Orbit Characteristics and View Angle Effects on the Global Cloud Field Brent C. Maddux1,2 Steven A. Ackerman1 Steve Platnick3 Paul Hubanks4 1Cooperative Institute for Meteorological Satellite Studies 2Department of Atmospheric and Oceanic Sciences-U of Wisconsin 3NASA Goddard Space Flight Center, Greenbelt, Maryland 4Wyle Information Systems, McLean, VA 22102

Comparison to other cloud datasets (apples-to-apples) Motivation Comparison to other cloud datasets (apples-to-apples) Place ‘error bars’ on global mean statistics Quantify cloud variability globally

Daytime Cloud Fraction Coast lines evident Aqua 7 Year Mean Global Means Daytime Cloud Fraction Coast lines evident Differences where expected: maritime stratocumulus decks, SH land, etc. Terra 9 Year Mean Spatial differences aren’t very pronounced, some differences over stratocumulus and land areas. 0 .25 .5 .75 1.0

MODIS Cloud Fraction Terra (Red) and Aqua (Black) Global Cloud Amount Anomaly CA Anomaly (%) 1970 1975 1980 1985 1990 1995 2000 2005 2010 3 2 1 -1 -2 -3 Time series is very stable and Aqua and Terra MODIS agree very well

Cloud Fraction vs Viewing Angle 7 years of Aqua and Terra 16% increase from near nadir to edge of scan View angle effect not constant for all cloud types Cloud Fraction vs Sensor Zenith Angle Sensor Zenith Angle Cloud Fraction (%)

Single Day in the MODIS Orbit Geostationary-like view 16 day orbit procession U-Shape at edges of convection and cloudy regions Increase in pixel size with view angle Terra Daytime Cloud Fraction 0 .25 0.5 .75 1.0

Daytime Cloud Fraction Mean Nadir to 10° 60° to edge of scan° Nadir to 10° minus 60 and greater Differences not uniform Largest differences not where thin high clouds exist

Nadir vs Edge of Scan Changes are not uniform Near Nadir Near Edge of Scan Nadir vs Edge of Scan Cloud Fraction (%) 0 0.2 0.4 0.6 0.8 1 Changes are not uniform Largest changes in CF aren’t the same as largest changes in optical properties Near Nadir ≤10° Near Edge of Scan ≥50° Cloud Top Pressure (hPa) 200 375 550 775 900 Cloud Effective Radius Ice (μm) 15 20 25 30 35 Cloud Effective Radius Liquid (μm) 5 10 15 20 25 Cloud Optical Depth Ice 15 20 25 30 35 Cloud Optical Depth Liquid 5 10 15 20 25

L3 Statistical Uncertainty Study: Zonal Retrieval Statistics vs L3 Statistical Uncertainty Study: Zonal Retrieval Statistics vs. VZA Effective Radius, water clouds Similar biases for all categories, and reasonably symmetric => partly cloudy pixel but not shadowing impact OR vertical microphysical structure Significance of these biases? What about regional biases?

L3 Statistical Uncertainty Study: Global Retrieval Statistics vs L3 Statistical Uncertainty Study: Global Retrieval Statistics vs. Zenith Angle Effective Radius, water clouds, MODIS Aqua, April 2005 operational (all bins) backward oblique bin (-53° to -65° VZA) forward oblique bin (53° to 65° VZA) fwd. nadir bin 12 ≥24

Mean Cloud Fraction Difference (MOD35-MOD06) This is the 7 year mean difference of the cloud mask and the retrieval fraction. You can see that the differences are greatest over regions where broken cumulus, aerosols and highly reflective surfaces. 0 10 20 30 40 (%) MOD06 cloud fraction is a quality assured subset of MOD35 to retrieve better cloud optical properties

Mean Cloud Fraction Difference in Percent (MOD35-MOD06)/MOD35 This is the 7 year mean difference of the cloud mask and cloud retrieval fraction divided by that cloud mask fraction. By looking at the difference in percent it is a little more obvious where clouds are being missed relative to the cloud fraction, but these regions tend to have few clouds anyway, e.g. the Sahara or mountain ranges. 0 30 60 (%) Differences are due to the QA stuff(clear sky restoral and cloud edges, thin clouds, and surfaces influences).

