Examining the Recent Climate through the Lens of Ecology: Inferences from Temporal Pattern Analysis. Paul Hessburg & Ellen Kuhlmann, USDA Forest Service Pacific Northwest Research Station, Wenatchee, WA, and Thomas Swetnam, University of Arizona, Tree-Ring Research Laboratory The climate of a region exerts top-down control on regional ecosystem patterns and processes across space and through time Knowing this climatologists want to program their GCC models to realistically grow & burn regional landscapes under varying climatic regimes This has important ramifications for carbon budgeting in future climate modeling Ecologists and climatologists now want to know how fire regimes, sites, tree growth, and regeneration will vary with the climate One might begin by correlating past climate anomalies with natural records of those patterns or processes. For example, information useful to interpreting fire frequency, severity, and spatial extent is recorded in fire scars of living and dead trees. Other useful information is recorded in the regenerated cohorts that fill in the landscape between the recorder trees. 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 1650 1625 1600 (h) (1680-96) (1712-34) (1696-1712) (1734-58) (1798-1816) (1816-40) (1758-98) (1840-1923) (1923-46) (1946-77) (1977- ) (g) (1600s) (1630s) (1640s-60s) (1680s-90s) (1660s-80s) (1690s-1710s) (1770s) (1680s-90s) (1700s-20s) (1820s) (1810s) (1770s-1820s) (1720s-70s) (1850s) (1820s-70s) (1870s-1920s) (1870s-80s) (1880s-1920s) (1920s-30s) (1920s-40s) (f) (1980s-90s) (1940s-70s) (e) (1646-68) (1777-88) (1688-1738) (1739-44) (1756-60) (1745-55) (1761-76) (1682-87) (1669-81) (1800-38) (1839-53) (1854-69) (1870-93) (1894-1916) (1917-36) (1795-99) (1789-94) (d) (1745-69) (1785-1809) (1885-1909) (1825-49) (1920s-30s) (c) (1922-31) (1977) (b) (1600-50) (1650-90) (1700-60) (1860-90) (1920-79) (a) 1975 1950 (1919-40) 1925 1875 1850 1825 1800 1775 1750 1725 1700 1675 1650 1900 (1900-19) approximate starting/ending of cool or wet phase cool or wet phase aproximate starting/ending of warm or dry phase warm or dry phase 1625 1600 - Available studies… 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 1650 1625 1600 1) Lacked pre-defined climatic regions 2) Did not identify climatic regions that coherent with respect to the signals they were interested in 3) Lacked specific start and end dates for climatic periods, and 4) Lacked information on the unique features of each climatic regime (h) (1680-96) (1712-34) (1696-1712) (1734-58) (1798-1816) (1816-40) (1758-98) (1840-1923) (1923-46) (1946-77) (1977- ) (g) (1600s) (1630s) (1640s-60s) (1680s-90s) (1660s-80s) (1690s-1710s) (1770s) (1680s-90s) (1700s-20s) (1820s) (1810s) (1770s-1820s) (1720s-70s) (1850s) (1820s-70s) (1870s-1920s) (1870s-80s) (1880s-1920s) (1920s-30s) (1920s-40s) (f) (1980s-90s) (1940s-70s) (e) (1646-68) (1777-88) (1688-1738) (1739-44) (1756-60) (1745-55) (1761-76) (1682-87) (1669-81) (1800-38) (1839-53) (1854-69) (1870-93) (1894-1916) (1917-36) (1795-99) (1789-94) (d) (1745-69) (1785-1809) (1885-1909) (1825-49) (1920s-30s) (c) (1922-31) (1977) (b) (1600-50) (1650-90) (1700-60) (1860-90) (1920-79) (a) 1975 1950 (1919-40) 1925 1875 1850 1825 1800 1775 1750 1725 1700 1675 1650 1900 (1900-19) approximate starting/ending of cool or wet phase cool or wet phase aproximate