Natural Hazard Uncertainty Assessment E-Book

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: Geophysical Monograph Series
• Sprache: Englisch

Seitenzahl: 956

Veröffentlichungsjahr: 2016

Beschreibung

Uncertainties are pervasive in natural hazards, and it is crucial to develop robust and meaningful approaches to characterize and communicate uncertainties to inform modeling efforts. In this monograph we provide a broad, cross-disciplinary overview of issues relating to uncertainties faced in natural hazard and risk assessment. We introduce some basic tenets of uncertainty analysis, discuss issues related to communication and decision support, and offer numerous examples of analyses and modeling approaches that vary by context and scope. Contributors include scientists from across the full breath of the natural hazard scientific community, from those in real-time analysis of natural hazards to those in the research community from academia and government. Key themes and highlights include: * Substantial breadth and depth of analysis in terms of the types of natural hazards addressed, the disciplinary perspectives represented, and the number of studies included * Targeted, application-centered analyses with a focus on development and use of modeling techniques to address various sources of uncertainty * Emphasis on the impacts of climate change on natural hazard processes and outcomes * Recommendations for cross-disciplinary and science transfer across natural hazard sciences This volume will be an excellent resource for those interested in the current work on uncertainty classification/quantification and will document common and emergent research themes to allow all to learn from each other and build a more connected but still diverse and ever growing community of scientists. Read an interview with the editors to find out more: https://eos.org/editors-vox/reducing-uncertainty-in-hazard-prediction

