GEO-CAPE Ocean Color Study: The influence of cloud coverage space and time scales on the retrieval of ocean surface reflectance and chlorophyll fluorescence measurements. John R. Moisan NASA, Goddard Space Flight Center Wallops Flight Facility, Wallops Island, VA 23337 Ph: 757-824-1312; Cell: 443-735-0086 e-mail: John.R.Moisan@NASA.gov Previous Cloud Analysis Results In the previous year, hourly cloud observations for 2007 of the U.S. east and west coasts were obtained from the GOES Cloud Sounder Effective Cloud Amount (ECA) observations from the University of Wisconsin’s Cooperative Institute for Meteorological Satellite Studies (Li et al, 2009). Effective Cloud Amount is the product of cloud fraction and emittance. These observations are at a nominal spatial resolution of 12 km. A comparable cloud product is available using the MODIS instrument at spatial scales of 5 km (Baum and Platnick, 2006). The full time series of 8760 hourly observations was analyzed to derive the probabilities of acquiring ocean color observations from areas where the ECA was 0. The first component of the analysis was to use the daily local noon observations in order to essentially develop a “baseline” probability (Figure 1) of acquiring ocean color imagery for a traditional Low Earth Orbiting (LEO) ocean color satellite such as MODIS/Aqua or SeaWiFS. The next set of analysis focused on understanding how more frequent opportunities of acquiring an image over the period of a day could improve the level of probability of getting at least one pixel imaged per day. The results (Figure 2) show that there is a high probability over all coastal ocean regions for getting at least one image per day for a specific pixel/region. These results are directly comparable with the previous LEO results from Figure 1. By having a satellite in geostationary orbit the probability of acquiring a cloud free image once per day is much higher than of that from a satellite in LEO. Figure 1. The probability of obtaining a clear pixel from a local noon GOES observation. Areas of high likelihood of obtaining clear pixels (such as the Sea of Cortez) have values close to 1, while chronically cloudy areas such as off the North Pacific coast have low probability values. The overlap in the image is because there are 2 separate satellite views employed in the analysis. Figure 2. The probability of acquiring at least 1 clear pixel during the daytime. The high probabilities are a derived from the fact that there are multiple times during the daytime that could be imaged and that cloud formations (fronts and storms systems) tend to move through a region in less than half a day. Further analysis focused on estimating the probability of obtaining multiple observations over the period of one day (6AM to 6PM) period. These probability maps essentially morph from the high probability image of acquiring at least 1 clear pixel per day to the lower probability image of obtaining a local noon cloud free pixel. For ocean color applications that require only one image per day a geostationary imager will provide much more observations that from that of a LEO satellite. But, for those applications that require more than one cloud free image per day, such as studies to investigate the diurnal variability of phytoplankton fluorescence, the probabilities of obtaining the necessary observations will vary depending on the level of daily observations that are required to resolve the signals. Proposed FY10 Work Plan In the coming year, the proposed work will focus on improving our understanding of the spatial variability of the clouds in coastal areas. The GOES cloud product is at a nominal 12 km resolution. There are a number of satellite cloud products that have higher spatial resolutions (MODIS: 1km; Landsat: 30m), but with much reduced temporal resolution. The work in the coming year will investigate the spatial scales of clouds using these finer spatial resolution data sets and attempt to modify the GOES cloud product to improve the present cloud masking by optimizing it with present ocean color cloud masks. The goal here is to improve our understanding of the spatial scales of cloud signals in the scales ranging from 30 m to 12 km. In addition to this, work will progress towards completion of an instantaneous hourly GOES PAR product that will provide the necessary time periodicity for supporting GEO-CAPE studies aimed at understanding the likely diurnal signals from phytoplankton fluorescence. Task 1: Comparison of GOES ECA data with MODIS cloud mask data: MODIS produces a 1 km Level-2 cloud mask product (MOD35). The objectives of this task will be to compare GOES ECA data to its time-adjacent MODIS cloud mask in order to establish what level of ECA’s compare best with the MODIS cloud mask. The GOES cloud data set will be expanded to include 2000-2002. This will yield about 1095 individual scenes and allow for comparison of various seasons as well as annual means. This period of time was chosen because it