Cli t and Climate d Topographic T g hi Influences I fl on Fire Fi Regime R gi Attributes Att ib t in the Northern Cascade Range Range, g , Washington Washington, g , USA C Alina C. Ali C Cansler l 1 and d Donald D ld McKenzie M K i 2 Northern Cascade Range Burn Severity Images, Images 1984-2008 1984 2008 Climate Fire Occurrence Climate, Occurrence, and Fire Size 0.5 1.0 1.5 -2 10 8 -1 0 100000 2 2 3 2 R =0.20, P<0.05 10 000 10000 Area burned was significantly i ifi tl correlated l t d with ith J l maximum July i average t temperature t and d JJuly l minimum i i average relative l i humidity. 10 00 10 000 15 1.5 1 July RH 10 00 10 1.0 6 1 10000 100000 10 00 05 0.5 12 14 0 R =0.46, P<0.001 -3 3 -2 2 Snow Water Equivalent q Develop p a burn severityy atlas for the northern Cascade Range g based on local field data. -1 July Temperature 10000 Are ea Burned (ha) -1.5 1 5 -1.0 1 0 -0.5 0 5 0.0 00 Combined model with Julyy temperature p and snow water equivalent best explained fire occurrence 2 -3 Snow Water Equivalent Obj ti Objectives PDE* = 0.16, P<0.01 100000 -1.5 -1.0 -0.5 0.0 Fire occurrence was significantly correlated with May 1st snow water equivalent July maximum equivalent, a erage temperature, average temperat re and J l minimum July i i average relative l ti humidity. h idit 4 Num mber o of Fire es perr Yearr 12 10 8 6 2 4 Num mber o of Fire es perr Yearr 12 10 8 6 4 2 Num mber o of Fire es perr Yearr PDE* = 0.35, P<0.001 14 PDE* = 0.20, P<0.001 14 We examined the influence of annual climate and topography p g p y on fire occurrence, size, severity, y and within-fire severity yp pattern across 14,455 , square q kilometers of federallyy managed g land in the northern Cascade Range g of Washington g State,, USA. Landsat Thematic Mapper pp ((LTM)) data were used to quantify q y burn severityy for all fires greater than 10 ha (n = 125) that occurred during a 25 year period (1984 (1984-2008). 2008). Categorical burn severity images were developed from an index of burn severity (Relative differenced Normalized Burn Ratio) derived from LTM data and parameterized with data from 639 field plots plots. Fires in the northern Cascade Range respond both to local topographical controls and large large-scale scale annual climatic variation. variation Topographical complexity was positively correlated with patch density and negatively correlated with within within-fire fire spatial aggregation, aggregation indicating that the within-fire within fire severity mosaic reflects the underlying topographic complexity complexity. Fire size was positively correlated with the proportion of area burned at high severity and spatial aggregation of the high severity class. class Summer temperature was positively correlated with fire occurrence, occurrence annual area burned, burned the proportion of area burned at high severity severity, and spatial aggregation of th hi the high h severity it class. l The Th relationship l ti hi between b t climate li t drivers di and d fire fi regime i attributes tt ib t identified id tifi d in i this thi study t d adds dd nuance to t the th climate-area li t burned b d relationship l ti hi documented d t d in i previous i research. h 10 000 Ab t Abstract t -1 1 0 1 -1 1 0 1 July y Temperature p 2 Area burn was not correlated with May 1st snow water equivalent. equivalent 3 Julyy RH The dNBR is the change from prefire to postfire NBR. NBR The RdNBR normalizes the dNBR byy the initial image g reflectance to account for spatial p variation within the image. g The dNBRoffset corrects for o mismatches s a c es in phenology p e o ogy between be ee image age pairs. pa s The e RdNBR d has as been shown to work better when extrapolating to fires not included in the original classification (Miller et al. al 2009). 