Leseprobe

Table of Contents

COVER

TITLE PAGE

CONTRIBUTORS

1 Uncertainty in Natural Hazards, Modeling and Decision Support

1.1. INTRODUCTION

1.2. ORIGINS AND OBJECTIVES OF THIS VOLUME

1.3. STRUCTURE

1.4. A SYNTHESIS: LEARNING FROM THIS MONOGRAPH

1.5. CONCLUSION

ACKNOWLEDGMENTS

Part I: Uncertainty, Communication, and Decision Support

2 Natural Hazard Modeling and Uncertainty Analysis

2.1. INTRODUCTION

2.2. IDENTIFYING AND CLASSIFYING UNCERTAINTIES

2.3. GUIDANCE FOR IDENTIFYING AND CLASSIFYING UNCERTAINTIES

2.4. TECHNIQUES FOR EVALUATING UNCERTAINTY

2.5. DISCUSSION

REFERENCES

3 Understanding Uncertainty as a Key Interdisciplinary Problem in Earth System Science

ACKNOWLEDGMENTS

APPENDIX

REFERENCES

4 Uncertainty and Probability in Wildfire Management Decision Support

4.1. INTRODUCTION

4.2. WILDFIRE MANAGEMENT

4.3. PROBABILISTIC INFORMATION AND RISK‐BASED WILDFIRE DECISION SUPPORT

4.4. FUTURE DIRECTIONS FOR WILDFIRE DECISION SUPPORT

4.5. CONCLUSION

REFERENCES

5 Role of Uncertainty in Decision Support for Volcanic Ash Cloud Modeling

5.1. INTRODUCTION

5.2. CLASSIFYING THE UNCERTAINTY

5.3. QUANTIFYING THE UNCERTAINTY

5.4. MITIGATING THE UNCERTAINTY AND INCREASING CONFIDENCE

5.5. APPLICATION OF UNCERTAINTY INTO THE DECISION SUPPORT SYSTEM

5.6. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

Part II: Geological Hazards

6 Building an Uncertainty Modeling Framework for Real‐Time VATD

6.1. INTRODUCTION AND BACKGROUND

6.2. METHODOLOGY

6.3. PROBABILISTIC MODELING RESULTS

6.4. DISCUSSION

6.5. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

7 Uncertainties in Estimating Magma Source Parameters from InSAR Observation

7.1. INTRODUCTION

7.2. VOLCANO DEFORMATION FROM INSAR AND THE ASSOCIATED UNCERTAINTIES

7.3. RETRIEVAL OF MAGMA SOURCE PARAMETERS FROM INSAR AND ASSOCIATED MEASUREMENT UNCERTAINTIES

7.4. DISCUSSION ON IMPACTS OF OTHER GEOPHYSICAL ASSUMPTIONS ON THE MOGI SOURCE MODEL

7.5. CONCLUSION

ACKNOWLEDGMENTS

REFERENCES

8 Improving Model Simulations of Volcanic Emission Clouds and Assessing Model Uncertainties

8.1. INTRODUCTION

8.2. METHODS

8.3. EXAMPLE CASE STUDY: THE GRÍMSVÖTN 2011 ERUPTION

8.4. SUMMARY AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

9 Uncertainty Assessment of Pyroclastic Density Currents at Mount Vesuvius (Italy) Simulated Through the Energy Cone Model

9.1. INTRODUCTION

9.2. METHODS

9.3. RESULTS

9.4. DISCUSSION

ACKNOWLEDGMENTS

REFERENCES

10 Earthquake Loss Estimation in the Gyeongju Area, Southeastern Korea, Using a Site Classification Map

10.1. INTRODUCTION

10.2. METHODS

10.3. SEISMICITY IN THE KOREAN PENINSULA

10.4. SITE CLASSIFICATION MAP

10.5. SCENARIO EARTHQUAKE AND INVENTORY DATA FOR THE EARTHQUAKE LOSS ESTIMATION

10.6. LOSS ESTIMATION RESULTS

10.7. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

11 Implications of Different Digital Elevation Models and Preprocessing Techniques to Delineate Debris Flow Inundation Hazard Zones in El Salvador

