Validation Sources¶

In addition to the core data products, POWER curates and provides several point-based in-situ validation data sets for the express purpose of validating POWER data. These are the same data sets that the POWER team use to validate a subset of the core data sets that are detailed within the methodology pages. They are not made available to validate derived parameters.

They are available for the public in the Data Access Viewer and through an API endpoint for specific parameters.

The sources below have been vetted for quality of the data and availability for global and temporal coverage.

Source	Abbreviation	Temporal	Number of Locations	Start Year	POWER Parameters
NOAA NCEI Global Summary of the Day	NOAA NCEI GSOD	Daily	2599	1981	T2M, T2MDEW, WS10M_MAX
NOAA NCEI Integrated Surface Database (Lite)	NOAA NCEI ISD Lite	Hourly	2599	1981	T2M, T2MDEW, PS, PRECTOTCORR, WS10M, WD10M
Baseline Solar Radiation Network	BSRN	Hourly, Daily	78	1992	ALLSKY_SFC_LW_DWN, ALLSKY_SFC_SW_DWN, ALLSKY_SFC_SW_DNI, ALLSKY_SFC_SW_DIFF
Aerosol Robotic Network	AERONET	Hourly, Daily	1825	1993	AOD_84, AOD_55, PW

NOAA NCEI GSOD¶

Data sourcing and processing¶

As the name suggests, NOAA NCEI summarizes weather station observation data to daily averages. For details regarding how that was done, please consult the GSOD documentation. The POWER Validation API reads the daily data values directly from the source, which is publicly available through the Registry of Open Data on AWS. The surface site data is converted to the correct units and compared against the nearest MERRA-2 data point.

Migration to new surface site data¶

The Global Summary of the Day has been decommissioned since 29 August 2025. In the future, POWER will migrate to using the NCEI Synoptic Summary of the Day (SSODv2), which is NCEI's replacement daily observation summary. See official documentation here.

NOAA ISD (Lite)¶

Data sourcing¶

Like GSOD, the most accurate hourly subset ISD is available through the AWS Open Data Registry, but POWER accesses values from a lighter data set ISD Lite. The data values are mostly the same and considered suitable for timeseries analysis, which is exactly what POWER tools enable. Several of the fields that are available in the true NOAA subset like accuracy flags, etc. are absent in the Lite version.

Processing¶

For temperature, dew point, wind speed, and precipitation, the actual data values are read directly from ISD lite, divided by 10, and converted to the correct units. Wind direction remains the same without dividing by 10. Air Pressure requires on-the-fly reprocessing.

Pressure conversion from sea level pressure to station level pressure¶

The only available ISD pressure is sea level pressure (SLP). For POWER Surface Pressure validation, the ISD SLP is converted to station level pressure, accounting for surface elevation with the following equation.

$\begin{align} \text{PS} = \frac{\text{SLP}}{e^{\frac{\text{Elev} \times g}{R \times \text{Temp}}}} \end{align}$

$\begin{align} Where: \\ Elev: & \text{ Elevation of the reporting surface site, in meters. } \\ Temp: & \text{ Temperature of reporting surface site, in Kelvin. } \\ PS: & \text{ Surface site pressure, in desired output unit(hPa, kPa). } \\ SLP: & \text{ Surface site sea level pressure. } \\ R: & \text{ Gas constant for dry air, 287.1 JK⁻¹kg⁻¹ } \\ g: & \text{ Gravitational constant, 9.81 } \text{m} \text{s}^{-2} \\ \end{align}$

Precipitation¶

ISD Lite captures "trace" precipitation, which is interpreted as 1/10 millimeter per hour for DAV and API usage.

Migration to new surface site data¶

Like the Global Summary of the Day, the Integrated Surface Database with hourly averages has not been updated since 29 August 2025 and has been superseded by the Global Historical Climatology Network hourly (GHCNh). Like SSOD, POWER will migrate hourly surface observations to GHCNh.

NOAA sites subset

For both NOAA sources, the 2,599 surface sites were identified based on the overall long-term timeseries consistency. All sites have some data gaps but are generally good for POWER validation. This is only a small subset of all NOAA sites.

