Poster Session P1.15 Validation of Clouds & The Earths Radiant Energy System (CERES) Synoptic (SYN) data product

Monday, 28 June 2010
Exhibit Hall (DoubleTree by Hilton Portland)
David A. Rutan, SSAI, Hampton, VA; and D. R. Doelling, T. P. Charlock, F. G. Rose, and L. T. C. Nguyen

Handout (832.8 kB)

The Clouds and the Earths Radiant Energy System (CERES) recently released the SYN (for "synoptic") global, gridded, 3-hourly data product that includes observed top-of-atmosphere (TOA) and calculated flux profiles for both the Terra and Aqua satellite data streams. The CERES team uses MODIS pixels collocated within CERES footprints to derive cloud properties (independently of the MODIS team cloud products) for input into the radiation transfer model when CERES observations are available. To execute the model for hours where CERES data are not available, the CERES time/interpolation/spatial averaging group has amassed geostationary observations where available and produced cloud properties and estimates of TOA flux approximately every three hours around the globe across the time frame of CERES processing. To fill remaining intervening hours, where neither CERES nor GOES data are available, cloud properties and TOA radiation are interpolated across time. The radiation transfer model used is a highly modified version of the Fu & Liou radiation transfer (RT) code and calculations are done hourly though archived on the SYN every three hours at five atmospheric levels for each global 1-degree by 1-degree grid cell. For validation purposes, CERES observations supply a 'truth' against which the model can be compared at the TOA at overpass times and broadband surface flux observations provide the same at the Earth's surface throughout the entire day. We also consider comparisons of the model to TOA fluxes estimated from the geostationary data and the interpolated TOA fluxes.

Validation of the SYN calculations relies on systematic comparison with broadband radiometric observations: by CERES instruments at the TOA (sufficient for roughly 15% of the spatial domain) and by quality surface stations (continuous but of very limited number). Our aggregated surface site statistics can unfortunately mask problems inherent in RT calculations based on differing sources of cloud properties and offsetting positive/negative biases at differing sites as they average to smaller overall biases. Results concerning surface validation can also be affected by the fact that surface radiometers do not adequately represent a grid box that might be 111km square. Looking at nine well characterized surface sites we find mean bias (RMS) of model results minus observations, at the hour of the Aqua overpass where we have MODIS information for cloud properties, of -3W(22W) for LW down and 7W(103W) for SW. Though biases appear small, (RMS is expected to be large when looking at only 1 hour box of the day) the maximum positive LW bias of 10W at the COVE site, a coastal site that has a positive bias due to its location in a coastal grid box, balances with the larger negative error of -17W at the De Aar S. Africa site where we know of a large negative bias due to a negative bias of near surface air temperature in the input atmospheric profiles. Results will also be compared to the CERES CRS data product, which contains similar RT computations but at the CERES footprint level. Each product runs the RT model twice, first with original inputs and a second time with adjustments to various inputs based on comparisons to observed TOA fluxes, a “tuning” calculation. Results between the two data streams will be compared to determine the impact of tuning on both TOA and surface calculations.

Results from the two time series of data products (Terra SYN and Aqua SYN) can be compared at the same local time showing the viability of using geostationary derived cloud properties versus the higher quality cloud properties derived using MODIS information. In each case when one product is compared against the other we find an approximate increase in RMS of 5% to 10% when using geostationary (or interpolated) cloud properties in the RT calculations as compared to using the MODIS derived cloud properties. Ultimately though one hopes to find that the inclusion of geostationary products improves the diurnal modeling of the radiation incoming at the surface of the Earth across the day. A special run of the SYN code utilizing cloud properties derived at CERES observation times, which are then interpolated to all intervening hours, is compared to the full SYN results using geostationary properties to enhance diurnal modeling, bears this out. Comparisons show for the Aqua time series, in all sky SW surface insolation, little change in bias but a decrease of SW RMS from 79W to 69W, though in surface LW down there is no appreciable change. The Terra time series shows an almost identical drop in surface SW RMS with a slight decrease in surface LW RMS.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner