Validation of Daytime CRTM Performance Using AVHRR IR 3.7 μm Band and Suggested Improvements

The community radiative transfer model (CRTM) is a key part of the Advanced Clear-Sky Processor for Oceans (ACSPO) newly developed at NESDIS. CRTM is used in ACSPO in conjunction with the upper air fields of the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) and Reynolds sea surface temperature (SST) to calculate model clear-sky brightness temperatures (BT) in AVHRR Ch3B (3.7 µm), Ch4 (11 µm) and Ch5 (12 µm) onboard NOAA-16, -17, -18, -19 and MetOp-A. Simulated BTs are subsequently used in ACSPO to improve cloud masking and quality control, and to explore physical SST inversions to potentially replace the current regression algorithms. Consistency between model and observations is critically important for these applications.

A near real-time, web-based tool that monitors IR clear-sky radiances over oceans for SST (MICROS; www.star.nesdis.noaa.gov/sod/sst/micros/) has been established to monitor global M-O biases and facilitate physical SST retrievals, improved cloud detection and quality control of AVHRR BTs in ACSPO. CRTM performance at night has been validated elsewhere and found to be sufficiently accurate for these ACSPO applications. During daytime, however, global M-O distribution in Ch3B shows an unrealistic cold bias (~20 K) in sun glint areas and an unexplained warm bias (~+5 K) elsewhere. Special analyses have shown that these anomalies are due to the surface reflectance model employed in CRTM version 1.2.

An instantaneous specular reflectance model, or ISRM, based on Cox-Munk (1954) facet distribution and summarised by e.g. Breon (1993) and Gordon (1997), was therefore tested to see if the accuracy of CTRM BTs can be improved. The ISRM dramatically reduces the M-O bias in Ch3B, to ~-2 K in sun glint area and to ~-1 K elsewhere. Based on these analyses, the ISRM will be employed in CRTM version 2.

Work is underway to better understand and minimize the remaining M-O biases in CRTM version 2. The negative bias outside glint area is likely due to missing aerosol in CRTM and the use of Reynolds SST, which currently does not resolve the diurnal cycle. Exploring a simple model to adjust Reynolds SST for the effect of diurnal cycle is currently underway. The positive bias in sun glint area, on the other hand, is deemed to be due to the remaining systematic biases in the empirical Cox-Munk model. To check this hypothesis, facet probability density function (PDF) was empirically estimated by inverting the CRTM daytime radiative transfer equation, and the inverted PDF was compared with the Cox-Munk PDF. Unlike the Cox-Munk PDF, the inverted PDF does not asymptotically approach zero as one moves away from sun glint, consistently with the negative M-O bias observed away from sun glint due to missing aerosol in CRTM and unresolved diurnal cycle in Reynolds SST. In the glint area (near 0° of facet slope angle), the Cox-Munk PDF is overestimated for wind speed <2 m/s, matches the inverted PDF for medium wind speeds (6-8 m/s) and becomes underestimated at higher wind speeds.

Future work will focus on reducing uncertainty in the derived PDF (in particular, by including aerosols in CRTM and by correcting Reynolds SST for the effect of diurnal cycle). Once accurate empirical PDF is established, it will be either parameterized analytically (by, for example, recalculating the coefficients in the Cox-Munk model or testing a new non-Gaussian fit function) or as a look-up table. The new ISRM will then be comprehensively tested using not only polar but also geostationary data (e.g., MSG/SEVIRI and GOES-R/ABI), which progress through the full illumination cycle daily and, therefore, cover the glint angle range more fully and uniformly.