2009) 1.0 0 0.8 0.6 0.4 0.2 0 0.0 Prop portion a at Un ncha anged d Se everitty 1.0 0 0.8 0.6 0.4 0.2 0 0.0 1 -3 -2 1 1.0 J l Temperature July T t 2 R =0.32, P=0.006 0.8 100 ha 1000 ha 10000 ha 100000 ha 0 Annual Area Burned 100 ha 1000 ha 10000 ha 100000 ha 2 R =0.25, P=0.05 0.6 Annual Area Burned -1 0.4 Hi High h severity it spatial ti l dispersion, di i as measured by the Normalized Landscape Shape Index, was positively correlated with May 1st snow water equivalent 0 2 R =0.36, P=0.003 100 ha 1000 ha 10000 ha 100000 ha 0..2 Both indices are based on the Normalized Burn Ratio, which uses two Landsat TM bands, Band 4 (R4, 0.76–0.90μm, near-infrared) and Band 7 ((R7, 2.08–2.35μm, μ mid-infrared), ) to assess burn severity. y -1 High Se everiity NLSI N Proportion of area burned at unchanged severity was negatively correlated with July temperature, temperature and positively correlated with August relative humidity. humidity -2 Annual Area Burned J l Temperature July T t 0.25 5 Geospatial fire occurrence records from federal land management agencies were used to identify all fires over ten hectares (n = 125) that occurred in the study area during between 1984-2008. 1984 2008 Two remotely sensed indices of burn severity, severity the differenced Normalized Burn Ratio (dNBR) (Key and Benson 2006) and the Relative differenced Normalized Burn Ratio (RdNBR) (Miller and Thode 2007), 2007) were evaluated for use in this study. study -3 0.20 0 P Proportion ti off area b burned d att high hi h severity was positively correlated with July temperature. 0.1 15 B Burn S Severity ity Indices I di 0.10 0 Climate influences both burn severity and the spatial pattern: 2 R =0.27, P=0.012 100 ha 1000 ha 10000 ha 100000 ha 0.05 0 Methods Annual Area Burned 0.0 00 Quantify the relationship between climate and fire regime attributes. attributes Climate, Cli t S Severity, ity and d Within-fire Severity Pattern Quantifyy the relationship p between fire size and fire regime g attributes. P ortio Prop on off High Se everiity Core C Area a Quantify the relationship between topography and fire regime attributes attributes. Prop P portio on att Hig gh Se everrity *PDE = Proportion of deviance explained in a GLM -3 3 -2 2 -1 1 0 July Temperature 1 -1.5 15 -1.0 10 -0.5 05 00 0.0 05 0.5 10 1.0 15 1.5 July Temperature C Conclusions l i Fi ld Validation Field V lid ti and d IImage g Cl Classification ifi ti In the northern Cascade Range fire season climate (i.e. July temperature) is more important than antecedent climate (i.e. ( spring p g snowpack) p ) in p predicting g fire occurrence and area burned. Fire spatial patterns had a weak but significant relationship with both fire season climate and antecedent climate. climate Climate – Local climate observations from Remote Automated Weather Stations (RAWS) and snow telemetry station (SNOTEL) were converted to climate anomalies for analysis analysis. Contagion Index – increases with spatial aggregation aggregation. Aggregation Index – increases with spatial aggregation. 10000 0 1000 10 00 10 150 200 0 250 30 00 1000 10000 1.3 1.4 85 70 1.2 1.5 2 80 8 60 65 70 0 75 7 A k Acknowledgements l d t 55 1.1 1.2 1.3 1.4 Under expected warmer future climate, fires may not only be larger in size, but may also create a more spatially p y aggregated gg g landscape. p Future research is needed to address how large g aggregated gg g patches of high severity influence species composition, composition rates of succession, succession and other ecosystem functions functions. R =0.13, 0 13 P P=0.001 0 001 60 50 40 1.0 1.5 Topographic Surface-Area Ratio 1.0 1.1 1.2 1.3 1.4 1.5 Topographic Surface-Area Ratio 1.0 2 0.6 0..8 R =0.19,, P<0.001 0.4 10 100 1.0 Fire Size (ha) 1000 10000 100000 Fire Size (ha) 2 2 R = 0.06, P<0.01 0.6 100 1000 Fire Size (ha) 10000 100000 100 1000 Fire Size (ha) 10000 100000 Proportion p of area burned at low and unchanged g severities decreases. High severity core area increases. increases High severity spatial dispersion, as measured by the Normalized Landscape Shape Index ,decreased. 