starting/ending of warm or dry phase warm or dry phase 1625 1600 - Evaluating Temporal Patterns of Drought: 1) Pros and cons of PCA and FA PCA and FA are helpful for extracting variance held in common among a set of related sites. These approaches assume that climatic signals will be expressed in common modes of variance or in shared variance. But what if the signals are not just monotonic but also mixed and based on amplitude differences? Are there more appropriate multivariate techniques for finding these? Evaluating Temporal Patterns of Drought: 1) Pros and cons of PCA and FA 2) Why correspondence analysis? Correspondence analysis, also an eigenanalysis method, is better suited to nonlinear and nonmonotonic variables. We used correspondence analysis as implemented in TWINSPAN because it combined ordination with a classification function, allowing direct group formation. Flow of Analysis (1) Obtain modeled grid point PDSI reconstructions for the CONUS (Cook et al.1999, Cook 2000) (2) Regionalize PDSI grid points into climatic regions (3) Establish a Northwestern United States (NW) climatic region, including grid points 1-3, 8-10, and 16-18 (Fig. 2) (4) Convert the PDSI values for the 9 grid point time series in the one matrix to presence/absence values corresponding with 4 defined severity ranges, making 4 matrices from the one (Figs. 4 and 5) MildExtreme ModerateExtreme SevereExtreme Extreme PDSI ≤ -1 -1 to 1 ≥1 PDSI ≤ -2 -2 to 2 ≥2 PDSI ≤ -3 -3 to 3 ≥3 PDSI ≤ -4 -4 to 4 ≥4 Grid Points from other climate zones STOP Combine the 9 grid point time series for the NW into a single matrix, with years as rows and annual PDSI values as columns (1) Obtain modeled grid point PDSI reconstructions for the CONUS (Cook et al.1999, Cook 2000) (11) Perform ARIMA analysis on the timelines derived by PCA, FA, and TWINSPAN; using only the 20thcentury portion of the timelines, test the shifts identified by each method as intervention variables on the PDO time series of Mantua et al. (1997, Table 1). STOP (5) Convert the presence/absence values to proportions of 1, 2, and 3 decade forward moving bins (Fig. 6) 1d 1d 1d 1d 2d 2d 2d Mod Mild Sev Ext2d 3d 3d 3d Mod Mild Sev Ext3d Mod Mild Sev Ext (6) Conduct TWINSPAN analysis on the 12 resulting matrices, considering 4 dry/wet severity levels and 3 bin sizes (Fig. 6) (7) Evaluate the robustness of the 12 sets of TWINSPAN groupings via discriminant analysis with 4-fold cross-validation. (2) Regionalize PDSI grid points into climatic regions (10) Conduct PCA and FA using the same matrices analyzed by TWINSPAN in step 6; compare the PCA, FA,and TWINSPAN results (Fig. 8) (9) Characterize differences among the various regimes, especially among the anomalies (Fig. 9) (3) Establish a Northwestern United States (NW) climatic region, including grid points 1-3, 8-10, and 16-18 (Fig. 2) (8) Plot the 12 DISCRIM validated groupings of years to a timeline (Fig. 7) and estimate the most appropriate start/stop dates of each period using TWINSPAN Signal Strength (TwSS) and TWINSPAN Intervention Analysis (TwIN) analysis methods. A BB C C1 NWT C2 NCA C3 ENI GP1 GP2 GP3 GP8 GP9 GP10 GP16 GP17 GP18 GP4 GP5 GP6 GP11 GP12 GP13 GP19 GP20 GP21 GP27 GP28 GP29 GP38 C4 WMT GP25 GP26 GP34 GP35 C5 EMT GP36 GP44 GP45 GP46 GP54 GP55 GP56 GP65 C6 NSD GP64 GP66 GP67 GP75 GP76 GP77 GP78 GP79 Flow