Random Forest Nowcasts

Published: 31 May 2020| Version 3 | DOI: 10.17632/smxkyhtdvj.3
Contributor:
Yiwen Mao

Description

A binary classification model is trained by random forest using data from 41 stations in Norway to predict the precipitation in a given hour. The predictors consist of precipitation predictions from radar nowcasts and numerical weather predictions (AROME), as well as other variables from the numerical weather predictions. The results demonstrate that the random forest model can improve the precipitation predictions by the radar nowcasts and the numerical weather predictions. This dataset provides the predictand (i.e. precipitation observation) and 17 predictors (radar nowcasts and AROME variables) for building a random forest model to process the radar-based nowcasts at 41 stations in Norway for the years 2017 and 2018. The geographic information of the stations is also provided. Explanations of variable names in Data_2017.csv and Data_2018.csv: RR1hr: accumulated precipitation in 1 hour (mm) RR6hr: accumulated precipitation in 6 hours (mm) radar: radar nowcasts obs: observations SLP: sea level pressure AF: area fraction LowC: low type cloud MediumC: medium type cloud HighC: hight type cloud ABLT: Atmospheric boundary layer thickness WSNT: Wind speed normal to the topographic aspect DD(1000-500mb): Average dew point depression from 1000 mb to 500 mb K: K index rand: a sequence of random number Observations are recorded at each weather station. Other meteorological fields are the average of the grid point value closest to the station and the mean grid values weighted by the inverse of the distance from the center of a square of 100km X 100km, centered at the grid point closest to the station. The original data are provided by the Norwegian Meteorological Institute.

Files