Students as Ground Observers for Satellite Cloud Retrieval

Students as Ground Observers for Satellite Cloud Retrieval Validation Lin H. Chambers, P. Kay Costulis, David F. Young NASA Langley Research Center, Hampton, VA Tina M. Rogerson Science Applications International Corporation, Hampton, VA 13 th Conference on Satellite Meteorology & Oceanography Norfolk, VA Sept. 2004

OUTLINE • Sources of student data • What is the S’COOL Project? • Comparisons – Proof of Concept – New Comparisons – Bright Surfaces • Conclusions

Sources of Student Data Ø The CERES S’COOL Project – More than 35, 000 complete observations – 9, 172 now have corresponding CERES data – http: //scool. larc. nasa. gov • The GLOBE Program – More than 2. 5 M cloud data points – http: //www. globe. gov

What is S’COOL? • Began Jan. 1997 • Students’ Cloud Observations On-Line • K-12 Education and outreach portion of CERES: Clouds and the Earth's Radiant Energy System • 1700+ participants in 65 countries • Focused on obtaining ground-based cloud observations for validation of the CERES data

The S’COOL Concept • Students provide ground observations for CERES overpass 1. Determine satellite overpass time 2. Observe cloud properties 3. Transmit results to NASA 4. Compare to satellite-retrieved properties • Data of value to CERES scientists • Real-world learning for students

Data Collected Cloud type Contrails Cloud cover Visual opacity Surface Cover Surface Measurements Comments

Comparing to Satellite • S’COOL Site Matched to 1 degree Satellite Region • Observation Times Within 15 Minutes

FIRST S’COOL Comparison Cloud Observations Over Gloucester, VA January 13 & 17, 1997

Initial Comparisons: Proof of Concept • Measurements in 1998 • CERES on TRMM only (~50 correspondences) • Augmented with AVHRR and geostationary data (~50): Analyzed by hand

Cloud Amount Comparison 62% in complete agreement 0% in complete disagreement Stats: Chi-Squared value of 82; significant to 5 e-12

Cloud Layer Comparison

Interim Conclusions • Clearly some useful information • Insight into cloud layering • Insight into sparse, thin cirrus • Educationally a big success

New Comparisons - 2004 • New CERES angular models (see talk by Loeb this afternoon) • CERES on TRMM, Terra, Aqua • Feb. 1998 to April 2004 • Production data products

Data Available Max: 479 (High School in Pennsylvania) Min: 1 (70 schools)

Cloud Amount Comparison 54. 5% in complete agreement 2% in complete disagreement Stats: Chi-Squared value of 5636; significant!!!

Students Overcast vs. Satellite Clear (48 cases) • Spatial Mismatch? : >1/3 are schools located less than 0. 1 degree from the edge of a lat/long grid box. • Universal Time? : 3 cases with incorrect UT • Student/Satellite error? : remaining cases have no clear explanation. Study needed. • Snow: 10 cases, yet the satellite still reports clear sky.

Students Clear vs. Satellite Overcast (143 cases) • Spatial mismatch? : About 22% • Universal Time? : ~10 • Snow? : 18 cases students report snow. • Only one satellite retrieval is suspect: low cloud temperature 2. 5 K below the surface temperature. • Satellite/Student error? : stratus = clear?

Cloud Amount Comparison 191 3 -class errors (2%) - ~1/3 easily explainable 711 2 -class errors (8%) - need more study 3271 1 -class errors (36%) - may be near-matches

First look at 1 -class errors Students say 0 -5% cloud Satellite says 5 -50% cloud 20% of CERES has 10 < fc < 15 24% of CERES has 5 < fc < 10

Subvisual Cirrus? • MODIS vs GLOBE cloud type comparison indicated some subvisual cirrus (Stephens and Rogers, 2004). • This CERES/S’COOL dataset: • 19 cases where ttot< 3 • None high cloud only • 5 cases where ttot < 1 • None high clouds • No evidence of subvisual cirrus in this dataset • May be due to location of the S’COOL student data, over land with few data points in the Tropics.

Cloud Layer Comparison

Effect of Bright Surfaces • 1057 reports (~11%) with snow or ice in ground report • Data from 1/4 of respondents Max - 86 (4 th grade in NH) Min - 1 (31 schools)

Snow Effect on Cloud Amount Ground Observers Chi-squared = 671 All S A T Clr PC MC OV Snow/Ice S A T Clr PC MC OV Clr 1415 624 193 143 Clr 147 67 35 18 PC MC OV 350 101 48 897 575 230 553 695 680 187 489 1992 Ground Observers PC MC OV 27 8 10 75 35 19 59 68 95 33 52 309

Snow Effect on Cloud Amount All - scaled S A T Clr PC MC OV Snow/Ice S A T Clr PC MC OV Ground Observers Clr 163 72 22 16 Clr 147 67 35 18 PC MC 40 12 103 66 64 80 22 56 Ground Observers PC MC 27 8 75 35 59 68 33 52 OV 6 27 78 230 OV 10 19 95 309

Snow Effect on Cloud Layers Ground Observers All S A T No 1 -layer Multi. Cloud layer No Cld 950 615 100 1 -layer 306 2030 581 Multi 249 3306 1035 Ground Observers Snow/Ice No 1 -layer Multi. Cloud layer No Cld 92 73 5 S A 1 -layer 42 295 54 T Multi 31 400 65

Snow Effect on Cloud Layers Ground Observers All - scaled S A T No 1 -layer Multi. Cloud layer No Cld 110 71 12 1 -layer 35 235 67 Multi 29 383 120 Ground Observers Snow/Ice No 1 -layer Multi. Cloud layer No Cld 92 73 5 S A 1 -layer 42 295 54 T Multi 31 400 65

Conclusions First major analysis of student ground observer data to validate cloud retrievals from a satellite instrument. A few pitfalls are evident. Useful information can be derived.

Future Plans Inviting S’COOL participants to do detailed analysis of their correspondences More analysis to be done (2 -class and 1 -class errors, cloud levels, opacity…. ) Data available via the Internet for analysis: http: //asd-www. larc. nasa. gov/SCOOL/usedata. html

Acknowledgments Science and Education support from NASA’s Earth Science Enterprise. This work would not be possible without the participation of our extended network of educators and their students, and we thank them most sincerely for their efforts.