Difference in high and low cloud fraction CTP histogram from CTP (red) and CTPvsOD (blue) histogram Pressure Level (hPa) This is the mean CTP histogram (red) SDS for 1 year of data and the cloud top pressure histogram from the mean CTP vs OD (water and ice) for global data. These are relative numbers, i.e. all red dots add to 100 and all blue dots add to 100. This shows that after the optical properties are excluding significantly more low cloud. Differences are due to the QA (clear sky restoral and cloud edges, thin clouds, and surfaces influences). Cloud Fraction Difference in high and low cloud fraction

Cloud Fraction Day Difference for 7 yrs (Aqua minus Terra) OCT JUL APR JAN NOV AUG MAY FEB DEC SEP JUN MAR -30 0 30

Grid Cell Size and Swath Overlap Cloud Top Pressure vs Latitude Swath overlap causes more cloud to be averaged into the middle cloud bin 50 40 30 Cloud Fraction 20 At 1 degree resolution when the mean gridcell cloud fraction is between 700 and 400 hPa the clouds in that gridcell are only mid clouds about 21 % of the time. Compared to ¼ degree resolution where the number of mid clouds is 56. 10 EQ 15 30 45 60 75 90 Latitudes Included

Grid Cell Size and Swath Overlap Cloud Top Pressure Histogram High Cloud <400 hPa Mid Cloud >400 and <700 hPa Low Cloud >700 hPa At 1 degree resolution when the mean gridcell cloud fraction is between 700 and 400 hPa the clouds in that gridcell are only mid clouds about 21 % of the time. Compared to ¼ degree resolution where the number of mid clouds is 56.

Cloud fractions need to characterize the global cloud trend within ±1% Summary Cloud fractions need to characterize the global cloud trend within ±1% View angle dependencies are large across swath 16% cloud fraction (>60 locally) 30hPa for cloud top pressure (>200 hPa locally) 2μm for effective radii (10μm locally) 2 for optical depth (20 locally)

Pixel vs Area Weighting Not a uniform offset Doesn’t change long term mean (.2%) Polar regions oscillate opposite mid-latitudes MODIS Terra Cloud Fraction Area (Red) and Pixel (Blue) If data is weighted by pixel counts, such that each pixel is included equally then the long term mean is virtually unchanged BUT the timeseries is not only different is is out of phase. This is not simply an aggregation issue, but an issue of what data to include, the red line is equivalent to a single overpass data set, the other is a multiple over pass data set.

L3 Statistical Uncertainty Study: Zonal Retrieval Statistics vs L3 Statistical Uncertainty Study: Zonal Retrieval Statistics vs. VZA Optical Thickness, water clouds Bias is within the range of the mean pixel-level uncertainties for nadir views. Not that bias should be directly equated with these mean uncertainties, but helpful – or at least practically useful – to know that the mean uncertainties appear to provide a bias limit.

Summary of VZA Results For water clouds: Zonal mean statistics: at all latitudes (land and/or ocean) VZA dependencies for t and re are within product’s mean instantaneous uncertainties Optical Thickness zonal means: ocean clouds show some asymmetry (slight indication of shadowing on sun view side of scan?) land clouds symmetric. Effective Radius: no obvious shadowing effect (symmetric), both zonal and some regional analysis

This figure would argue for not using the edge of scan, especially at higher latitudes.

Cloud Retrieval Fraction for one month (MOD06 optical properties cloud fraction) This is the cloud retrieval fraction for a single July. It is simply the fraction (successful retrievals)/(all possible retrievals). So this cloud fraction will be smaller over regions of small/broken clouds, or over surfaces with hard to retrieve clouds, i.e. sunglint, desert, ice… 0.0 0.5 1.0

Cloud Fraction for one month (MOD35) This is the MODIS cloud mask cloud fraction for a single july. This is the actual cloud fraction regardless of whether a microphysical cloud retrieval can be run. 0.0 0.5 1.0

Mean Cloud Fraction Difference for on month (MOD35 minus MOD06) This is the difference cloud mask minus the cloud retrieval fraction for the first two slides. 0.0 0.2 0.4

Mass Concentration for 1 orbit day

Cloud Fraction MODIS cloud mask (MOD35, described in Ackerman et. al 1999) Up to 19 channels in vis and IR 1km to 250m resolution For Level-3 statistics there are clear and cloudy pixels

Two types of variability Natural: Artificial: Introduction Two types of variability Natural: Time and spatial scales Artificial: Instrument (Aqua-Terra Differences) Orbit (polar vs geostationary) Algorithm- channels choices

Find distribution of re and od vs solar angles Pdfs from poster 2.1 and 37 differences TTD: CTPDH vs angle for Aqua 2004 Find distribution of re and od vs solar angles Pdfs from poster Make a error bar plot on dataset Include figures on angle attribution for two main influence, pixel size and thin clouds at oblique angles