FIRE AND CLIMATE - Using the past to predict the future: Figures 34.2 and 34.3 in color versions.

Figure 34.1 Examples of annual area burned time series for various countries: (a) Canada (Van Wagner, 1988, https:// cwfis.cfs.nrcan.gc.ca/ datam art; last accessed on May 23, 2023); (b) United States of America (www.nifc.gov/ fireI nfo/ fireIn fo_ s tats tot alFi res.html; last accessed on February 22, 2022); (c) Chile (https:// gfmc.onl ine/ invent ory/ cl_ sta tist ic_1 964- 2004.html; last accessed on February 28, 2022); (d) European Southern Member States (Portugal, Spain, France, Italy and Greece (SCHMUCK et al., 2013); (e) Komi Republic (Drobyshev and Niklasson, 2004); and (f) Russian Federation (Goldammer et al., 2007). Parentheses indicate the periods covered by each fire statisticFigure 34.3 Summary of the main analytical steps required for reconstructing past fire frequency from sedimentary charcoal records. (a) Raw charcoal values expressed in Charcoal Accumulation Rate (CHAR mm2.cm- 2.yr- 1) were smoothed using a locally weighted regression (LOESS) with a 500- year window width, which enabled us to discriminate the CHARpeak (b) and CHARback (c) components. CHARback (c) represented longdistance fires, charcoal redeposition and noise while the CHARpeak (b) represented local fires. Noise is also present in the remaining CHARpeak component and a Gaussian mixture model is generally used to discriminate the two CHARpeak subpopulations: CHARnoise and CHARfire. The CHARfire component represented the occurrence of one or more fires locally around lakes (<10 km). The Gaussian mixture could be applied globally to the entire records (such as in this example) or locally in order to better fit the centennial to millennial variations in CHAR. Analytical steps (CHARraw filtering technique and window width) that are able to provide well- discriminated fire events are not unique (Higuera et al., 2007) and techniques that aimed at replicating the analysis of a record using ensembles have been proposed (Blarquez et al., 2013). Therein, 1,000 replicates of the steps (a– c) are made using multiple smoothing techniques and window width, and the cumulative number of fires by the ensemble is calculated for each fire event date (d). From this ensemble of fire event dates, it is possible to assess fire frequency (or fire return intervals). In the example illustrated in (e), the tinted bands represent the distribution of the fire frequencies obtained from the 1,000 iterations of the model parameters. From these, one can be confident that the changes from low to high fire frequency ca. 3,000 cal BP are not a consequence of the selection of model parameters but, rather, reflect important changes that have occurred in burning activity (Blarquez et al., 2013).

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.