Mapping burn severity in heterogeneous landscapes with a relativized version of the delta Normalized Burn Ratio (dnbr)

Transcription

1 Mapping burn severity in heterogeneous landscapes with a relativized version of the delta Normalized Burn Ratio (dnbr) Jay D. Miller USDA Forest Service 3237 Peacekeeper Way McClellan, CA jaymiller@fs.fed.us Andrea E. Thode School of Forestry Northern Arizona University P.O. Box Flagstaff, Arizona andi.thode@nau.edu RS2006

2 Applications of Vegetation Change Detection Change in landcover type Absolute change in vegetation related variables live biomass, LAI, cover Magnitude of impact due to disturbance, i.e. Severity

3 What is Severity? Effect of the disturbance event on the organism, community, or ecosystem. Turner 1998, White and Pickett 1985 Fire severity describes the magnitude of the disturbance and, therefore, reflects the degree of change in ecosystems. Neary et al 2005 Fire Severity cannot be expressed as a single quantitative measure that relates to resource impact. Therefore, relative magnitude of fire impacts, expressed in terms of the percent of soil organisms affected by the fire, proportion of trees and shrubs killed, number of postfire resproutings of the trees or shrubs, or other measures of disturbance, are better placed into broadly defined discrete classes ranging from low to medium to high fire severities. DeBano et al 1998

4 Fire Severity Categories Unchanged Unable to detect a change pre- to post-fire by satellite. This does not mean the site did not experience fire. Low Fire is generally nonlethal to the dominant vegetation and does not substantially change the structure of the dominant vegetation. Derived from Understory fire regime Brown & Smith 2000 Moderate Fire causes selective mortality in the dominant vegetation, depending on different species susceptibility to fire, or varies between understory and stand replacement. Derived from Mixed severity fire regime Brown & Smith 2000 High Fire kills aboveground parts of the dominant vegetation, changing the aboveground structure substantially. Derived from Stand replacing severity fire regime Brown & Smith 2000 Brown, J.K. & Smith, J.K., (Editors) (2000). Wildland fire in ecosystems: effects of fire on flora. General Technical Report, RMRS-GTR-42-vol. 2. Ogden, UT: USDA Forest Service, Rocky Mountain Research Station, 257 pp.

5 Some Thoughts If all stand replacing fire is defined as high severity, then the definition of severity is independent of how much living vegetation was in the stand prior to any fire occurring. The high severity category (stand replacing fire) should produce the highest producer accuracies since it should be the easiest to detect. From a management perspective, we want high severity patches to have high accuracies.

6 Change Detection Algorithm Methods used in past change detection studies. Table adapted from Singh 1989 Analysis technique used to detect change Raw Data Band Difference Band Ratio Vegetation Index Difference Regression Principal Components Change Vector Post- Classification Comparison Thresholds X X X X X Supervised Classification Unsupervised Classification X X X X X X X X X Normalized Burn Ratio (NBR) [(Landsat TM Band 4 Band 7) / (Band 4 + Band 7)] * 1000 Key, C.H. & Benson, N.C. (2005). Landscape assessment: remote sensing of severity, the Normalized Burn Ratio. In: D.C. Lutes et al. (Editors), FIREMON: Fire Effects Monitoring and Inventory System., Ogden, UT: USDA Forest Service, Rocky Mountain Research Station, General Technical Report, RMRS-GTR-164-CD:LA1-LA51.

7 Prefire Moderate % Canopy Cover NBR = 400 Postfire NBR = 25 dnbr = 375 Stand replacing Fire High % Canopy Cover NBR = 800 NBR = 50 dnbr = 750 Stand replacing Fire Both of these sites experienced stand replacing fire, i.e. high severity fire.

8 Prefire Moderate % Canopy Cover NBR = 400 Postfire NBR = 25 dnbr = 375 Stand replacing Fire High % Canopy Cover NBR = 800 NBR = 400 dnbr = 400 Mixed lethal fire If both of these sites occurred within the same fire perimeter, one of them would get miss-classified when thresholding dnbr.