of Analysis (1) Obtain modeled grid point PDSI reconstructions for the CONUS (Cook et al.1999, Cook 2000) (2) Regionalize PDSI grid points into climatic regions (3) Establish a Northwestern United States (NW) climatic region, including grid points 1-3, 8-10, and 16-18 (Fig. 2) (4) Convert the PDSI values for the 9 grid point time series in the one matrix to presence/absence values corresponding with 4 defined severity ranges, making 4 matrices from the one (Figs. 4 and 5) MildExtreme ModerateExtreme SevereExtreme Extreme PDSI ≤ -1 -1 to 1 ≥1 PDSI ≤ -2 -2 to 2 ≥2 PDSI ≤ -3 -3 to 3 ≥3 PDSI ≤ -4 -4 to 4 ≥4 Grid Points from other climate zones STOP Combine the 9 grid point time series for the NW into a single matrix, with years as rows and annual PDSI values as columns (11) Perform ARIMA analysis on the timelines derived by PCA, FA, and TWINSPAN; using only the 20thcentury portion of the timelines, test the shifts identified by each method as intervention variables on the PDO time series of Mantua et al. (1997, Table 1). STOP (5) Convert the presence/absence values to proportions of 1, 2, and 3 decade forward moving bins (Fig. 6) 1d 1d 1d 1d 2d 2d 2d Mod Mild Sev Ext2d 3d 3d 3d Mod Mild Sev Ext3d Mod Mild Sev Ext (6) Conduct TWINSPAN analysis on the 12 resulting matrices, considering 4 dry/wet severity levels and 3 bin sizes (Fig. 6) (7) Evaluate the robustness of the 12 sets of TWINSPAN groupings via discriminant analysis with 4-fold cross-validation. (10) Conduct PCA and FA using the same matrices analyzed by TWINSPAN in step 6; compare the PCA, FA,and TWINSPAN results (Fig. 8) (9) Characterize differences among the various regimes, especially among the anomalies (Fig. 9) (8) Plot the 12 DISCRIM validated groupings of years to a timeline (Fig. 7) and estimate the most appropriate start/stop dates of each period using TWINSPAN Signal Strength (TwSS) and TWINSPAN Intervention Analysis (TwIN) analysis methods. Combine the 9 grid point time series for the NW into a single matrix, with years as rows and annual PDSI values as columns YEAR 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 . . . GP1 0.4 2.1 -1.8 -0.1 -1.3 0.3 -0.4 -1.2 -0.3 1.2 0.4 0.5 -2.2 1.4 0.0 0.4 . . . GP2 1.3 1.7 -2.3 0.3 -0.8 0.2 0.7 -1.3 -0.1 0.5 0.9 0.3 -1.2 0.5 0.2 0.0 . . . PDSI Value GP3 -0.5 0.9 -1.6 -0.5 -1.3 0.4 -0.4 -0.4 -0.1 1.3 0.4 -0.8 -0.9 1.8 0.5 0.2 . . . GP8 1.0 -0.7 -2.9 2.4 -1.1 -0.5 3.4 -2.9 3.0 -2.9 1.4 -0.2 1.0 -2.6 0.9 -0.3 . . . GP9 0.4 -0.9 -2.0 0.5 -0.8 -1.1 2.1 -0.5 1.0 0.7 0.9 -1.0 2.3 -1.4 2.3 0.3 . . . … … … … … … … … … … … … … … … … … Flow of Analysis (1) Obtain modeled grid point PDSI reconstructions for the CONUS (Cook et al.1999, Cook 2000) (2) Regionalize PDSI grid points into climatic regions (3) Establish a Northwestern United States (NW) climatic region, including grid points 1-3, 8-10, and 16-18 (Fig. 2) (4) Convert the PDSI values for the 9 grid point time series in the one matrix to presence/absence values corresponding with 4 defined severity ranges, making 4 matrices from the one (Figs. 4 and 5) MildExtreme ModerateExtreme SevereExtreme Extreme PDSI ≤ -1 -1 to 1 ≥1 PDSI ≤ -2 -2 to 2 ≥2 PDSI ≤ -3 -3 to 3 ≥3 PDSI ≤ -4 -4 to 4 ≥4 Grid Points from other climate zones STOP Combine the 9 grid point time series for the NW into a single matrix, with years as rows and annual PDSI values