This has been mentioned in other posts (e.g. here), but for the Netherlands HARMONIE 2.5km is the best short-range prediction model by a long shot. It is for example the only model that accurately predicts sea breeze, which is very relevant in this coastal country. HARMONIE-40 is the operational short-range model for the Dutch Meteorological Institute (KNMI). Note that ECMWF is decent and also used by the KNMI but only for medium-range forecasts.

I read somewhere else on this forum that 'more models will only confuse', but for the Netherlands HARMONIE 2.5km and ECMWF are really the only models you'll need.

HARMONIE can be downloaded for free from the KNMI datashare, you'll find more information here.

Also, I think replies with maps like this create confusion about the usefulness of regional short range models. So to clarify a bit:

1. AROME is really good, for France.
2. Some might understand that all the models on the map are basically the same with minor tweaks here and there, and that one of them (AROME) isn't quite different from another (HARMONIE-AROME). This is untrue. The map is historical and orginisational rather than consequential. For example, HARMONIE-AROME and AROME are very different models, see e.g. Bengtsson et al.. Importantly, it's also the boundary conditions and initialisation that may (and do) make forcasts vastly different between these models.

Again, I do understand that adding HARMONIE to Windy is a choice between effort and number of users from the Netherlands or any other country that uses HARMONIE. But I think it would be super cool if we could get the discussion beyond 'some other model tries to do this as well' and talk about actual quality.

Best wishes,

Bastiaan

And one other addition: there are quite a lot of countries whose short-range forecasts are based on HARMONIE: "... Denmark, Estonia, Finland, Iceland, Ireland, Lithuania, the Netherlands, Norway, Spain, and Sweden." (Bengtsson et al., 2017).

The challenge will be to get those predictions from their respective national institutes. At least for the Netherlands they are freely available (see link in my first post).

I think it would be worth the effort to include HARMONIE-40 2.5km in Windy, if it works out for the Netherlands maybe those other countries might follow relatively easily.

Thanks @idefix37. I am aware that other models predict sea breeze, but for the Dutch coast they are not accurate enough. Apart from accurately predicting sea breeze HARMONIE in general is really the go-to short-range model for the Netherlands.

EDIT: to elaborate a bit, HARMONIE is specifically tailored to the Netherlands by the (only) meteorological institute and it really shows. As said it is the official operational short-range model and there's no discussion here that other models are in a different ballpark, especially for wind predictions.
EDIT2: just historically it's nice to know that HARMONIE is based on AROME... :)

Hi @tjebbe,

The link to the KNMI data share you provided is to the 5.5km Europe version of HARMONIE, not to the 2.5km Dutch version. The 2.5km is the correct one, and can be found here.

Adding HARMONIE would change the game for Windy in the Netherlands from medium range to short range. I have elaborated a bit in another post, cross-link for reference here.

P.S. I've answered this in a couple other posts as well, so I'm sorry for the redundancy, but just to be sure that the correct model is implemented!

Bastiaan

Hi @idefix37,

There is a reason national institutes optimise their models; it greatly improves regional forecasts. I honestly do not believe it is useful to point out that other models can do short range as well if in a practical sense no one believes their outcome. In the Netherlands AROME is not a reasonable short range prediction model, and HARMONIE is. The added benefit is that it is available for free and that many other countries use it so there might be portability afterwards (see my other posts about this). The same might be true for German models; it might not be as bleak as the map suggests for many countries.

If Windy wants to do short range, at least for the Netherlands the way to go is HARMONIE. I understand that implementing it is a cost-benefit analysis. But let's not spin the discussion to 'AROME is good enough'.

Bastiaan

Also, the discussion here about HARMONIE vs AROME is relevant:

https://community.windy.com/topic/12545/dutch-harmonie-weather-model

Hi all,

ECMWF is really not meant to accurately predict thunderstorms the next day. It's a medium range forecast model that's best used for 2+ days ahead, and is very useful for initialization of short-range high resolution models that are much better for local predictions like thunderstorms. Currently Windy does not include those models except for France (AROME). The AROME model does include regions outside France like Belgium but those are meant to set the boundary conditions for France and not to accurately predict weather in those regions.

