Rapid advances in various spectrally-selective technologies, such as PV and multi-coated glazings, have put a lot of pressure on the solar radiation community to provide demanding measurements of the solar spectrum, as well as appropriate models to predict its variations. Only a very few institutions, such as NREL in Colorado, are actually measuring spectral irradiance on a permanent basis. Most other institutions measuring the solar spectrum do it for experimental reasons, and therefore on a sporadic basis. Indeed, obtaining good-quality spectra is not easy because considerable resources are required: costly instrumentation, frequent and expensive recalibrations, and highly skilled personnel. All these conditions greatly limit the availability of the reference spectral irradiance databanks that are necessary to serve the development of spectrally-selective technologies.

This lack of measurements can be compensated for in large part by the use of appropriate radiative transfer models. A variety of such models have been developed for the needs of atmospheric scientists and of the remote sensing and climate change communities. Examples of such models include MODTRAN, SBDART, libRadtran, and 6S. There are many reasons, however, why such models are not convenient for the engineering applications envisioned here. One essential drawback of these models, besides their complexity and considerable execution time, is that they do not address the essential case of spectral irradiance incident on tilted surfaces.

Since its development began many years ago, the SMARTS model's conceptual idea has always been to offer fast and accurate predictions of spectral irradiance on any tilted surface without the difficulties and limitations associated with the atmospheric models mentioned above. The SMARTS model is now used by thousands of scientists worldwide, for a large variety of applications. This is actually made possible by the model's versatility and accuracy, which have been discussed in various scientific papers [1], [2]. Although the model currently accommodates the case of cloudless skies only, it is hoped that funding will become eventually available to expand its scope through the development of an all-sky version, which could be used to simulate the yearly performance of spectrally-selective devices, for instance.

SMARTS Versions

Three versions of SMARTS are currently available: 2.9.2, 2.9.5, and 2.9.9.

Version 2.9.2 of the model is freely available in two different packages (PC and Mac-Classic) from

A newer and much expanded version, 2.9.5, is also available for three platforms (Linux, Mac-OSX and PC-Windows). It can be downloaded here. Among other refinements and improvements, this version has the option to use an ISO-sanctioned extraterrestrial spectrum, which can de obtained here. Windows-only graphic interfaces are also available for both version 2.9.2 and 2.9.5, making them easier to use by novice users.

A python wrapper exists for version 2.9.5. It has been developed by Silvana Ayala of NREL, and is available here.

The User's Manual in pdf format is included in each original distribution package.

The latest version, 2.9.9, is now available in two different packages (PC and Mac-Intel). Among other improvements, the code is now faster, many more spectral albedo data files have been added, and a new default extraterrestrial spectrum has been introduced, based on recent developments [3], [4].

As before, version 2.9.9 is intended for education or pure civilian research, and is free for use for these specific applications ONLY. The License Agreement is available here. If you fully agree to the terms and conditions of this License, you can request v2.9.9 (specify Mac or Windows version) by contacting us directly through the contact page).


Sample outputs produced by SMARTS, compared with actual spectroradiometric measurements from high-performance instruments, are shown in the figures below.

Typical Applications

SMARTS v2.9.2 has been used to produce Reference terrestrial spectra for standardization purposes, including ASTM Standards G173, G177 and G197, and IEC 60904-3.

Many applications involve the prediction of the components (direct, diffuse, or global) of the solar spectrum under ideal or realistic conditions. In the latter case, the accurracy of the modeled spectra is essentially constrained by that of the input data (e.g., aerosol optical depth, precipitable water, or ozone amount). Such realistic predictions can be compared to actual measurements, as shown in the figures above.

Whenever such spectral measurements are available, it is also possible to use SMARTS in a reverse way, to evaluate the main atmospheric variables (such as aerosol optical depth at many wavelengths, or precipitable water) at the time of measurement.

Finally, SMARTS can also be used to predict the broadband irradiance components incident on horizontal or tilted surfaces when high-accuracy results are necessary [5], [6], or when estimates of the circumsolar impact are desirable. For instance, the figure below shows the annual-mean clear-sky broadband global horizontal irradiance (GHI) and direct normal irradiance (DNI) over the world at ≈1° spatial resolution. In related developments, SMARTS has been used to derive some broadband radiation models for direct irradiance (LWMT, LWMT2, REST) or direct, diffuse and global irradiance (REST2), as well as models for illuminance, luminous efficacy, and photosynthetically active radiation (PAR). See the Publications page.

More Technical Details

SMARTS computes clear sky spectral irradiances (direct beam, circumsolar, hemispherical diffuse and total on a prescribed receiver plane -- tilted or horizontal) for one set of atmospheric conditions (user specified, or selected from 10 standard atmospheres); and for one to many points in time or solar geometries. The algorithms were developed to match the output from the MODTRAN sophisticated band models within 2%. The algorithms are implemented in compiled FORTRAN code for various platforms. The algorithms are used in conjunction with files for atmospheric absorption of atmospheric components and spectral albedo functions. The spectral resolution is 0.5 nm from 280 nm to 400 nm, 1 nm from 400 nm to 1750 nm, and 10 nm from 1750 nm to 4000 nm.

The user constructs a text file of between 20 and 30 lines of simple text and numbers specifying input conditions and various spectral output parameters. The user may specify field-of-view angles for direct beam computations, which then include the circumsolar radiance component. Gaussian or triangular smoothing functions with user-defined bandwidth may be specified to compare model results with measurements made with the specified pass band. The user may specify only Ultraviolet (280 to 400 nm) computations for erythemal dose, UV index, etc. Photometric (luminous flux) computations, weighted by a selected photopic response curve, may also be specified. Output is spreadsheet-compatible ASCII text files and header information with prescribed conditions.

[1] C.A. Gueymard, Interdisciplinary applications of a versatile spectral solar irradiance model: A review. Energy, vol. 30, 1551-1576 (2005).
[2] C.A. Gueymard, Prediction and validation of cloudless shortwave solar spectra incident on horizontal, tilted, or tracking surfaces. Solar Energy, vol. 82, 260-271 (2008).
[3] C.A. Gueymard, Revised composite extraterrestrial spectrum based on recent solar irradiance observations. Solar Energy, vol. 169, 434-440 (2018).
[4] C.A. Gueymard, A reevaluation of the solar constant based on a 42-year total solar irradiance time series and a reconciliation of spaceborne observations. Solar Energy, vol. 168, 2–9. (2018).
[5] J.A. Ruiz-Arias and C.A. Gueymard, A multi-model benchmarking of direct and global clear-sky solar irradiance predictions at arid sites using a reference physical radiative transfer model. Solar Energy, vol. 171, 447–465 (2018).
[6] J.A. Ruiz-Arias, C.A. Gueymard and T. Cebecauer, Direct Normal Irradiance Modeling: Evaluating the Impact on Accuracy of Worldwide Gridded Aerosol Databases. SolarPACES Conf., Casablanca, Morocco (2018).