as columns (11) Perform ARIMA analysis on the timelines derived by PCA, FA, and TWINSPAN; using only the 20thcentury portion of the timelines, test the shifts identified by each method as intervention variables on the PDO time series of Mantua et al. (1997, Table 1). STOP (5) Convert the presence/absence values to proportions of 1, 2, and 3 decade forward moving bins (Fig. 6) 1d 1d 1d 1d 2d 2d 2d Mod Mild Sev Ext2d 3d 3d 3d Mod Mild Sev Ext3d Mod Mild Sev Ext (6) Conduct TWINSPAN analysis on the 12 resulting matrices, considering 4 dry/wet severity levels and 3 bin sizes (Fig. 6) (7) Evaluate the robustness of the 12 sets of TWINSPAN groupings via discriminant analysis with 4-fold cross-validation. (10) Conduct PCA and FA using the same matrices analyzed by TWINSPAN in step 6; compare the PCA, FA,and TWINSPAN results (Fig. 8) (9) Characterize differences among the various regimes, especially among the anomalies (Fig. 9) (8) Plot the 12 DISCRIM validated groupings of years to a timeline (Fig. 7) and estimate the most appropriate start/stop dates of each period using TWINSPAN Signal Strength (TwSS) and TWINSPAN Intervention Analysis (TwIN) analysis methods. (4) Convert the PDSI values for the 9 grid point time series in the one matrix to presence/absence values corresponding with 4 defined severity ranges, making 4 matrices from the one (Figs. 4 and 5) MildExtreme ModerateExtreme SevereExtreme Extreme PDSI ≤ -1 -1 to 1 ≥1 PDSI ≤ -2 -2 to 2 ≥2 PDSI ≤ -3 -3 to 3 ≥3 PDSI ≤ -4 -4 to 4 ≥4 (5) Convert the presence/absence values to proportions of 1, 2, and 3 decade forward moving bins (Fig. 6) 1d 1d 2d 2d Mod Mild 3d 3d Mod Mild Mod Mild 1d 2d Sev 3d Sev Sev 1d Ext2d Ext3d Ext Value of the Palmer Drought Severity Index (PDSI) Dry/Wet Severity -1 Temperate Portion 0 Mild-Extreme -6 Dry Portion -5 -4 -3 -2 1 Wet Portion 2 3 4 5 6 Moderate- Extreme -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Severe-Extreme -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Extreme -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Flow of Analysis (1) Obtain modeled grid point PDSI reconstructions for the CONUS (Cook et al.1999, Cook 2000) (2) Regionalize PDSI grid points into climatic regions (3) Establish a Northwestern United States (NW) climatic region, including grid points 1-3, 8-10, and 16-18 (Fig. 2) (4) Convert the PDSI values for the 9 grid point time series in the one matrix to presence/absence values corresponding with 4 defined severity ranges, making 4 matrices from the one (Figs. 4 and 5) MildExtreme ModerateExtreme SevereExtreme Extreme PDSI ≤ -1 -1 to 1 ≥1 PDSI ≤ -2 -2 to 2 ≥2 PDSI ≤ -3 -3 to 3 ≥3 PDSI ≤ -4 -4 to 4 ≥4 Grid Points from other climate zones STOP Combine the 9 grid point time series for the NW into a single matrix, with years as rows and annual PDSI values as columns (11) Perform ARIMA analysis on the timelines derived by PCA, FA, and TWINSPAN; using only the 20thcentury portion of the timelines, test the shifts identified by each method as intervention variables on the PDO time series of Mantua et al. (1997, Table 1). STOP (5) Convert the presence/absence values to proportions of 1, 2, and 3 decade forward moving bins (Fig. 6) 1d 1d 1d 1d 2d Mod 2d 2d Mild Sev Ext2d 3d Mod 3d 3d Mild Sev Ext3d Mod Mild Sev Ext (6) Conduct TWINSPAN analysis on the 12 resulting