I'm mentioning this because I think if you're using Windy predictions for thunderstorms you should know about this, and because it would be very cool if those short-range predictions would be included in Windy. It is currently the main missing feature as far as weather prediction goes (I'm trying to get coverage in my own country via other posts), and most of them are available for free from their respective meteorological institutes.

See you on the water!

Also, I think replies with maps like this create confusion about the usefulness of regional short range models. So to clarify a bit:

1. AROME is really good, for France.
2. Some might understand that all the models on the map are basically the same with minor tweaks here and there, and that one of them (AROME) isn't quite different from another (HARMONIE-AROME). This is untrue. The map is historical and orginisational rather than consequential. For example, HARMONIE-AROME and AROME are very different models, see e.g. Bengtsson et al.. Importantly, it's also the boundary conditions and initialisation that may (and do) make forcasts vastly different between these models.

Again, I do understand that adding HARMONIE to Windy is a choice between effort and number of users from the Netherlands or any other country that uses HARMONIE. But I think it would be super cool if we could get the discussion beyond 'some other model tries to do this as well' and talk about actual quality.

Best wishes,

Bastiaan

Hi @idefix37,

There is a reason national institutes optimise their models; it greatly improves regional forecasts. I honestly do not believe it is useful to point out that other models can do short range as well if in a practical sense no one believes their outcome. In the Netherlands AROME is not a reasonable short range prediction model, and HARMONIE is. The added benefit is that it is available for free and that many other countries use it so there might be portability afterwards (see my other posts about this). The same might be true for German models; it might not be as bleak as the map suggests for many countries.

If Windy wants to do short range, at least for the Netherlands the way to go is HARMONIE. I understand that implementing it is a cost-benefit analysis. But let's not spin the discussion to 'AROME is good enough'.

Bastiaan

Hi @tjebbe,

The link to the KNMI data share you provided is to the 5.5km Europe version of HARMONIE, not to the 2.5km Dutch version. The 2.5km is the correct one, and can be found here.

Adding HARMONIE would change the game for Windy in the Netherlands from medium range to short range. I have elaborated a bit in another post, cross-link for reference here.

P.S. I've answered this in a couple other posts as well, so I'm sorry for the redundancy, but just to be sure that the correct model is implemented!

Bastiaan

Hi @tjebbe and @ivo,

The link to the KNMI data share you provided is to the 5.5km Europe version of HARMONIE, not to the 2.5km Dutch version. The 2.5km is the correct one, and can be found here https://data.knmi.nl/datasets/harmonie_arome_cy40_p1/0.2?q=harmonie.

Bastiaan

@TheRadarGuy said in FMI Hirlam Weather Model on Amazon AWS S3:

find any free source for HARMONIE models

I have created a new topic here for HARMONIE, including the location of the Dutch data share.

And one other addition: there are quite a lot of countries whose short-range forecasts are based on HARMONIE: "... Denmark, Estonia, Finland, Iceland, Ireland, Lithuania, the Netherlands, Norway, Spain, and Sweden." (Bengtsson et al., 2017).

The challenge will be to get those predictions from their respective national institutes. At least for the Netherlands they are freely available (see link in my first post).

I think it would be worth the effort to include HARMONIE-40 2.5km in Windy, if it works out for the Netherlands maybe those other countries might follow relatively easily.

Thanks @idefix37. I am aware that other models predict sea breeze, but for the Dutch coast they are not accurate enough. Apart from accurately predicting sea breeze HARMONIE in general is really the go-to short-range model for the Netherlands.

EDIT: to elaborate a bit, HARMONIE is specifically tailored to the Netherlands by the (only) meteorological institute and it really shows. As said it is the official operational short-range model and there's no discussion here that other models are in a different ballpark, especially for wind predictions.
EDIT2: just historically it's nice to know that HARMONIE is based on AROME... :)

This has been mentioned in other posts (e.g. here), but for the Netherlands HARMONIE 2.5km is the best short-range prediction model by a long shot. It is for example the only model that accurately predicts sea breeze, which is very relevant in this coastal country. HARMONIE-40 is the operational short-range model for the Dutch Meteorological Institute (KNMI). Note that ECMWF is decent and also used by the KNMI but only for medium-range forecasts.

I read somewhere else on this forum that 'more models will only confuse', but for the Netherlands HARMONIE 2.5km and ECMWF are really the only models you'll need.

HARMONIE can be downloaded for free from the KNMI datashare, you'll find more information here.