NSRDB Comparison

Solcast's historical solar irradiance database was released in 2019, with a mission to provide highly accurate, validated data to the solar development community.

Using a proprietary cloud detection and identification model, utilising the new fleet of third-generation geostationary satellites and refined over a four-year period, we’ve been able to produce the best performing solar database available. We've always thought developers should focus on planning and assessment, with fast and easy access to high quality, bankable data.

The US National Renewable Energy Agency (NREL) maintains the National Solar Radiation Database (NSRDB). The database is free-to-access but has limited coverage and higher uncertainty when compared with Solcast data.

Summary

Source: Evaluation of the National Solar Radiation Database (NSRDB Version 2): 1998–2015

Accuracy & Validation

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. You can request a copy of our validation data, see accuracy and validation.

The NSRDB has been independently validated in the USA, but no public information exists validating the data for South Asia or Central America. The NSRDB has been validated at 9 sites, a sample size too small to pull representative statistics from.

As there is such a low number of validation sites, and a portion of the database has no validation document published, it is difficult to objectively evaluate the two data sets. As the NSRDB is not global, it is difficult to establish the consistency of the methodology used across different regions. Further, the NSRDB has a larger resolution than the Solcast database. That being said, we have still presented a summary of Solcast's and NSRDB's hourly mean GHI validation results below.

Solcast v NSRDB Bias Distribution

Understanding bias helps to understand a possible error of the long-term estimate. Bias variability and RMSE are important for estimating the accuracy of energy simulations and operational calculations. The Solcast model is symmetrically centred with an average bias error of 0.0%. The NSRDB has a positive bias of 0.3%.

The Solcast data shows significantly less variability than NSRDB. A lower bias deviation of ±2.0% v ±2.7% means Solcast data is represented by a narrower probability distribution.

RMSE, a measure of the differences between the values predicted by the Solcast model and the values actually observed, shows Solcast outperforms the NSRDB. With much less variability in the dataset, developers can trust the accuracy of their energy simulations and operational calculations.

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

Model Inputs

The Solcast methodology is semi-empirical and satellite-derived. We begin using validated, published models that offer excellent performance to build clear sky models. We combine this with the ‘secret sauce', our in-house methods for detecting and tracking clouds. This lets us model the amount and type of solar radiation reaching any particular 1-2 km² grid cell. For more details see inputs and algorithms.

The NSRDB methodology is similar. It is semi-empirical and satellite-derived, relying on validated, published models to build a clear sky model, and for cloud detection. The NSRDB cloud model has a lower resolution than the Solcast model, based around a 4 km² grid.

Solcast only uses data from recent generation geostationary meteorological satellites (GMS). For this reason, we do not use data prior to 2007. Prior to this time, we find that too many of the GMS satellites have significant data quality problems including geolocation, poor temporal sampling frequency, and unreliable, bad or drifting data signatures. This ensures our entire database is based on data of the highest possible resolution, whilst still containing 12+ years of data history from which to understand the interannual variability. Users can be sure our accuracy and validation statistics apply to all our data, not just the most recent half of it.

We don't believe in using data from older satellites. With longer refresh rates and larger spatial resolutions, it's just not possible to produce bankable solar radiation assessments with the certainty demanded by the solar community. We're firm of the opinion that bad data in, bad data out.

Solargis has an extensive database, compiled using high-resolution new satellites, and lower-resolution, pre-21st century satellites.

Data Availability

Solcast offers near-global coverage (latitudes higher than 60-70 degrees are not effectively covered by geostationary satellite images). The NSRDB is limited to North America, Central America, and South Asia.

NSRDB Coverage

Solcast Coverage

Solcast offers data from as recent as the last week. NSRDB data is available only until 2014. Without the provision of recent data, developers are not able to make one-to-one comparisons with ground observations to better understand the accuracy of data. Further, it is not possible to improve the accuracy of a long-term estimate by quantifying systematic deviations between modeled data and high-quality ground observations or use regular updates of recent data to monitor the operational performance of PV power plants. Data parameters available are shown below.

Data parameters available

Solcast strives to make our data easy to use, immediately. Our data is available within 1 to 9 minutes of ordering. Our data is available in the following ready-to-import formats.

Data files available

Pricing

Being a free-to-access database, NSRDB can appear to be cheaper from the first viewpoint. However, the proven bankability of Solcast's database offers better value by lowering risk for developers, investors, and operators of solar power plants. When decisions on development or operation of high-value solar assets are to be made, higher risk associated with the use of high-uncertainty data can be very expensive. View Solcast prices