11.1. INTRODUCTION

11.2. METHODS

11.3. RESULTS

11.4. DISCUSSION

11.5. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

12 Evaluating the Performance of FLO2D for Simulating Past Lahar Events at the Most Active Mexican Volcanoes

12.1. INTRODUCTION

12.2. METHODOLOGY

12.3. RESULTS AND DISCUSSION

12.4. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

Part III: Biophysical and Climatic Hazards

13 An Uncertainty Analysis of Wildfire Modeling

13.1. INTRODUCTION

13.2. METHODS

13.3. RESULTS

13.4. DISCUSSION

13.5. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

14 Fire and Smoke Remote Sensing and Modeling Uncertainties

14.1. INTRODUCTION

14.2. METHODS

14.3. RESULTS AND DISCUSSION

14.4. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

15 Uncertainty and Complexity Tradeoffs When Integrating Fire Spread with Hydroecological Projections

15.1. INTRODUCTION

15.2. METHODS

15.3. RESULTS

15.4. DISCUSSION

ACKNOWLEDGMENTS

REFERENCES

16 Uncertainty Quantification and Propagation for Projections of Extremes in Monthly Area Burned Under Climate Change

16.1. INTRODUCTION

16.2. DATA

16.3. METHODS

16.4. RESULTS

16.5. DISCUSSION

ACKNOWLEDGMENTS

REFERENCES

17 Simulating Vegetation Change, Carbon Cycling, and Fire Over the Western United States Using CMIP5 Climate Projections

17.1. INTRODUCTION

17.2. METHODS

17.3. RESULTS

17.4. DISCUSSION AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

18 Sensitivity of Vegetation Fires to Climate, Vegetation, and Anthropogenic Drivers in the HESFIRE Model

18.1. INTRODUCTION

18.2. METHODS

18.3. DISCUSSION

REFERENCES

19 Uncertainties in Predicting Debris Flow Hazards Following Wildfire

19.1. INTRODUCTION

19.2 BIOPHYSICAL SETTING

19.3. FIRE PROCESSES

19.4. FIRE EFFECTS

19.5. RAINFALL TRIGGERS

19.6 DEBRIS FLOW INITIATION, MOBILIZATION, AND DEPOSITION

19.7. VALUES‐AT‐RISK

19.8. CONCLUSIONS

REFERENCES

20 Capturing Spatiotemporal Variation in Wildfires for Improving Postwildfire Debris‐Flow Hazard Assessments

20.1. INTRODUCTION

20.2. METHODS

20.3. RESULTS

20.4. DISCUSSION

20.5. CONCLUSION

REFERENCES

21 Uncertainty in Estimation of Debris‐Flow Triggering Rainfall

21.1. INTRODUCTION

21.2. METHODS

21.3. RESULTS

21.4. DISCUSSION AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

22 Prospects in Landslide Prediction

22.1. INTRODUCTION

22.2 DATA

22.3 PRECIPITATION‐BASED LANDSLIDE PREDICTION

22.4 EXTENDING LANDSLIDE PREDICTION TO CLIMATE CHANGE SCENARIOS

22.5 COMPARISON OF MODEL RESULTS

22.6. DISCUSSION AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

INDEX

END USER LICENSE AGREEMENT

List of Tables

Chapter 01

Table 1.1 Emergent Themes Based on Synthesis of This Volume

Table 1.2 Selected Techniques and Methods for Handling Uncertainty in the Natural Hazards

Chapter 02

Table 2.1 Definitions and Examples of the Nature Dimension of Uncertainty

Table 2.2 Definitions and Examples of the Location Dimension of Uncertainty

Table 2.3 Definitions and Examples of the Level Dimension of Uncertainty

Table 2.4 Stylized Uncertainty Matrix Illustrating How Sources of Uncertainty Can Be Analyzed According to Each Dimension: Nature, Location, Level

Chapter 04

Table 4.1 Response Functions and Relative Importance Weights for Two Stylized HVRAs

Table 4.2 Select Set of Identified Cognitive Biases That Influence How Fire Managers Perceive and Respond to Risk

Chapter 05

Table 5.1 Inputs Needed for VATD Models to Simulate Volcanic Ash Cloud Location and Airborne Ash Concentrations After an Eruption Has Been Detected

Table 5.2 Range in Possible Mass Eruption Rate (MER) Given 0.25 km Accuracy in Plume Height Measurements

Chapter 07

Table 7.1 Mean and Standard Deviation (σ) of Estimated Mogi Source Parameters With

ϵ

u

Errors in Deformation Measurements

Table 7.2 Mean and Standard Deviation (σ) of Estimated Mogi Source Parameters When the Deformation Measurement Contains

ϵ

a

and

ϵ

l

, Respectively

Table 7.3 Biased Estimation From Flat‐Surface Assumption in Mogi Model

Table 7.4 Standard Deviation (σ) of Estimated Mogi Source Parameters With Single Track and Multitrack Deformation Maps

Chapter 09

Table 9.1 Shape Parameters of the Selected Probability Density Functions, PDFs

Table 9.2 Relative Maximum Expected Deviations From Aleatory Uncertainty Considering Area of PDC Invasion and Maximum Runout (

δ

A

,

δ

MR

) and Four Different Sources of Epistemic Uncertainty

Table 9.3 Two‐Sample Kolmogorov‐Smirnov Tests Performed by Comparing the Aleatory‐Uncertainty Output Empirical Cumulative Distribution Function (ECDF) Against Diverse Alternate Output ECDFs (Epistemic‐Uncertainty Configurations) for Two Different Output Variables (Area of Invasion and Maximum Runout) and Three Eruption Sizes at Mount Vesuvius (Italy)

Table 9.4 Conditional Probabilities of PDC Invasion (

CP

, in Percentage) on Five Selected Points Over the Surroundings of Mount Vesuvius (Italy)

Chapter 10

Table 10.1 Summary of Correlations Among the Earthquake Hazard Reduction Program (NEHRP) Site Classifications, Geological and Topographical Information, and Shear‐Wave Velocity

Chapter 11

Table 11.1 DEM

S

ources and

C

haracteristics

Table 11.2 Hydrological

C

onditioning Techniques

Table 11.3 Description of How Simulated Versus Observed Debris Inundation Zones Are Classified to Determine Overall Accuracy

