1. Stennis Space Center Using MODIS Data to Detect Historical Gypsy Moth Defoliation Joseph P. Spruce Science Systems and Applications, Inc. John C. Stennis Space Center USDA Forest Service, Asheville, NC September 6-7, 2007

2. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 2National Aeronautics and Space Administration Technical Contributors to Study SSAI –Joseph P. Spruce, Robert E. Ryan, James C. Smoot, Philip Kuper, Roxzana F. Moore, Kenton W. Ross, Slawomir Blonski, Mary Pagnutti CSC –Don Prados; Jeffrey A. Russell; Ronald D. Vaughan, Jr. Lockheed Martin –Jerry Gasser University of Maine –Steven Sader NASA –Rodney McKellip ITD –George May

3. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 3National Aeronautics and Space Administration Presentation Outline Project Background – Including Review of Previous Work Developing MODIS-based Defoliation Detection Methods Visualizing Gypsy Moth Defoliation on MODIS Data Maps of Gypsy Moth Defoliation Detection –From Image Classification –From Image Differencing Change Detection Miscellaneous Topics Conclusions and Suggested Future Steps

4. Stennis Space Center Background of Project

5. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 5National Aeronautics and Space Administration Review of Project Background Project initiated to help assess the potential of multitemporal MODIS inputs to the forest threat EWS –Gypsy moth is a major forest insect threat to eastern U.S. hardwoods Selected Mid-Appalachian Highlands for this study as a region prone to gypsy moth defoliation –Area included gypsy moth defoliation for multiple years during the MODIS era (2000–Present) –Area has frequent cloud cover and significant topographic relief –Area has large acreages of oak-dominated forest Initial work demonstrated that gypsy moth defoliation detection was possible with MODIS data, although more effort was needed to refine detection techniques and to validate products

6. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 6National Aeronautics and Space Administration Changes Since April Meeting Development of improved means for detecting gypsy moth defoliation, compared to initial work –Previous approach used June 2000 MODIS data and July 2001 MODIS data to detect 2001 gypsy moth defoliation –Current approach uses the entire 2000–2006 time series to estimate maximum NDVI during peak defoliation for individual years and for the entire time series, resulting in means to detect defoliation of individual years –Current approach produces imagery that is mostly much less noisy and is more useful for application at hand –Current approach also employs image enhancement techniques that improve contrast of defoliated areas on the MODIS (and Landsat) data visualization products

7. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 7National Aeronautics and Space Administration Gypsy Moth Study Area Location Study Area Outlined in Yellow Below (Total ~15.5 Million Acre ) This Area Received Extensive Gypsy Moth Defoliation in 2000-2001 Landsat Circa 2000 Mosaic is in Foreground MODIS Terra 2001 Mosaic is in Background W-VA VA PA MD Study Area - 199 x 310 km

8. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 8National Aeronautics and Space Administration USFS Sketch Maps of GM Defoliation (Localized Subset of Study Area) Sketch maps indicate that extensive defoliation occurred in 2000 and 2001 W-VA VA 2000 – CYAN 2001 – PURPLE 2002 – YELLOW 2003 – RED 2004 – MAGENTA 2005 – ORANGE 2006 – BLUE 04 km

9. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 9National Aeronautics and Space Administration Project Objectives 1.Assess multitemporal MODIS data for visualizing extent and intensity of gypsy moth defoliation 2.Assess MODIS-based gypsy moth defoliation detection products –Multitemporal image classification products –Image differencing change detection products Automated detection – via percent change in maximum NDVI during the peak defoliation period of a given year compared to maximum peak defoliation NDVI across the entire 2000–2006 time frame (considers defoliated and non-defoliated years) –Accuracy of detection products compared to reference data High-resolution NASA satellite data (Landsat, ASTER, ALI, Hyperion) USFS geospatial data (defoliation sketch maps and other data) Focus on 2001 because of extensive defoliation and ample reference data, though some assessment was also done for 2003