matrices, considering 4 dry/wet severity levels and 3 bin sizes (Fig. 6) (7) Evaluate the robustness of the 12 sets of TWINSPAN groupings via discriminant analysis with 4-fold cross-validation. (10) Conduct PCA and FA using the same matrices analyzed by TWINSPAN in step 6; compare the PCA, FA,and TWINSPAN results (Fig. 8) (9) Characterize differences among the various regimes, especially among the anomalies (Fig. 9) (8) Plot the 12 DISCRIM validated groupings of years to a timeline (Fig. 7) and estimate the most appropriate start/stop dates of each period using TWINSPAN Signal Strength (TwSS) and TWINSPAN Intervention Analysis (TwIN) analysis methods. (6) Conduct TWINSPAN analysis on the 12 resulting matrices, considering 4 dry/wet severity levels and 3 bin sizes (Fig. 6) (7) Evaluate the robustness of the 12 sets of TWINSPAN groupings via discriminant analysis with 4-fold cross-validation. (9) Characterize differences among the various regimes, especially among the anomalies (Fig. 9) (8) Plot the 12 DISCRIM validated groupings of years to a timeline (Fig. 7) and estimate the most appropriate start/stop dates of each period using TWINSPAN Signal Strength (TwSS) and TWINSPAN Intervention Analysis (TwIN) analysis methods. Flow of Analysis (1) Obtain modeled grid point PDSI reconstructions for the CONUS (Cook et al.1999, Cook 2000) (2) Regionalize PDSI grid points into climatic regions (3) Establish a Northwestern United States (NW) climatic region, including grid points 1-3, 8-10, and 16-18 (Fig. 2) (4) Convert the PDSI values for the 9 grid point time series in the one matrix to presence/absence values corresponding with 4 defined severity ranges, making 4 matrices from the one (Figs. 4 and 5) MildExtreme ModerateExtreme SevereExtreme Extreme PDSI ≤ -1 -1 to 1 ≥1 PDSI ≤ -2 -2 to 2 ≥2 PDSI ≤ -3 -3 to 3 ≥3 PDSI ≤ -4 -4 to 4 ≥4 Grid Points from other climate zones STOP Combine the 9 grid point time series for the NW into a single matrix, with years as rows and annual PDSI values as columns (11) Perform ARIMA analysis on the timelines derived by PCA, FA, and TWINSPAN; using only the 20thcentury portion of the timelines, test the shifts identified by each method as intervention variables on the PDO time series of Mantua et al. (1997, Table 1). STOP (5) Convert the presence/absence values to proportions of 1, 2, and 3 decade forward moving bins (Fig. 6) 1d 1d 1d 1d 2d 2d 2d Mod Mild Sev Ext2d 3d 3d 3d Mod Mild Sev Ext3d Mod Mild Sev Ext (6) Conduct TWINSPAN analysis on the 12 resulting matrices, considering 4 dry/wet severity levels and 3 bin sizes (Fig. 6) (7) Evaluate the robustness of the 12 sets of TWINSPAN groupings via discriminant analysis with 4-fold cross-validation. (10) Conduct PCA and FA using the same matrices analyzed by TWINSPAN in step 6; compare the PCA, FA,and TWINSPAN results (Fig. 8) (9) Characterize differences among the various regimes, especially among the anomalies (Fig. 9) (11) Perform ARIMA analysis on the timelines derived by PCA, FA, and TWINSPAN; using only the 20thcentury portion of the timelines, test the shifts identified by each method as intervention variables on the PDO time series of Mantua et al. (1997, Table 1). STOP (8) Plot the 12 DISCRIM validated groupings of years to a timeline (Fig. 7) and estimate the most appropriate start/stop dates of each period