This is an old revision of the document!
Camera Images and Ceiling (CIC) database
The purpose of this database is to be able to provide a constant, real-time/up-to-date set of data points will then be used to train Machine Learning (ML) algorithm that can produce it's own real-time cloud ceiling data along with the ability to forecast ceiling heights.
Documentation and information is provided below; related to the collection of Camera Images and Ceiling (CIC) data and creation of a database for a Machine Learning (ML) algorithm which will be trained on the combined data from three types of data sites within a given geographical area available for the public:
- Surface CAM images (by extracting images and features from CAM imagery such as weather cameras, road cameras, etc),
- Ceilometers (by processing data and numbers from METAR, ASOS, AWOS, etc) and
- Satellite (by extracting images and features from Satellite imagery such as cloud top height).
The resulting training based on the publicly available real time data should provide sufficient data points for the ML Algorithm to create it's own real time ceiling data which can be cross referenced with ceilometers.
For example, the UND Dept. of Atmospheric Sciences Skycam images, KGFK METAR, and GOES Satellite imagery can be utilized to determine cloud ceilings in the Grand Forks City area. The ML algorithm will be trained using regularly updated images and data points, enabling it to provide real-time ceiling figures more frequently than the hourly intervals provided by METAR and then produce a set of forecast data points for the location.
Publicly Available Data
Creating the Database
Data and information retrieved from publicly available databases will be grouped under where they fit within the three basic ceiling categories below:
- Ceiling < 200ft AGL
- 200 < Ceiling < 400ft AGL
- Ceiling > 4000ft AGL
Once classified, they will be documented and paired with metadata such as:
- Location
- Date
- Camera Location
- Camera Features
- METAR/ASOS Features
- Ceiling
This categorization and metadata pairing will allow for organized and structured access to the ceiling data within the CIC database.