BSRN¶

Data Sourcing¶

Shortwave irradiance validation data come from the Baseline Surface Radiation Budget. The POWER team manually pulls data from 78 independent BSRN sites, some of which have been tracking solar irradiance since 1992. Every site/month data set is regorously reviewed for quality before being published and assigned a DOI. Only data with published DOIs are compiled by the POWER team for POWER validation analysis.

Processing¶

Raw BSRN data sets typically have readings at 1, 2, 3, or 5 minute intervals. Reading directly from the source DOIs to validate a timeseries of POWER data would be computationally intensive with a slow turnaround time for users. Instead the data is precomputed into hourly and daily averages.

For hourly and daily POWER validation, the minute-based BSRN data must be aggregated appropriately. The following process accounts for missing data.

BSRN Temporal Processing Diagram

15-minute averages require at least one valid minute value from the raw data. 1-hour averages require valid values for all 4 15-minute periods within that hour. Daily averages require 5 valid daytime hourly values per day, regardless of LST or UTC designation. Quality flags in the source data are accounted for. Additional there are other contitions that must be met related to which latitude region the site is located within and the length of the gap between valid hourly data values within a day.

There are also scientific processes for aggregating the data into 3-hour and monthly averages, which are important for the development of the POWER team validations in the methodology docs, but are not made available to users for hourly and daily validation.

You can read more about the process of aggregating solar data from minute to hourly and daily in this paper on validating GEWEX SRB with BSRN.

BSRN irradiance data are in W/m^2. Users can validate POWER irradiance parameters in the same unit.

Data Latency¶

The latency for individual sites varies, but the surface site data is never available near-real time due to the process of publishing monthly data sets with a DOI. Adding to the consistency is the decentralized nature of BSRN, with each site publishing data sets independently and on different timelines. The POWER team reviews all available BSRN data on an annual basis.

Migration to new surface site data¶

There is no plan to migrate to a new validation data set for solar irradiance as updates to BSRN are ongoing.

AERONET¶

Data Sourcing¶

Aerosol validation data for POWER comes from the Aerosol Robotic Network from NASA's Goddard Space Flight Center. As with BSRN, the AERONET data is pulled and curated significanly before it is made available for user validation due to also being minute-based. POWER uses Level 2.0 AERONET data.

Processing¶

Precipitable Water are copied simply aggregated before being converted to the correct units on the fly. The Aerosol Optical Depth (AOD) parameters, on the other hand, need additional processing to be suitable for validation.

Aerosol Optical Depth¶

POWER AOD parameters are for 550 and 840 nanometers, but AERONET does not track those exact values. Instead, POWER interpolates the 550 and 840 amounts from at least two of AERONET's valid readings. Typically the 500 or 870 values are reliable to inpterpolate from, but in their absense the 550 and 870 values are interpolated from other AERONET values. If the data for a particular reading cannot be interpolated, POWER recognizes no valid data.

Temporal Aggregation¶

Hourly and daily AERONET in POWER are sourced differently, where daily averages are already available from the AERONET project, hourly must be aggregated from raw minute-based data. Daily AOD 550 and 840 are interpolated as described above and do not need temporal aggregation.

Raw AERONET data is aggregated to hourly values for POWER users in a similar method as how BSRN is handled. The minute-based data is averaged to 15-minute and require at least 1 valid value. The hourly values are averaged from 4 valid 15-minute values or else no value can be assigned. Daily values are a result of the average of at least 5 hourly values. Since POWER does not have sub-hourly core data, the minute-based and 15-minute averages are discarded and only the hourly values are kept.

Once it is curated and saved in POWER's storage solution, the data is ready for users to validate with it in the DAV or API.

Data Latency¶

While AERONET updates their data sets daily, POWER accesses and processes the complete data set on a monthly basis.

Migration to new surface site data¶

There is no plan to migrate to a new validation data set for aerosol data as updates to AERONET are ongoing.

Significant AERONET data gaps

Due to the nature of the data, AERONET has noticeably more gaps than other validation sources. One key reason for this is that particulate matter and other aerosols cannot be detected accurately except in clear sky conditions.

Limited BSRN and AERONET temporal availability

One caveat to consider is that not all sites have data for the complete expected timeseries between and the present time, so timeseries data availability is specific to each site. This is different from having data gaps. For example, a site may have consistent data from 2016–2018 and cannot help for analysis before or after that range.