NG31A-01 INVITED
Joint modeling of human dwellings and the natural ecosystem at the wildland-urban interface helps mitigation of forest-fire risk
The late summer of 2007 has seen again a large number of catastrophic forest fires in the Western United States and Southern Europe. These fires arose in or spread to human habitats at the so-called wildland-urban interface (WUI). Within the conterminous United States alone, the WUI occupies just under 10 percent of the surface and contains almost 40 percent of all housing units. Recent dry spells associated with climate variability and climate change make the impact of such catastrophic fires a matter of urgency for decision makers, scientists and the general public. In order to explore the qualitative influence of the presence of houses on fire spread, we considered only uniform landscapes and fire spread as a simple percolation process, with given house densities d and vegetation flammabilities p. Wind, topography, fuel heterogeneities, firebrands and weather affect actual fire spread. The present theoretical results would therefore, need to be integrated into more detailed fire models before practical, quantitative applications of the present results. Our simple fire-spread model, along with housing and vegetation data, shows that fire-size probability distributions can be strongly modified by the density d and flammability of houses. We highlight a sharp transition zone in the parameter space of vegetation flammability p and house density d. The sharpness of this transition is related to the critical thresholds that arise in percolation theory for an infinite domain; it is their translation into our model's finite-area domain, which is a more realistic representation of actual fire landscapes. Many actual fire landscapes in the United States appear to have spreading properties close to this transition zone. Hence, and despite having neglected additional complexities, our idealized model's results indicate that more detailed models used for assessing fire risk in the WUI should integrate the density and flammability of houses in these areas. Furthermore, our results imply that fire proofing houses and their immediate surroundings within the WUI would not only reduce the houses' flammability and increase the security of the inhabitants, but also reduce fire risk for the entire landscape. http://www.environnement.ens.fr/
NG31A-02
On Earthquake Statistics: Fault and Seismicity Models, ETAS and BASS
There are two fundamentally different approaches to assessing the probabilistic risk of earthquake occurrence. The first is fault based. The statistical occurrence of earthquakes is determined for mapped faults. The applicable models are renewal models in that a tectonic loading of faults is included. The second approach is seismicity based. The risk of future earthquakes is based on the past seismicity in the region. These are also known as cluster models. An example is the epidemic type aftershock sequence (ETAS) model. In this paper we discuss an alternative branching aftershock sequence (BASS) model. In the BASS model an initial, or seed, earthquake is specified. The subsequent earthquakes are obtained from statistical distributions of magnitude, time, and location. The magnitude scaling is based on a combination of the Gutenberg-Richter scaling relation and the modified Båth's law for the scaling relation of aftershocks relative to the magnitude of the seed earthquake. Omori's law specifies the distribution of earthquake times, and a modified form of Omori's law specifies the distribution of earthquake locations. Unlike the ETAS model, the BASS model is fully self-similar, and is not sensitive to the low magnitude cutoff.
NG31A-03
Statistical Analysis of Seismicity in the Sumatra Region
We examine the effect of the great M=9.0 Boxing day 2004 earthquake on the statistics of seismicity in the Sumatra region by dividing data from the NEIC catalogue into two time windows before and after the earthquake. First we determine a completeness threshold of magnitude 4.5 for the whole dataset from the stability of the maximum likelihood b-value with respect to changes in the threshold. The split data sets have similar statistical sampling, with 2563 events before and 3701 after the event. Temporal clustering is first quantified broadly by the fractal dimension of the time series to be respectively 0.137, 0.259 and 0.222 before, after and for the whole dataset, compared to a Poisson null hypothesis of 0, indicating a significant increase in temporal clustering after the event associated with aftershocks. To quantify this further we apply the Epidemic Type Aftershock Sequence (ETAS) model. The background random seismicity rate £g and the coefficient ƒÑ, a measure of an efficiency of a magnitude of an earthquake in generating its aftershocks, do not change significantly when averaged over the two time periods. In contrast the amplitude A of aftershock generation changes by a factor 4 or so, and there is a small but statistically significant increase in the Omori decay exponent p, indicating a faster decay rate of the aftershocks after the Sumatra earthquake. The ETAS model parameters are calculated for different magnitude threshold (i.e. 4.5, 5.0, 5.5) with similar results for the different magnitude thresholds. The ƒÑ values increases from near 1 to near 1.5, possibly reflecting known changes in the scaling exponent between scalar moment and magnitude with increasing magnitude. A simple relation of magnitude and span of aftershock activity indicates that detectable aftershock activity of the Sumatra earthquake may last up to 8.7 years. Earthquakes are predominantly in the depth range 30-40 km before 20-30 km after the mainshock, compared to a CMT centroid depth of the earthquake of 28.6 km.