10 Proportion P ti off area b burned d att high hi h severity it increases. i Fire size was not significantly correlated with spatial aggregation when measured across all severity classes by the Contagion Index or the Aggregation Index Index. Alana Lautensleger, Lautensleger Andrew Larson, Larson Joe Restaino, Restaino Seth Cowdery, Cowdery and Whitney Albright for help in the field Jim Lutz Lutz, Research Associate, Associate School of Forest Resources Resources, University of Washington Washington, Seattle Seattle, WA Karen Kopper, Kopper Fire Ecologist Ecologist, North Cascades National Park, Park Seattle, Seattle WA North Cascades National Park Complex Fire Management Management, Marblemount, Marblemount WA Robert Norheim, GIS Analyst, School of Forest Resources, University of Washington, Seattle, WA Susan Prichard,, Research Ecologist. g Fire and Environmental Applications pp Team,, US Forest Service,, and School of Forest Resources,, Universityy of Washington, g , Seattle,, WA Stephen p Howard, USGS, EROS Data Center, Sioux Falls, SD Funding provided by US Forest Service, Pacific Northwest Research Station, through a cooperative agreement with the University of Washington, School of Forest Resources. References 0.6 0.8 0.8 8 R =0.41, P<0.001 10 Fire Size, Severity, and Withi fi Severity Within-fire S ity Pattern P tt As fire size increases: 0..2 100000 0.4 Normalized Landscape Shape Index (NLSI) – decreases with spatial aggregation aggregation, and increase with spatial dispersion dispersion. 100 0.2 Proportion of landscape made up of core area (area > 90m from edge of patch). 10 Pro oportio on off High h Sevverity Core e Area Area weighted mean patch size 1.1 T Topographic hi S Surface-Area f A Ratio R ti 2 0.0 1.0 P Proportion ti off landscape l d att high, hi h moderate, d t low, l and d unchanged h d severity. it Edge g densityy – decreases with spatial p aggregation. gg g 2 R =0.22,, P<0.001 0.0 The spatial pattern of fires fires, both for the whole landscape and within a given severity class, class was quantified using FRAGSTATS (McGarigal and Marks 1995). 1995) Specific metrics used include: 1.0 Fire climate interactions affect landscape pattern not just through changes in the annual area burned but also by influencing the severity mosaic within individual fires. burned, fires The influence of climate on fire fi can be b seen directly, di tl via i itits effect ff t on severity it pattern tt relationships, l ti hi and d iindirectly, di tl via i itits effect on fire size, which in turn affects severity patterns. Prop portio on at High Seve erity Topographical complexity - the “surface to area ratio” was calculated using digital elevation models (DEMs) data for each fire (Jenness 2004). 1.5 30 3 Topographical complexity was not significantly correlated with class level spatial pattern. Quantifying Climate, Climate Topography, Topography and Within-fire Within fire Severity Pattern 1.4 20 0 CBI Conta C agion n Inde ex Patch density increased and patch shape becomes less complex, complex probably due to an increased proportion of small simple patches across the landscape (not shown). shown) 30 3.0 1.3 R =0.09, 0 09 P P=0.001 0 001 0.4 25 2.5 1.2 2 R =0.08, P=0.001 T Topographic hi S Surface-Area f A Ratio R ti 0.2 20 2.0 Pro oportion a at Uncchang ged S Severrity 15 1.5 High Sevverityy NLS SI 10 1.0 0..8 05 0.5 0.6 00 0.0 1.1 Ag ggreg gation Ind dex 10 000 The spatial pattern of the whole fire , as measured by the Contagion Index and Aggregation Index becomes less spatially aggregated. gg g 2 R =0.10, P<0.001 1.0 Area-weighted Area weighted mean patch size increased. increased 0.4 RdNBR images were used to produce categorical images for all fires in the study. study Edge density increased. 