using TWINSPAN Signal Strength (TwSS) and TWINSPAN Intervention Analysis (TwIN) analysis methods. (10) Conduct PCA and FA using the same matrices analyzed by TWINSPAN in step 6; compare the PCA, FA,and TWINSPAN results (Fig. 8) Results High/Mixed Moderate/Mixed Low/Dry PACIFIC 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 Twinspan Signal Strength Analysis (1675-1714) (1715-30) (1739-48) (1756-65) (1766-1921 ) (1922-43) (1944-72) (1973-) (1675-1714) (1715-30) (1731-55) (1756-65) (1766-1925) (1926-43) (1944-76) (1977-) Twinspan Intervention Analysis Factor Analysis 1716 1742 1801 1742 1801 1852 1917 1940 1917 1940 High/Mixed Moderate/Mixed Low/Dry PACIFIC 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 Principal Components Analysis tEXT/DW t MOD/DW 1736-48 1756-65 1715-30 tLOW/D tTEMP 1922-43 1977- Moderate/Mixed Low/Dry PACIFIC 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 High/Mixed High/Mixed -6 Pacific 4 PDSI 2 0 -2 -4 Low/Dry -6 Moderate/Mixed Mean Dry/Wet Percent Composition 80 70 * 60 50 40 30 * - PDSI 20 10 +PDSI 0 Low/Dry Low/D Mod/Mixed Mod/DW High/Mixed Ext/DW PACIFIC Temp PDSI – Tree rings 1900–1921 Pacific 1922–1943 (PDSI – Tree Low /Dry 1930 rings) 1944–1973 Pacific 1973–1978 Moderate/Mixed 1950 1960 1980 PC1 of the PDO 4 2 0 -2 -4 1900 1910 1920 1940 1970 Year 1925† shift 1947† shift 1977†, shift ‡ 1990 2000 Table 1. Comparison of PDO phase shift dates of Mantua et. al. (1997) to the dates derived by TWINSPAN Signal Strength, PCA Intervention, and FA Intervention methods. Each set of shift dates were tested as intervention variables on the same PDO time series used in Mantua et. al. (1997). Intervention analysis methodology follows that of Box and Tiao (1976). TWINSPAN PCA Mantua et.al Signal Strength Intervention (1997) PDO phase shift date p value date p value date p value cool to warm 1925 0.017* 1922 0.003 1917 0.460 warm to cool 1947 0.000 1943 0.000 1940 0.260 cool to warm 1977 0.000 1973 0.000 FA Intervention date p value 1917 0.460 1940 0.260 *Value reported here differs from that published in Mantua et.al. (1997), however it is correct (personal communication, Stephen Hare) PDSI – Instrumental 1977–1986 High/Mixed (PDSI – Instrumental) 1924 1986–1994 Pacific 1944 1900–1924 Pacific 1924–1944 Moderate/Dry 1910 1930 1994–1995 Moderate/Dry 1986 1994 1944–1977 Pacif ic PC1 of the PDO 4 2 0 -2 -4 1900 1920 1940 1950 1960 1970 1980 1990 Year 1925† step 1947† step 1977†, ‡ step 1989‡ step 2000 Table 2. Comparison of PDO phase shift dates of Mantua et. al. (1997) to the dates derived by TWINSPAN Signal Strength analysis of the Cook (2000) instrumental PDSI time series for the northwestern US. Each set of shift dates were tested as intervention variables on the same PDO time series used in Mantua et. al. (1997). Intervention analysis methodology follows that of Box and Tiao (1976). PDO phase shift Cool to warm Warm to cool Cool to warm Warm to cool Cool to warm Mantua et. al. (1997) date p value 1925 0.017* 1947 0.000 1977 0.000 TWINSPAN Signal Strength date p value 1924 0.000 1944 0.000 1977 0.000 1986 0.021 1994 0.473 *Value reported here differs from that published in Mantua et. al. (1997), however it is correct (Stephen Hare, personal communication). Ecological Applications: Low Energy/Dry anomaly (1922-1943) corresponds well with period of strong