Context

Our aim is to improve the accuracy of the meteorological model with Machine Learning. To do so we need a database that contains input variables (meteorological model results) and output data (actual data from a meteorological station). Dependent variables are outputs of the meteorological model. Independent variables are measured by the meteorological station. The trained Machine Learning will take the variables from the meteorological model and forecast a meteorological variable (wind variable in our case). The meteorological model is a WRF model maintained by Meteogalicia, a public meteorological service from Galicia (Spain). The model has a resolution of 4 Km. We get the nearest points outputs provided by the model from the station. This dataset is focused on two meteorological stations. Coron at latitude: 42.5801 N and longitude: 8.80471 W. Cortegada at latitude: 42.626 N and longitude: 8.784 W. The stations are in Ria Arousa (Spain). We try to improve the forecast of wind variables. At the figure below you can see the meteorological model points and the stations (Cortegada and Coron) location

Content
The dataset contains:

1. Results of a meteorological model
2. Real meteorological data from a meteorological station
3. Artificial intelligence algorithms and algorithm descriptions and performances.

The main menu contains files with two kinds of formats:

1.-Format stationname_all.csv: wind values observed every ten minutes at the "station name" and meteorological model results at the nearest point from the station. More details in the description of each file.

2.-Format stationnameD1res4K.csv: wind variables at the "station name" and meteorological model results at the four closest points from the station. Points label are NE, SE, SW, and NW. D1 refers to model forecasts from 24 hours to 48 hours forecast. res4K means the spatial resolution of the model: 4 Km.

The algorithm menu contains a submenu for every station. The submenu includes two kinds of files:
1.- Files with extension ".h5":
A binary file provides the trained Artificial Intelligence algorithm. The file name represents the station name, the meteorological model range (D1 as described before), the model spatial resolution (res4K 4Km ), algorithm kind (NN neural networks, or DT decision tree) b for Beaufort and q for quintiles and dir for wind direction. Notebooks with operational names use these files to get the output. variable

1. Files with extension ".pptx":
   Description and performance of each algorithm provided as a PowerPoint file. The PowerPoint file compares the performance of every meteorological model point near the station and the performance of the AI algorithm. And, of course, the performance of the AI algorithm is better than the meteorological model at every near point

Acknowledgment
The meteorological model that I used to build the database is maintained by Meteogalicia (a public meteorological service). Meteogalicia supplied a WRF model applied in the Galicia region (Northwest Spain). They use a THREDDS (Thematic Realtime Environmental Distributed Data Service), a connectivity tool between scientific data providers and end-users. We can get a historical model from Meteogalicia WRF archives.
The actual meteorological data are obtained from the meteorological stations at Coron and Cortegada. Coron and Cortegada meteorological stations are provided by Meteogalicia

Inspiration
I aim to find new algorithms that can improve meteorological model accuracy. We can set up supervised or unsupervised Machine learning algorithms to achieve our goal. The dataset would be a good start point