0..2 Both indices performed similarly, but RdNBR had slightly g y higher g classification accuracyy (62% ( vs. 59%). ) As topographical complexity increases: 50 00 Both dNBR and RdNBR models were evaluated based on model fit and severity class categorization accuracy. accuracy Topography and Withi fi Severity Within-fire S ity Pattern P tt 0 DNBR and RdNBR values that would serve as thresholds between severity classes were determined. RdNB R BR 1500 1 0 1.74 Results 100 Physical measurements of overstory tree canopy scorch, scorch char height, height tree mortality, mortality and surface fuel consumption (forest plots) were translated into CBI values. values 388 physical plots were installed on the 70 000 ha Tripod fire 70,000 fire. To T validate lid t and d classify l if burn b severity it images i non-linear li RdNBR = 187.1869 126.7CBI r-squared = 0.470, P < 0.001 regression i was used d to t model d l dNBR and d RdNBR as a f function i off CBI. CBI Ed dge D Denssity The Composite Burn Index (CBI) combines ecologically significant burn severity variables into one numeric site index. CBI was employed on 251 plots over 4 fires. 0.0 0 Are ea-w weighted Mean M n Pattch Size S Two methodologies were used to assess burn severity in the field: Jenness, J. S. 2004. Calculating landscape surface area from digital elevation models. Wildlife Society Bulletin 32:829839. K Key, C C.H. H and dN N.C. C Benson. B 2006 2006. L Landscape d A Assessment: t G Ground d measure off severity, it the th Composite C it Burn B Index; I d and d R Remote t sensing i off severity, it the th Normalized N li d Burn B R Ratio. ti IIn D D.C. C L Lutes; t R R.E. E K Keane; JJ.F. F Caratti; C tti C C.H. H Key; K N.C. N C Benson; B S S. Sutherland; and L L.J. J Gangi. Gangi 2006 2006. FIREMON: Fire Effects Monitoring and Inventory System. System USDA Forest Service Service, Rocky Mountain Research Station Station, Ogden Ogden, UT UT. Gen Gen. Tech Tech. Rep Rep. RMRS RMRS-GTR-164-CD: GTR 164 CD: LA 1 1-51. 51 McGarigal K McGarigal, K., S.A. S A Cushman, Cushman M.C. M C Neel, Neel and E. E Ene. Ene 2002. 2002 FRAGSTATS: Spatial Pattern Analysis Program for Categorical Maps Maps. Computer software program produced by the authors at the University of Massachusetts, Massachusetts Amherst Amherst. Available at the following web site: http://www http://www.umass.edu/landeco/research/fragstats/fragstats.html. umass edu/landeco/research/fragstats/fragstats html Miller, J. D. and A. E. Thode. 2007. Quantifying burn severity in a heterogeneous landscape with a relative version of the delta Normalized Burn Ratio (dNBR). ( ) Remote Sensing g of Environment 109:66-80. Miller,, J. D.,, E. E. Knapp, pp, C. H. Key, y, C. N. Skinner,, C. J. Isbell,, R. M. Creasy, y, and J. W. Sherlock. 2009. Calibration and validation of the relative differenced Normalized Burn Ratio (RdNBR) ( ) to three measures of fire severityy in the Sierra Nevada and Klamath Mountains, California, USA. Remote Sensing of Environment 113:645-656. 1 School of Forest Resources, University of Washington, Seattle, WA 98195-2100. 2 Pacific Wildland Fire Sciences Lab, USDA Forest Service, 400 N 34th Street, Suite 201, Seattle, WA 98103.