mountain hemlock establishment (1921-1945, Woodward et al. 1995) tEXT/DW t MOD/DW 1736-48 1756-65 1715-30 tLOW/D tTEMP 1922-43 1977- 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 Ecological Applications: Low Energy/Dry anomaly (1922-1943) corresponds well with period of strong mountain hemlock establishment (1921-1945, Woodward et al. 1995) Our results align well with those of Keen (1937) who found periods of poor ponderosa pine growth from 1739-44, 1756-1760 with adjacent periods of good growth from 1745-55 and 1761-76. tEXT/DW t MOD/DW 1736-48 1756-65 1715-30 tLOW/D tTEMP 1922-43 1977- 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 Ecological Applications: Recently, we reconstructed fire severity of pre-management era (A.D. 1900) forests of 3 eco-subregions in the eastern Cascades of OR & WA using forest structural attributes. ESR5 ESR11 ESR13 (Hessburg et al. 2000) Ecological Applications (examples): We found that regardless of forest type or biophysical setting, most forests of the pre-management era were influenced by mixed severity fire (MSF) High 16% Low 12% Low 9% High 44% Mixed 47% Mixed 72% ESR 5 - E. WA Cacades High 27% Low 11% Mixed 62% Study Area - E. WA Cacades ESR 11 - E. WA Cacades High 20% Low 13% Mixed 67% ESR 13 - E. WA Cacades 100 ESR5 Low Mixed High 80 60 40 20 0 * * wd-PP/DF/GF cm-DF/GF Total 60 Percentage area 50 ESR11 40 30 20 10 0 80 70 60 50 40 30 20 10 0 wd-PP/DF/GF cm-DF/GF Total wd-PP/DF/GF cm-DF/GF Total (warm-dry) (cool moist) ESR13 Percentage area 70 60 Study Area 50 40 30 20 * * 10 0 wd-PP/DF/GF cm-DF/GF (warm-dry) (cool moist) Total 45 Eastern OR & WA Study Area MSF, wd-PP/DF/GF 40 %Patches %Area 30 25 20 15 10 Overstory canopy percent 91-100% 81-90% 71-80% 61-70% 51-60% 0 41-50% 5 31-40% Percentage 35 Ecological Applications: These results suggested that non-equilibrium rather than equilibrium fire dynamics were at work. Occurrence of equilibrium dynamics would be represented by strong dominance of low severity fires (LSF). With LSF dominating, we’d see abundant large tree and fire tolerant species dominated forest structures. High/Mixed -6 Pacific 4 PDSI 2 0 -2 -4 Low/Dry -6 Moderate/Mixed The observation of non-equilibrium fire dynamics is consistent with the recent 300 yr PNW climatic history: one dominated by a low variance but mixed background Pacific climate that is punctuated with moderate and high variance anomalies. Conclusions: TWINSPAN applied to temporal pattern analysis was useful; it provided start and end dates, and features of signals for comparison with natural records of ecological phenomena. Unexpected finding – the variance level of a climate anomaly can be a primary dimension and hence it is an important descriptor. t EXT/DW t MOD/DW 1736-48 t LOW/D 1756-65 1715-30 t TEMP 1922-43 1977- 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675 Recommendations for Future Studies: Stratify data prior to pattern analysis by climatic region Relax assumptions about the features of climate signals. Employ a variety of classification and ordination methods in pattern analysis, see more dimensions. Consider the characteristics of the background climate and how anomalies may stand out in contrast. t EXT/DW t MOD/DW 1736-48 t LOW/D 1756-65 1715-30 t TEMP 1922-43 1977- 1975 1950 1925 1900 1875 1850 1825 1800 1775 1750 1725 1700 1675