Appendix E. Fire occurrence estimates obtained from CGCM3 simulations.
|
|
| FIG. E1. (a), (c) and (e): Fire occurrence (n large fire events) as predicted from tree-ring based drought reconstructions (period 1769–1998) and simulated data from the CGCM3 model (periods 2047–2065 and 2082–2100). (b), (d) and (f): Occurrence rate of extreme fire years. The extreme events were detected using median smoothing (median smoothing parameter k = 100 and a distance to the median threshold of z = 2.5; see Mudelsee 2006 for details). The extreme events were further analyzed using a Gaussian kernel, a bandwidth of h = 15 years and bootstrap simulations (nsim = 10,000). The bandwidth was determined using a cross-validation criterion (Mudelsee et al. 2004: Eq. 5 therein). This yielded occurrence rate (bold line) and 90% confidence bands (shaded area). A sensitivity analysis, in which the parameters k, z, and h were varied around the used values, confirmed the robustness of the results. |
LITERATURE CITED
Mudelsee, M., M. Börngen, G. Tetzlaff, and U. Grünewald. 2004. Extreme floods in central Europe over the past 500 years: Role of cyclone pathway “Zugstrasse Vb”. Journal of Geophysical Research 109(D23), D23101 doi: 10.1029/2004JD005034.
Mudelsee, M. 2006. CLIM-X-DETECT: A Fortran 90 program for robust detection of extremes against a time-dependent background in climate records. Computers and Geosciences 32:141–144.