STAR Joint Polar Satellite System Algorithms & Data Products website National Oceanographic & Atmospheric Administration website NOAA Center for Satellite Applications and Research website
Suomi NPP Albedo Global map - click to enlarge

Suomi NPP Albedo Global map - click to enlarge

Surface Albedo

Team Lead: Yunyue (Bob) Yu

Background

Land surface albedo (LSA), defined as the ratio between solar radiation reflected by Earth's land surface and solar radiation incident at the surface, is a function of both solar illumination and the reflective properties of land. LSA is an essential variable linking the land surface and the climate system. It is a unique property for studying how land surface changes affect the energy balance and the overall climate system.

Surface albedo is generated routinely from VIIRS as an operational product in the form of an environmental data record (EDR). The surface albedo EDR has global coverage and consists of ocean surface albedo and snow or sea ice surface albedo in addition to LSA. Since the launching of Suomi NPP, the LSA algorithm has gone through two major improvements. Maturity status of the Suomi NPP product generation is defined as beta, provisional and validated versions. The LSA beta and provisional productions were started in June 2013 and April 2014, respectively. The validated V1 version readiness review was approved in December 2015.

Validation results show the errors of the current non-snow LSA retrievals are well smaller than L1RD threshold. The performance of snow LSA is also comparable (slightly better) than the existing albedo product.

Algorithm Science and Data Access

Products and data:

EDR Long Term-Monitoring

Documentation

The VIIRS LSA is retrieved with a direct estimation method, which directly links surface broadband albedo with VIIRS TOA reflectance through statistical modeling. The training data used to establish the regression models are obtained through simulation of physical models.

The original LSA algorithm uses a spectral library as the input for atmospheric radiative transfer simulation. An assumption of Lambertian surface is implied in the process. It has been found that this simplification of surface reflectance will result in retrieval uncertainties and lead to angular dependence in some cases. To address this issue, a new look-up table (LUT) that considers the anisotropy of surface reflectance was established. A band construction method is used to obtain a BRDF database in VIIRS bands from MODIS BRDF data, where VIIRS reflectance is expressed as the linear combination of MODIS spectral reflectances. To obtain a representative training dataset, MODIS BRDF products with the highest quality are collected over various surface types throughout the year. The surface bidirectional reflection factor (BRF) in predefined angular bins is converted from MODIS bands to VIIRS bands for each record in the MODIS BRDF database. The derived VIIRS BRF data will be used as inputs to atmospheric radiative transfer to obtain TOA reflectance in these bins for various atmospheric conditions. Linear regression is then performed to obtain the coefficients that relate LSA to TOA reflectance. The regression coefficients are stored in a LUT, indexed by viewing geometry.

The official JPSS Surface Albedo EDR product can be accessed from CLASS.

Users

Scientists and modelers in many fields may need the LSA data, including NOAA NWS Environmental Modeling Center, USDA Agricultural Research Services, USDA Forest Service, STAR, and NCDC as well as universities and other research institutions throughout the world.

Calibration and Validation

We collected LSA measurements from various ground monitoring networks across the world. However, because VIIRS data are available after 2012, not many stations currently provide measurements of these recent years. After a thorough examination, data at 35 sites, including AmeriFlux, BSRN, GC-Net and SURFRAD, were obtained. High spatial-resolution satellite imagery was then used to evaluate the spatial representativeness of the ground measurements. The analysis of spatial homogeneousness selected 25 sites for data validation. In addition to in situ data, VIIRS albedo was also compared with MODIS data. VIIRS and MODIS both produce high-quality surface albedo over non-snow pixels with root mean square error (RMSE) of 0.024 and 0.032 respectively. The VIIRS retrievals have a small negative bias of 0.006, whereas MODIS underestimate snow-free albedo by 0.026. For all the land cover types other than sparsely vegetated ground, VIIRS data have smaller bias and uncertainties. Compared to snow-free data, estimation of snow albedo generally faces more challenges. Both VIIRS and MODIS data significantly underestimate snow albedo. RMSEs of snow albedo are also much higher, twice greater than snow-free cases. The temporal filter can effectively exclude pixels affected by undetected cloud or cloud shadow. After filtering, bias of VIIRS retrievals is reduced from -0.039 to -0.023 and RMSE is reduced from 0.084 to 0.065.

Ongoing Improvements

Although the VIIRS LSA has passed criteria of the validated V1 release, the team is still working hard on some critical improvements. The planned data improvement includes:

  • Update LUT of regression coefficients for estimating sea ice albedo;
  • Develop a separate LUT for snow pixels and other major land surface types;
  • Implement a temporal filtering to improve both quality and continuity;
  • Propose a framework to generate gridded data set of LSA.

Publications

Wang, D., S. Liang, T. He, and Y. Yu (2013), Direct estimation of land surface albedo from VIIRS data: Algorithm improvement and preliminary validation, J. Geophys. Res. Atmos., 118, 12577-12586, doi: DOI: 10.1002/2013JD020417.