9 Study area

10 Strata and Factors are defined in FIREMON Landscape Assessment, Chapter 2, and on accompanying BI cheat sheet. BURN SEVERITY -- COMPOSITE BURN INDEX (BI) PD - Abridged Examiners: Fire Name: Registration Code Project Code Plot Number Field Date mmddyyyy / / Fire Date mmyyyy / Plot Aspect Plot % Slope UTM Zone Plot Radius Overstory UTM E plot center GPS Datum Plot Radius Understory UTM N plot center GPS Error (m) Number of Plot Photos Plot Photo IDs BI Long Form % Burned 20 m Plot = % Burned 30 m Plot = Fuel Photo Series = STRATA RATING FACTORS BURN SEVERITY SCALE No Effect Low Moderate High FACTOR SCORES A. SUBSTRATES % Pre-Fire Cover: Litter = Duff = Soil/Rock = Pre-Fire Depth (inches): Litter = Duff = Fuel Bed = Litter/Light Fuel Consumed Unchanged -- 50% litter % litter >80% light fuel 98% Light Fuel Duff Unchanged -- Light char -- 50% loss deep char -- Consumed Medium Fuel, 3-8 in. Unchanged -- 20% consumed -- 40% consumed -- >60% loss, deep ch Heavy Fuel, > 8 in. Unchanged -- 10% loss -- 25% loss, deep char -- >40% loss, deep ch Soil Cover/Color Unchanged -- 10% change -- 40% change -- >80% change B. HERBS, LOW SHRUBS AND TREES LESS THAN 1 METER: Pre-Fire Cover = Enhanced Growth Factor = % Foliage Altered (blk-brn) Unchanged -- 30% -- 80% 95% 100% + branch loss Frequency % Living 100% -- 90% -- 50% < 20% None Colonizers Unchanged -- Low -- Moderate High-Low Low to None Spp. Comp. - Rel. Abund. Unchanged -- Little change -- Moderate change -- High change C. TALL SHRUBS AND TREES 1 TO 5 METERS: Pre-Fire Cover = Enhanced Growth Factor = % Foliage Altered (blk-brn) 0% -- 20% % > 95% Signifcnt branch loss Frequency % Living 100% -- 90% -- 30% < 15% < 1% % Change in Cover Unchanged -- 15% -- 70% 90% 100% Spp. Comp. - Rel. Abund. Unchanged -- Little change -- Moderate change -- High Change D. INTERMEDIATE TREES (SUBCANOPY, POLE-SIZED TREES) Pre-Fire % Cover = Pre-Fire Number Living = Pre-Fire Number Dead = % Green (Unaltered) 100% -- 80% -- 40% < 10% None % Black (Torch) None % -- 60% > 85% 100% + branch loss % Brown (Scorch/Girdle) None % % < 40 or > 80% None due to torch % Canopy Mortality None -- 15% -- 60% 80% %100 Char Height None m m -- > 5 m Post Fire: %Girdled = %Felled = %Tree Mortality = E. BIG TREES (UPPER CANOPY, DOMINANT, CODOMNANT TREES) Pre-Fire % Cover = Pre-Fire Number Living = Pre-Fire Number Dead = % Green (Unaltered) 100% -- 95% -- 50% < 10% None % Black (Torch) None % -- 50% > 80% 100% + branch loss % Brown (Scorch/Girdle) None % % < 30 or > 70% None due to torch % Canopy Mortality None -- 10% -- 50% 70% %100 Char Height None m -- 4 m -- > 7 m Post Fire: %Girdled = %Felled = %Tree Mortality = Community Notes/Comments: CBI = Sum of Scores / N Rated: Sum of Scores N Rated CBI Understory (A+B+C) Overstory (D+E) Total Plot (A+B+C+D+E) = N = X = = N = X = = N = X = = N = X = = N = X = % Estimators: 20 m Plot: 314 m 2 1% = 1x3 m 5% = 3x5 m 10% = 5x6 m After, Key and Benson 1999, USGS NRMSC, Glacier Field Station. 30 m Plot: 707 m 2 1% = 1x7 m (<2x4 m) 5% = 5x7 m 10% = 7x10 m Version 3.0 May 18, 2004

11 Prefire NBR values vs. Field Measured Severity Values CBI Unchanged Low Moderate High Prefire NBR Fire severity is uncorrelated with pre-fire NBR values. This is the relationship we would like satellite measured values to emulate.

12 dnbr values vs. Prefire NBR values by Field Measured Severity dnbr Unchanged Low Moderate High Prefire NBR dnbr is correlated to pre-fire NBR values. (r = 0.53) This relationship does not lend itself to using thresholds to define severity categories.

13 Prefire NBR Postfire NBR Prefire NBR RdNBR = Prefire NBR Postfire NBR Sqrt (Prefire NBR / 1000) dnbr/prefire NBR Unchanged Low Moderate High RdNBR Unchanged Low Moderate High Prefire NBR Prefire NBR

14 Nonlinear Regression = * EXP(CBI * ) = * EXP(CBI * 0.388) R 2 = R 2 = N = 741 plots across 14 fires

15 Nonlinear Regression CBI Predicted dnbr Predicted RdNBR Unchanged Low Moderate High

16 Nonlinear Model Confusion Matrix dnbr Class Name Unchanged Low Moderate High Total User's Accuracy (%) Unchanged Low Moderate High Total Producer's Accuracy (%) Overall Kappa RdNBR Class Name Unchanged Low Moderate High Total User's Accuracy (%) Unchanged Low Moderate High Total Producer's Accuracy (%) Overall Kappa 0.421

17 2004 Power El Dorado National Forest Black Oak - Ponderosa Pine - Jeffrey Pine - Mixed Conifer - White Fir

18 Ponderosa Pine Jeffrey Pine Western Juniper 2004 Straylor Lassen National Forest

19 2004 Power Fire dnbr Unchanged Low Moderate High RdNBR Unchanged Low Moderate High PreFire NBR PreFire NBR 2004 Straylor Fire dnbr Unchanged Low Moderate High RdNBR Unchanged Low Moderate High PreFire NBR PreFire NBR Severity classes on these plots are colored using severity measured with CBI.

20 Image dates Sept 12, 2003 and Sept 1, Ignition date was July 22, 2004.

21 2002 Birch Inyo National Forest Pinyon Pine Big Sagebrush

22 These dnbr and RdNBR classifications were derived using the same thresholds as the Straylor. Image dates June 7, 2002 and June 10, Ignition date was July 1, 2002.

23

24

25

26 Conclusions Using a relative index instead of an absolute index to map fire severity has two primary advantages: 1. A relative index provides a more consistent definition of severity allowing comparison of fires across space and time. - This is a requirement for the successful analysis of landscape level processes. 2. Classification of a relative index should result in higher accuracies for the high severity category in heterogeneous landscapes over those resulting from classifying an absolute index. - This is important for site level recovery projects.

27 Conclusions (continued) While we have shown how a relative index is applicable to classifying fire severity, the methods presented here may translate to other ecological measures. There needs to be more awareness made in the remote sensing community of the implications of using absolute versus relative indices for change detection.