Open DMI Data — new release of forecast data
The Danish Meteorological Institute (DMI) has released more and more datasets throughout the previous years and has recently added forecast data to the types of data offered. I was happy to receive the above e-mail in late October.
While observation data is very interesting and useful, I find that forecast data can solve different problems. The types of forecast data that were released were:
· Storm surge model data.
· Wave model data.
· Weather model data.
These forecast data are generated by physics-based models running on super computers since it requires complex calculations to complete simulations of physical processes. This, of course, also implies a certain energy consumption and thus CO2 emissions in the generation of this data, which makes it even more important to share so that these calculations can benefit and impact as widely as possible.
Types of forecast data made available by DMI
Let’s have a closer look at the forecast data released. One thing to note that is common to all three types, is that the API only retains results of model runs begun during the previous 48 hours (Request & Response Examples (forecastdata) — DMI Open Data — Confluence (govcloud.dk)). This implies that it is necessary to build your own collection of data from historic model runs if you need to evaluate approaches or train machine learning models on past data. More details on the forecast data models, such as the areas they cover and some more information on how they work, can be found at Forecast Data — DMI Open Data — Confluence (govcloud.dk).
Storm surge model (DKSS) data
This model is run four times a day and forecasts are made for the “three-dimensional state of the sea” five days ahead in terms of water level, sea temperature, salinity, current, and ice thickness and concentration (About DKSS — DMI Open Data — Confluence (govcloud.dk)). The input to the model (forcings) is provided by the numerical weather model HARMONIE from DMI and the global weather model from the European Centre for Medium-Range Weather Forecasts (ECMWF).
Wave model (WAM) data
WAM forecasts various parameters related to waves at an hourly resolution, such as wave height, direction, and wave period. The model is run four times a day, and forecasts are made five and half days into the future. It has the same inputs (forcings) as the DKSS model.
Weather model (HARMONIE) data
The HARMONIE model forecasts many weather-related parameters such as temperature, wind, precipitation, and pressure. There are two different versions of this model, one covering Greenland and one covering Northwestern Europe. The model runs every hour, but only data from deterministic runs every third hour is collected. For the Northwestern Europe model, the forecasts cover 54 hours into the future. For the Greenland version, some runs cover the coming 54 hours and some cover the coming 48 hours (About HARMONIE — DMI Open Data — Confluence (govcloud.dk)). The inputs (forcings) to the HARMONIE model are obtained from ECMWF’s global weather model.
Use cases for open DMI data
I believe there are many more valuable use cases than those I have been able to think of so far. An obvious use of this data is for heating/cooling control in buildings, potentially in an automated way with the forecasted temperature as input.
Precipitation and wind forecasts are likely to be useful for logistics companies planning deliveries a few days in advance since harsher weather conditions often increase travel time. Automated route optimization would likely be improved by taking the weekly weather forecasts into account.
By combining precipitation data with elevation data, e.g. from Dataforsyningen, a system to give automatic alerts to areas at risk of flooding could be set up. This could allow people to take precautions such as moving sensitive electronic equipment or other items to a safe location before a flooding event.
Summary
There is a lot of useful information about how and when the forecast data is generated and how to obtain it on DMI’s Forecast Data webpage at Forecast Data — DMI Open Data — Confluence (govcloud.dk). I have summarized the forecast data types currently available and hope this overview may make it easier to get an idea of what is available, and hence which use cases this data may support.
Note: I first published this as an article on my LinkedIn profile.