corresponds with the period of time (19992003) that the Landsat-7 Scan Line Corrector was in operation, a requirement for Task 2 below. The ‘optimized’ ECA value will then be used as an effective cloud mask for the GOES ECA observations used in the remaining analysis. Task 2: Characterizing cloud spatial scales using GOES, MODIS and LANDSAT cloud observations: Landsat-7 Enhanced Thematic Mapper Plus (ETM+) data are now available for free via the web at landsat.usgs.gov. These observations are at 30 m resolution and are available on a 16-day repeat cycle. This data set was processed using the Landsat-7 Processing System that determines the percentage of cloud cover by applying the Automatic Cloud Cover Assessment (ACCA) algorithm (Figure 3). At 30m resolution, this data set provides 1,111 pixels for each MODIS 1km pixel and 160,000 pixels for each of the GOES 12km pixels and should provide the spatial coverage needed to characterize the spatial resolution requirements and for understanding the sub-pixel cloud contamination issues. The combination of 3 different spatial scale cloud products (30 m, 1 km, and 12 km) will be used to characterize the cloud spatial statistics over a wide range of spatial scales. The Landsat-7 cloud product will be used as a “truth” field and resampled/mapped into coarser resolution grids (100m x 100m; 200m x 200m; 500m x 500m; 1km x 1km, and 12km x 12km). The resulting courser cloud products will be compared to quantify how increased levels of spatial resolution improve the ability of a geostationary sensor. Those resampled cloud products that are at the spatial resolution of the MODIS and GOES cloud products will be compared to the spatial analysis of those. Figure 3. Example of the 30 m resolution Landsat-7 ACCA cloud product. The visible channel (A) shows a land area with several small spotted cloud features. The ACCA algorithm identifies these cloud features and creates an image mask (B), where the colors yellow and orange indicate clouds, blue indicates pixels adjacent to clouds, dark green is for cloud shadow areas, and light green is clear conditions. (Figure from Dr. E. Vermote, UMd) Task 3: Hourly instantaneous GOES surface PAR estimates: The MODIS PAR product is a course resolution 1 km global area coverage (GAC), daily-averaged PAR product that is really only meaningful when averaged over multiple days. A similar PAR product is available for SeaWiFS it may be a better product as it does not saturate over clouds so easily. However, for both of these, sunglint remains a much bigger problem for the PAR product. The focus of this task is to use the radiative transfer model of Gregg and Carder (1990) along with the GOES cloud mask data set to provide estimates of Photosynthetically Available Radiance (PAR) estimates in support of GEOCAPE studies that will investigate the feasibility of measuring chlorophyll fluorescence at dawn and dusk periods. This model is already operational and has been easily merged with the GOES cloud observations to yield hourly estimates of PAR observations during cloud free conditions (Figure 4). This task will be a collaborative effort with others in the GEO-CAPE team. Specifically, we will work with Dr. Hu (USF) who is proposing to develop a sun glint parameterization and with Drs. H. Sosik and T. Moisan who are independently proposing to investigate the issue of measuring sun-stimulated phytoplankton fluorescence. Finally, we will implement the Zheng et al. (2009) algorithm for instantaneous PAR using GOES observations and validate them using local ground based FLUXNET sites. Figure 4. January-March 2007 hourly instantaneous cloud free PAR (umE m-2 s-1) for a GOES coastal ocean pixel adjacent to the Chesapeake Light Tower. The red dots indicate GOES daytime PAR values, blue indicates nighttime conditions. The black dots indicate times of MODIS PAR estimates. References: Gregg, W. W., and K. L. Carder. A simple spectral solar irradiance model for cloudless maritime atmospheres. Limnol. Oceanogr. 35, 1657-1675, 1990. Baum, B. A., and S. Platnick, Introduction to MODIS Cloud Products, In: Earth Science Satellite Remote Sensing, Qu, J. J., W. Gao, M. Kafatos, R. E. Murphy, and V. V. Salomonson (eds.), Springer, Berlin, pp. 74-91, 2006. Li, Z.L., Li, J., Menzel, P., Nelson, J.P., Schmit, T.J., Weisz, E., & Ackerman, S.A. Forecasting and nowcasting improvement in cloudy regions with high temporal GOES sounder infrared radiance measurements. Journal of Geophysical Research - Atmospheres, 114, D09216, 2009. Wu, D.L., Ackerman, S.A. , Davies, R., Diner, D.J., Garay, M.J., Kahn, B.H., Maddux, B.C., Moroney, C.M., Stephens, G.L., Veefkind, J.P., & Vaughan, M.A. Vertical distributions and relationships of cloud occurrence frequency as observed by MISR, AIRS, MODIS, OMI, CALIPSO, and CloudSat. Geophysical Research Letters, 36, L09821, 2009. Zheng, T., S. Liang, and K. Wang, Estimation of incident PAR from GOES imagery, Journal of Applied Meteorology and Climatology, 47, 853-868, 2008.