NG31A-04
Aftershock identification problem via the nearest-neighbor analysis for marked point processes
The centennial observations on the world seismicity have revealed a wide variety of clustering phenomena that unfold in the space-time-energy domain and provide most reliable information about the earthquake dynamics. However, there is neither a unifying theory nor a convenient statistical apparatus that would naturally account for the different types of seismic clustering. In this talk we present a theoretical framework for nearest-neighbor analysis of marked processes and obtain new results on hierarchical approach to studying seismic clustering introduced by Baiesi and Paczuski (2004). Recall that under this approach one defines an asymmetric distance D in space-time-energy domain such that the nearest-neighbor spanning graph with respect to D becomes a time- oriented tree. We demonstrate how this approach can be used to detect earthquake clustering. We apply our analysis to the observed seismicity of California and synthetic catalogs from ETAS model and show that the earthquake clustering part is statistically different from the homogeneous part. This finding may serve as a basis for an objective aftershock identification procedure.
NG31A-05
On Long-Term Correlations in Climate and the Clustering of Extreme Events
This talk reviews our recent results [1-3] on the statistics of return intervals (a) in long-term correlated (monofractal) records where the autocorrelation function decreases by a power law and (b) in multifractal data sets where the linear autocorrelation function vanishes and only non-linear correlations are present. Both monofractal and multifractal data sets play an important role in the geosciences. I will show (i) how the long-term correlations affect the statistics of the return intervals, (ii) which role the non-linear correlations in the multifractal data sets play, and (iii) how both kinds of correlations can lead to a considerably improved risk estimation. [1] A. Bunde, J. F. Eichner, J.W. Kantelhardt, S. Havlin, Phys. Rev. Lett. 94, 048701 (2005) [2] J. F. Eichner, J.W. Kantelhardt, A. Bunde, S. Havlin, Phys. Rev. E 75, 011128 (2007) [3] M. Bogachev, J. F. Eichner, A. Bunde, preprint (2007)
NG31A-06
Space Weather Forecasting and Risk Assessment
Space weather hazards are driven mainly by the turbulent solar wind, which is monitored continuously by spacecraft at the Lagrange L1 point. The correlated database of the solar wind \– magnetosphere system has been used to develop forecasting tools based on nonlinear dynamical approaches. The magnetosphere however is inherently multiscale in nature and dynamical or deterministic forecasts are possible only in a mean field sense. The deviations from the deterministic forecasts are due to the multiscale features and can be forecast only in a statistical manner. The database has been used to compute conditional probabilities following a Bayesian approach. This combination of the deterministic and probabilistic techniques yields space weather forecasts and quantitative assessments of the risks. These techniques have been applied to the forecasting of substorms (Ukhorskiy et al., GRL, 2004; Chen et al., JGR, 2006) and relativistic electron intensity in the radiation belt (Ukhorskiy et al., GRL, 2004). The multiscale property underlying the distribution of extreme events are analyzed using the burst and waiting time distributions, which are combinations of a power law with an exponential cut-off and a log-normal function (Freeman et al., GRL, 2000). The distribution of scales in the magnetosphere deviates from the power law and shows a scaling in the waiting-time distribution with respect to the mean waiting-time, characteristic of long-term correlations (Bunde et al.,PRL, 2005). However this scaling has a dependence on the threshold, leading to a rescaled distribution that deviates from a stretched exponential. The implications of these features to risk assessment analysis will be presented.
NG31A-07
Changes in Spatial Distribution of North Atlantic Tropical Cyclones
Although considerable attention has been paid to basinwide trends in North Atlantic tropical cyclone activity, considerably less attention has been given to the quantification of the changing spatial distribution of various metrics of activity . We show that the increase in observed North Atlantic cyclone activity has take place almost entirely in the middle of the Atlantic Ocean, with no trend in the western Atlantic. The lack of trend in the west Atlantic is consistent with the lack of trend in landfall statistics and measures of normalized damage. The increase in observed activity in the middle of Atlantic Ocean may be due to improved observing practices (Landsea 2006, 2007), to increased east Atlantic SST (Emanuel 2005), other physical mechanisms Kossin and Vimont (2007), or a combination of these vrious explanation. For decision makers concerned about the consequences of hurricane landfalls, the localization of the increase to the middle Atlantic is as significant as any overall increase, possibly mitigating landfall consequences of any increased activity in overall Atlantic statistics. http://www.climateaudit.org/pdf/agu07
NG31A-08
Annual shoreline change between 2006 and 2007 of the Outer Banks, North Carolina
Shorelines undergo continuous change, primarily in response to the action of waves. New technologies are enabling researchers to measure shoreline behaviors over a range of spatial and temporal scales. In June 2006 and June 2007, repeat NSF-funded LIDAR surveys were collected along a 175 km stretch of the Atlantic coast of the Outer Banks, North Carolina, United States by the National Center for Airborne Laser Mapping (NCALM). Analysis of annual shoreline position change determined from these new surveys will be presented. Our analysis of past annual surveys of the Outer Banks has documented nonlinear patterns in shoreline position change and strong correlations with preexisting beach width and annual dune retreat. Our work also documented spatial variability, where different regions of the Outer Banks exhibit different patterns of coastal change during the same time interval. The different regions have different shoreline orientations which correspond to different effective wave climates. The shoreline orientation has been shown to successfully model the spatial variations in observed shoreline change. The new 2006 and 2007 LIDAR surveys, combined with annual change determined from analysis of previous surveys of the same coastline segments collected by NASA in 1997, 1998, 1999, and 2000, will allow the analysis of patterns of shoreline change over time intervals ranging from one to ten years.