13 May 2020

Meteonorm dataset limitations

The key limitations of Meteonorm's representativeness lie within the use of 1) interpolated and 2) synthetic data. Interpolations rely heavily on surface observation data, which are not equally distributed. Synthetic estimates are useful when high quality direct observational data does not exist. However, given the presence of a global fleet of high resolution weather satellites that directly image cloud cover - this is not the case.

Meteonorm introduction

While Meteonorm has been around for over 30 years and is used widely throughout the solar development community, the provided data has significant limitations in its ability accurately represent the impacts of cloud cover on the solar radiation resource.

Meteonorm draws users with its relatively low price point, however what a user may save in upfront data purchasing costs creates project risk downstream when mismatched revenue models meet with reality. Cloud cover is a highly dynamic variable, and its impacts on solar radiation are quite complex. Look no further than the latest live solar data in your region via our Solar Radiation Map Archive.

europe-solar-radiation-map-archive.png

Meteonorm dataset challenges

The key limitations of the dataset’s representativeness lie within the use of interpolated and synthetic data.

The interpolations rely heavily on surface observation data, which are not equally distributed. This creates difficulties in diagnosing the representativeness of the data for a given site. Namely - is the data you’ve extracted for your PV yield assessment built off of data 1km away, or 100km or more?

GHI IAV.png Pictured above: Solcast satellite based data in Australia vs. Meteonorm’s synthetic approach

Furthermore, surface observation stations are not always appropriately maintained and the quality of the data varies regionally depending on who is responsible for routine maintenance and calibration.

Then we come to the issue of synthetic data creation. While there are many high quality methods for creating synthetic radiation data, their strengths lie in the ability to model solar variability on finer scales or to downscale coarse datasets into local estimates. In essence, synthetic estimates are useful when high quality direct observational data do not exist.

However, given the presence of a global fleet of high resolution weather satellites that directly image cloud cover - this is not the case. Where Meteornorm leverages satellite imagery for their dataset generation, it is generally incorporated only as support information where no meteorological station is available, rather than as the basis for the dataset.

Pictured above: Variable cloud cover conditions over Spain on 11 May 2020. Notice the complexity in cloud type, location and thickness!

Meteonorm alternatives and high quality satellite imagery

Since 2015, 3rd generation weather satellites have begun going into operation. First Himawari-8 over the Asia-Pacific, and then followed by GOES-16 and GOES 17 in the years that followed. These satellites are now sampling the world’s cloud via high resolution, multi-spectral imagery at up to 5-minute intervals with 1km2 resolution.

Simply put, these new hardware are a game changer for the creation of high quality solar resource assessment data. And the formation of our company, Solcast, was directly forged in recognition of this new fleet of satellites, where we’ve re-built satellite to solar irradiance from the ground up (or perhaps, ‘space down?’).

Meteonorm and the Solcast advantage

By applying a new suite of modelling tools, including a proprietary cloud detection and aerosol models (pioneered in part by my PhD research) Solcast has now produced the best performing solar database available. We’ve approached this from the philosophy of enabling solar farm developers and financiers to focus on planning, assessment and building the solar powered future, rather than being stuffed around with low quality datasets based on ‘guesswork’ (e.g. synthetic and interpolated low price data).

How Does Solcast Compare? Our Summary

  Solcast Meteonorm
Free trial with instant access and data download?
Price point $$ $
Download wait time 1-9 minutes 1 minute
Mean Bias Error (GHI) 0.0% N/A
Bias Deviation (GHI) ± 2.0% N/A
90% of sites have bias (GHI) smaller than ± 2.5% N/A
Mean RMSE (GHI) 16.5% N/A
Comprehensive, global, independent validation
Validation Sites 46 N/A
Satellite based estimation
Global Coverage
Resolution of satellite data used 1-2 km 2-8 km
10+ years of satellite data at full temporal resolution
Ignores older, less reliable satellites

Source: Meteonorm Handbook part I: Software & Meteonorm Handbook part II: Theory

Meteonorm - How can we offer such dramatic improvements?

In the rare case where Meteonorm utilises satellite data to support ground measurements, only low resolution data is used (a 2-3 km resolution scan every hour (Europe) or an 8 km scan once every 3 hours (rest of world)). In my opinion, at these timescales, there is much room for biased and inaccurate corrections to be introduced, due to the highly variable nature of cloud formation, motion and opacity (i.e just take another look at that video over Spain above!).

This is a key point of differentiation. It’s no small feat to ingest over 10TB of satellite imagery every month, but the Solcast team has built a modern computational system upon the principles of ‘big data’ to do just that - and utilise every single time-step of satellite data available, at the highest resolutions and across multiple wavelengths. We know it is hard, and so we understand why many of our competitors choose to use older satellites and low resolution sampling.

Our approach to creating our solar irradiance data is well-documented on our website, and so I won’t dive into it here. For more details see inputs and algorithms.

Closing words: Can we directly compare Meteonorm to Solcast?

It is difficult to objectively compare Meteonorm with Solcast due to our highly differentiated approaches to creating solar data. However, Solcast’s model has been independently validated using multiple locations globally. Solcast’s model has been compared with BSRN ground station measurements, available in the public domain.

  Solcast Meteonorm
Mean Bias Error (GHI) 0.0% N/A
Bias Error Standard Deviation (GHI) ±2.0% N/A
90% of sites have bias (GHI) smaller than ± 2.5% N/A
Mean RMSE (GHI) 16.5% N/A

You can find more details on the Solcast dataset on the accuracy and validation page.

Blog Author

Dr. Nick Engerer

Dr. Nick Engerer

Solcast CTO

Nick is an expert in the field of solar radiation and distributed solar PV modelling, and has co-founded Solcast out of a sincere desire to enable others to build the solar-powered future.

Read more about Dr. Engerer’s scientific papers

Follow Nick on Twitter

Connect on LinkedIn

Subscribe to receive our latest blog posts (Solcast "Rapid Updates")!

Enter your email below, to always get the latest.