10. Stennis Space Center Methods for Detecting Gypsy Moth Defoliation from MODIS Data

11. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 11National Aeronautics and Space Administration RS Requirements for Detecting Gypsy Moth Defoliation Option 1: 2 dates of multispectral imagery from the growing season, including 1 defoliated and 1 non-defoliated date –Needs at least red and NIR bands –Defoliation time frame depends on weather, vegetation, and pest phenology Option 2: Single date multispectral data during defoliation –Best with red, NIR, and SWIR band coverage RS data needs to have sufficient spatial resolution to detect patches of defoliation –Moderate spatial resolution data (e.g., MODIS) for detecting extensive defoliation of larger stands within a large, broad region Effective detection requires cloud-free RS data

12. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 12National Aeronautics and Space Administration When to Look for Gypsy Moth Defoliation Within the Study Area A defoliation time frame needs to be defined as a basis for comparing forested vegetation greenness during defoliated versus non-defoliated years During outbreaks, gypsy defoliation is most visible for about a month and a half (actual time frame depends on latitude and weather) Visible peak defoliation occurs during a shorter interval, though multitemporal NDVI has potential for detecting defoliation within partially re-foliated forests Defining the defoliation time frame took into account sketch mapping schedules, gypsy moth phenology, and vegetation response as detected on various relevant Landsat and ASTER satellite data sets (see table below)

13. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 13National Aeronautics and Space Administration TSPT Preprocessing Method for MODIS Time Series Data For each year, input single date MODIS Terra data (e.g., MOD02) into TSPT –Only MODIS Terra was available for 2000 and 2001 Compute daily NDVI, masking out the clouds and cloud shadows (TSPT “Clear”) Remove aberrant temporal high/low spikes in NDVI values and eliminate extreme off nadir data (TSPT “Ideal”) Temporally filter and fill data voids to compute daily enhanced NDVI products (TSPT “Filtered”) Stack daily output from each processing step into hyper- temporal cubes (Clear, Ideal, and Filtered stacks) Repeat TSPT process for each year Settings Often Product Specific

14. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 14National Aeronautics and Space Administration Computing Maximum NDVI Composites Over Peak Defoliation Time Frame On per pixel basis, compute maximum NDVI composite for data within peak defoliation time frame (June 10–July 27) of each year Compute maximum NDVI during peak defoliation period occurring over the entire 7 year time series (2000–2006) Stack annual maximum NDVI peak defoliation images, along with maximum NDVI peak defoliation for whole 2000– 2006 time series Apply output to derive visualization, classification, and image differencing defoliation detection products

15. Stennis Space Center Visualizing Gypsy Moth Defoliation on Landsat and MODIS Data Single Date Raw Color Composite Images Multi-Temporal NDVI Color Composites

16. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 16National Aeronautics and Space Administration Views of Gypsy Moth Defoliation Using Landsat False Color Composites Gypsy moth defoliation is not evident on the June 10, 2000, Landsat scene, yet is common on the July 15, 2001, image Landsat ETM+ Acquired June 10, 2000 Scene Shows Minimal Gypsy Moth Defoliation Landsat ETM+ Acquired July 15, 2001 Includes Extensive Heavy GM Defoliation White Vectors – 2001 Sketch Map Defoliated Oak Forests White Vectors – 2001 Sketch Map Defoliated Oak Forests

17. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 17National Aeronautics and Space Administration Enhanced View of Defoliation Using 2 Dates of Landsat-Based NDVI Defoliation is enhanced with RGB visualization of 2 dates of Landsat NDVI, including a very light defoliation date (6/10/2000) and a heavy defoliation date (7/15/2001) Defoliation is further enhanced using a Gaussian look-up table stretch 2-Date Landsat NDVI RGB Color Composite ETM+ 2000 in Red; ETM 2001 in Blue & Green Landsat ETM+ Acquired July 15, 2001 Includes Heavy GM Defoliated Forest Stands White Vectors – 2001 Sketch Map Defoliated Oak Forests White Vectors – 2001 Sketch Map Defoliated Oak Forests

18. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 18National Aeronautics and Space Administration Enhanced Views of Multi-Date NDVI Products: Landsat 7 vs. MODIS MOD02 Maximum NDVI Red tones – defoliation in 2001, except some deep red is clouds on the Landsat data Both Landsat and MODIS show defoliation, though the MODIS image is cloud free MOD02 sampled time frame (for peak defoliation) is optimized compared to initial work Landsat – 6/10/2000 NDVI Loaded in Red; 7/15/2001 NDVI Loaded into Blue and Green MOD02 – Maximum NDVI Peak Defoliation All Years in Red; Same for 2001 in Blue in Green Defoliation Green Box – Location of Previous Slide Landsat 7 2-Date NDVI RGB MODIS 2-Date NDVI RGB (MOD02) Defoliation

19. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 19National Aeronautics and Space Administration Shown in April: 2-Date NDVI RGBs – LS7 vs. MODIS MOD13 – Initial Result RGB of MOD13 data above – maximum NDVI 5/25/00-7/11/2000 in Red, maximum NDVI 6/10/01–7/27/00 in Blue and Green MODIS display above shows defoliation with less contrast and more noise than previous slide because of data type, contrast stretch, and selected temporal time frames Landsat – 6/10/2000 NDVI Loaded in Red; 7/15/2001 NDVI Loaded into Blue and Green MOD13 – Maximum NDVI 2000 in Red; Maximum NDVI for 2001 in Blue in Green – see below MODIS 2-Date NDVI RGB (MOD13) Defoliation LS7 2-Date NDVI RGB Green Box – Location of Previous Slide Defoliation

20. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 20National Aeronautics and Space Administration Simulated VIIRS 400 mMOD13 250 m Noise MOD02 250 m Subtle Defoliation Shown in April: Multi-Date NDVI RGBs from MOD13, MOD2, and MOD02-based VIIRS RGB based on maximum NDVI: 5/25/00–7/11/00 in Red, 6/10/01–7/27/01 in Blue and Green Defoliation is shown in deepest red tones

21. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 21National Aeronautics and Space Administration MOD13 (16 day) 250 m Noise Gone MOD02 (Daily) 250 m Improved Contrast Simulated (Daily) VIIRS 400 m Above RGBs – Maximum NDVI Peak Defoliation All Years in Red; Same for 2001 in Blue in Green Peak Defoliation Time Frame – June 10 through July 27 Defoliation is shown in deepest red tones Enhanced: Multi-Date NDVI RGBs from MOD13, MOD2, and MOD02-based VIIRS

22. Stennis Space Center 2001 Classification Results MOD02, MOD09, MOD13, and MOD43 Product Examples Accuracy Assessment Based on Landsat / ASTER Image Interpretation

23. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 23National Aeronautics and Space Administration MODIS Products Used for Defoliation Classifications MOD02 – Radiance product – 250 to 500 m - daily MOD09 – Reflectance product – 250 to 500 m - daily MOD13 – Vegetation index products – 250 m – 16 day MOD43 – Nadir BRDF-adjusted reflectance – 8, 16 day

24. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 24National Aeronautics and Space Administration Method for Image Classification of Gypsy Moth Defoliation Produced ISODATA unsupervised classifications based on 2 date NDVI, including peak defoliation maximum NDVI for 2001 and peak defoliation maximum NDVI for whole time series (2000–2006) –Produced classifications for NDVI products from MOD02 250 m, MOD02 500 m, MOD02-based VIIRS 400 m, MOD43a 500 m, and MOD43B1 km data –Also did some testing for MOD02 500 m NDMI –Each case used same settings: 20 cluster classes, 100 iterations, and 99.5% convergence for output –Did some refinement for certain products via cluster busting technique Each case – regrouped initially into 3 classes (non-forest, green forest, and defoliated forest) Applied clump aggregation and elimination routines to dissolve patches smaller than 4 pixels for MODIS 250 m products –Dissolved patches of 2 pixels or smaller for 400 m to 500 m products Regrouped result into final classification of defoliated forest versus “other”

25. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 25National Aeronautics and Space Administration 2001 Defoliation Classification from Landsat 7 2 Date NDVIs – 30-meter Resolution 2001 Defoliation from MOD02 2 Channel Maximum Value NDVIs – 250-meter Resolution Each image shows classified defoliation areas in blue draped over a 2-date NDVI RGB 2001 Defoliation Maps from 2-Date NDVI RGBs: Landsat 7 vs. MODIS (MOD02) CLOUDS Zoom - Subset of Study Area

26. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 26National Aeronautics and Space Administration Enhanced 2-Date NDVI RGBs of 2001 Defoliation: Landsat 7 vs. MOD02 Maximum NDVI Red tones – defoliation in 2001, except some deep red is clouds on the Landsat data Both Landsat and MODIS show defoliation, although MODIS is cloud-free TSPT output MOD02 sampled time frame (for peak defoliation) is better defined than previous slide of earlier work Landsat RGB – 6/10/2000 NDVI Loaded in Red; 7/15/2001 NDVI Loaded into Blue and Green MOD02 – Maximum NDVI Peak Defoliation All Years in Red; Same for 2001 in Blue in Green Defoliation Landsat 7 2-Date NDVI RGB MODIS 2-Date NDVI RGB (MOD02) Defoliation Zoom - Subset of Study Area

27. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 27National Aeronautics and Space Administration 2001 Defoliation Classifications from MOD13, MOD2, and MOD02-based VIIRS MOD13 (16 day) 250 mMOD02 (Daily) 250 mSimulated VIIRS (Daily) 400 m Under ClassifiesBest OverallClose 2 nd Best Overall NON-FOREST – TAN HEALTHY FOREST – GREEN DEFOLIATED FOREST – RED FOREST MASK – NLCD 2001

28. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 28National Aeronautics and Space Administration Views of Defoliation from Maximum NDVI RGBs (MOD13, MOD2, and MOD02-based VIIRS) Above RGBs – Maximum NDVI Peak Defoliation All Years in Red; Same for 2001 in Blue in Green Peak Defoliation Time Frame – Set to June 10 through July 27 Defoliation is shown in red to dark red tones MOD13 (16 day) 250 mMOD02 (daily) 250 mSimulated VIIRS (daily) 400 m

29. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 29National Aeronautics and Space Administration MOD13 (16 day) 250 mMOD09 (daily) 250 mMOD 43A (8 day) 500 m Under Classifies Highest Omission NON-FOREST – TAN HEALTHY FOREST – GREEN DEFOLIATED FOREST – RED FOREST MASK – NLCD 2001 Under Classifies 2001 Defoliation Classifications from MOD13, MOD9, and MOD43A

30. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 30National Aeronautics and Space Administration Views of Defoliation from Maximum NDVI RGBs (MOD13, MOD09, and MOD43A) MOD13 and 09 – Maximum NDVI Peak Defoliation All Years in Red; Same for 2001 in Blue in Green MOD43 – Maximum NDVI using 5/17/00-6/17/00 time frame in Red; 6/18/01–8/4/01 time frame in Blue and Green Defoliation is shown in red to dark red tones MOD13 (16 day) 250 mMOD09 (daily) 250 mMOD 43A (8 day) 500 m

31. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 31National Aeronautics and Space Administration Best Overall – “Tied”~ to Best Overall NON-FOREST – TAN HEALTHY FOREST – GREEN DEFOLIATED FOREST – RED FOREST MASK – NLCD 2001 2001 Classifications from MOD02 500 m NDVI, MOD02 250 m NDVI, and MOD02 500 m NDMI MOD02 (daily) 250 m NDVIMOD02 (daily) 500 m NDVIMOD02 (daily) 500 m NDMI

32. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 32National Aeronautics and Space Administration Maximum NDVI RGBs - MOD02 250 m NDVI, MOD02 500 m NDVI, and MOD02 500 m NDMI MOD02 (daily) 250 m NDVIMOD02 (daily) 500 m NDVIMOD02 (daily) 500 m NDMI Above RGBs – Maximum NDVI Peak Defoliation All Years in Red; Same for 2001 in Blue in Green Peak Defoliation Time Frame – June 10 through July 27 Defoliation is shown in red to dark red tones

33. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 33National Aeronautics and Space Administration Classification Results for MOD43A versus MOD43B MOD43A (8 day) 500 mMOD43B (16 day) 1 KM NDVIMOD43B4 (16 day) 1 KM

34. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 34National Aeronautics and Space Administration Method for Accuracy Assessment of Defoliation Classifications Drew stratified random sample from best apparent 3 class classification (MOD02 result) with 50 samples or more selected per class An experienced image analyst interpreted apparent class value of each random sample on Landsat or ASTER RGB displays Interpretation results were cross-tabulated against each test classification (MOD02, MOD09, MOD13, MOD43A, MOD43B, VIIRS products; 250 m–1 km NDVI products; also 500 m NDMI) Final results were summarized by collapsing scheme into 2 basic classes (defoliated forest versus “other”) Computed overall % accuracy and Kappa statistics Computed % user accuracy, % producer accuracy, and Kappa statistic for each class Accuracy of the reference data is presumed higher than classifications but not “error free”

35. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 35National Aeronautics and Space Administration Location of Sample Points Used in Accuracy Assessment Sample Points (Yellow) Overlain onto MOD02 250 m NDVI RGB Location of 2001 Landsat, ASTER, and EO-1 Data Acquired During Gypsy Moth Defoliation of 2001 Landsat ASTER Hyperion

36. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 36National Aeronautics and Space Administration Relative Accuracy of Classification Products Note: PA = % Producer’s Agreement (# correct/total), UA = % User’s Agreement (# correct/total), Kappa = Kappa Statistic, OA = % Overall Agreement (# correct/total), and OK = Overall Kappa. MOD02 products yielded best agreement to reference data MOD02 400 m and 500 m NDVI had similar performance to MOD02 250m MOD09 and derivative products had higher omission but low commission error

37. Stennis Space Center Maximum NDVI Image Differencing Results for 2001 Defoliation Detection Results from MOD02 and VIIRS Accuracy Assessment of Image Differencing Results Accuracy Assessment of Defoliation Sketch Map

38. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 38National Aeronautics and Space Administration Image Differencing Method Used annual peak defoliation maximum NDVI as input Computed output as % change in maximum NDVI for a given year compared to peak defoliation maximum NDVI across entire time series (2000 – 2006) Applied threshold to output for 2001 so that >4% drop in maximum NDVI equated to defoliation –Threshold applies to MOD02 NDVI from planetary reflectance Applied clump aggregation and elimination routines to dissolve patches smaller than 4 pixels (for MOD02 250 m) –Doing so greatly reduced visually apparent commission error

39. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 39National Aeronautics and Space Administration MOD02 250m Defoliation Maps from Image Differencing and Classification MOD02 250 m Classification NON-FOREST – TAN HEALTHY FOREST – GREEN DEFOLIATED FOREST – RED FOREST MASK – NLCD 2001 Maximum NDVI Composite From Daily MOD02 250 m Data MOD02 Image Differencing Defoliation Map MOD02 image differencing product produced similar detection accuracy, compared to classification

40. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 40National Aeronautics and Space Administration Image Differencing versus Sketch Map Accuracy Assessment Methods For the image differencing products, used same Landsat/ ASTER image interpretation of same stratified random sample locations employed in classification analyses For the sketch map product, drew a new stratified random sample from the sketch map recoded to defoliated forest versus other – then interpreted class value on sample locations of Landsat/ASTER image displays –A new sample was drawn because the sketch map product is visually quite different from the MOD02 classification (therefore with different error distributions)

41. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 41National Aeronautics and Space Administration Relative Accuracy of Image Differencing Detection Products vs. Sketch Map 0.47 73.64 (95/129)0.47 78.67 (59/75) 76.62 (59/77)0.46 66.67 (36/54)) 73.47 (36/49) Defoliation from USFS Sketch Map 0.67 85.55 (148/173)0.62 87.60 (106/121) 91.38 (106/116)0.71 80.77 (42/52) 73.68 (42/57) MOD02-Simulated VIIRS NDVI 400m 0.72 87.28 (151/173)0.76 91.96 (103/112) 88.79 (103/116)0.68 78.69 (48/61) 84.21 (48/57)MOD02 NDVI 250m OKOAKappaUAPAKappaUAPA OverallOtherDefoliated Forest Product Note: PA = % Producer’s Agreement (# correct/total), UA = % User’s Agreement (# correct/total), Kappa = Kappa Statistic, OA = % Overall Agreement (# correct/total), and OK = Overall Kappa. MOD02 NDVI products from image differencing and classification both produced accurate maps of gypsy moth defoliation, leading to belief that automated detection is feasible.

42. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 42National Aeronautics and Space Administration Conclusions for Defoliation Mapping Products Use of mapping technique enabled regional map of gypsy moth defoliation from multitemporal MODIS data Use of higher temporal resolution data (daily) improved accuracy, compared to 8 and 16 day composites Optimizing temporal processing settings of daily data improved results over initial runs MOD02 250m and 500m defoliation maps were similar –1 km products seam to be too coarse MOD09 products did not perform as well as expected –Maximum NDVI compositing technique did not work as well for MOD09, perhaps due to correction and resampling artifacts –Improved processing with custom TSPT settings should help…

43. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 43National Aeronautics and Space Administration Additional Comments Clouds and atmospheric noise were mitigated using TSPT and multiple dates of MOD02 data collected during the peak defoliation time frame Use of cloud frequency statistics and pest phenology data aided MODIS-based detection of gypsy moth defoliation NDVI or NDMI based detection requires only requires processing of 2 bands per date, which is less compared to Tassel Cap This work was aided by NASA-funded project to assess potential of VIIRS data for mapping gypsy moth defoliation MOD02 data performed better than MOD43 data and maybe better for near real time requirements implicit to early warning systems

44. Stennis Space Center Miscellaneous Topics Annual Cloud Frequency Statistics During Peak Defoliation Preliminary Analysis of 2003 MODIS-Based Defoliation Data GIS Overlay of 2001 MOD02 Defoliation versus Sketch Map Effects of Patch Size on 2001 Detection Accuracy

45. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 45National Aeronautics and Space Administration Annual Cloud Frequency Statistics for Peak Defoliation Time Frame Cloud statistics for MOD02 250 m NDVI time series Gypsy moth study area – 2000 – 2006 time frame PD = Peak Defoliation The cloud frequency illustrates why cloud-free Landsat data during gypsy moth defoliation is very spotty and difficult to acquire The cloud frequency had to be considered even when developing defoliation detection techniques from the daily MODIS data

46. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 46National Aeronautics and Space Administration Defoliation 2003 – MODIS 2 Date NDVI Compared to Single Date Landsat Data Landsat RGB – July 23, 2003 Defoliation on MOD02 – red to deep red tones Defoliation on Landsat bands 4,5,3 RGB – yellow green tones 2003 sketch map data shown as yellow vectors – small patches in south MODIS 2003 Peak Defoliation

47. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 47National Aeronautics and Space Administration Single Date MODIS Color Composite versus Multi-Date MODIS NDVI RGB MODIS – June 6, 2003 MODIS – 2003 PD Time Frame Gypsy Moth PDTF Set to June 10–July 27 Tent Caterpillar Defoliation ??? Defoliation Still Visible Refoliation?

48. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 48National Aeronautics and Space Administration Single-Date 2003 MODIS RGB, Defoliation Map, and Sensor Zenith Angle Image MOD02 RGB – June 6, 2003 MODIS Defoliation - June 6, 2003SZA – MODIS – June 6, 2003 43.96 56.48 TSPT thresholds out SZA at 47 degrees or more

49. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 49National Aeronautics and Space Administration MOD02 250m NDVI Classification vs. Sketch Map Overlay of MOD02 250 m NDVI Classification and Sketch Map ASTER Scenes from 7/24/01Zoom of Overlay on Left Legend Little defoliation was evident on ASTER imagery for sketch mapped areas of moderate defoliation

50. Using MODIS Data to Detect Historical Gypsy Moth Defoliation Stennis Space Center 50National Aeronautics and Space Administration Relationships of Patch Size to MOD02 250-m NDVI Classification Accuracy

51. Participation in this work by Science Systems and Applications, Inc., and by Computer Sciences Corporation was supported by NASA at the John C. Stennis Space Center, Mississippi, under Task Order NNS04AB54T.