difference in total snow height between two consecutive days, which is not very accurate and it is only an approximation. The work was done at Slovenian Environment Agency (ARSO) where we prepared a method that gives us a better estimate of new snow that takes into account meteorological parameters. The method is based on machine learning and was developed in the programming language Python using the sklearn library, where we used multiple polynomial regression. The result of the method has reduced the mean absolute error of the difference method by about 20%. The method works best for solid form of precipitation. The estimated height of new snow can be used for locations where we do not have the classic measurement of newly fallen snow or for data control where the height is measured.
|