Table 11.4 True Positive Rates, TPR, of Each DEM Versus Fill Method

Table 11.5 False Positive Rates, FPR, of Each DEM Versus Fill Method

Table 11.6 Overall

A

ccuracy of Each DEM Versus Fill Method

Table 11.7 Ranges of Estimated Areas of Potential Inundation Summed for All Debris Flows Tested, at a 95% Confidence Interval

Chapter 13

Table 13.1 Factors Influencing Fire Extent and Intensity Across Planning Horizons, in Terms of Uncertain Information

Table 13.2 An Uncertainty Matrix Identifying and Classifying Uncertainties in Fire Modeling

Chapter 14

Table 14.1 Selected Current or Recent Satellite Sensors Providing Observations of Fires and Smoke That Are Relevant to This Study

Table 14.2 The Uncertainty Ranges of Satellite‐Derived Fire and Smoke Variables

Table 14.3 Root Mean Square Error (RMSE) Values Between WRF‐Chem AOD Simulations and MODIS AOD Retrievals for Terra and Aqua According to Bins of 25% Coverage of MODIS AOD Retrievals Over Each Sample Box Area Shown in Figure 14.3b.

Chapter 15

Table 15.1 WMFire_Beta Parameter Values, Interpretations, and Search Ranges

Chapter 16

Table 16.1 Brier Scores and Quantile Scores for the Ecological Model Based on Fivefold Cross Validation for Three EBA Thresholds Tested Against Six Prediction Thresholds

Table 16.2 Estimated Bayesian Model Averaging Probabilities

for the Climate and Seasonal Covariates (

X

t

) and Best‐Fit Values for the

β

j

Used to Estimate the Probability of Exceeding a Threshold (Hectares Burned Month

−1

),

β

1

, and the Magnitude of the Extreme Burned Area,

β

2

.

Table 16.3 BMA Model Weights for Each Emissions Scenario for the Empirically Downscaled Members of the CMIP3 GCM Ensemble Based on Observations and Simulations From the 20C3M Experiment

Chapter 17

Table 17.1 List of the 20 CMIP5 Climate Models That Were Downscaled and Used in This Project

Table 17.2 Brief Summary of the Two RCP (Representative Concentration Pathways) Scenarios Used in This Project With the Associated Climate Change Estimates for the Western United States Averaged Across 20 CMIP5 Climate Models for the Period 2081–2100 Relative to 1986–2005

Table 17.3 Changes in Areal Extent (in %) of the Vegetation Types from 1971–2000 to 2071–2100 Using Projected Climates with Two Greenhouse Gas Concentration Trajectories

Chapter 18

Table 18.1 HESFIRE Optimized Parameters, Which Were the Ones Also Selected for the Sensitivity Experiment

Table 18.2 Impacts of Alternative Input Data on HESFIRE Outputs

Chapter 19

Table 19.1 Summary of Uncertainties Associated with the Components of the Postfire Debris‐Flow Hazard Cascade

Chapter 20

Table 20.1 Watershed Inputs for Debris‐Flow Model

Table 20.2 Attributes of Analysis Recurrence Interval Storms

Table 20.3 Wildfire Events Summary Data for Each Watershed

Table 20.4 A Comparison of Postwildfire Debris Flow Hazards for the Worst‐Case Scenario Event Versus a Best‐Case Scenario Event

Chapter 22

Table 22.1 Average Models’ Test Set Performance Using Long‐Term Weather Patterns

List of Illustrations

Chapter 02

Figure 2.1 The four primary stages of a structured decision‐making process and their relation to the four primary stages of a risk modeling process.

Figure 2.2 Conceptual overview of a three‐step process for uncertainty analysis.

Figure 2.3 Three decision trees for classifying each dimension of uncertainty, modified from

Warmink et al.

[2010].

Chapter 03

Figure 3.1 Global temperature evolution and associated uncertainties. The simulated evolution of global average surface temperature for the historical period and the projected changes are shown until the year 2100 (data from the coupled model intercomparison project CMIP5, plot from IPCC AR5, Summary for policymakers,

IPCC

[2013]). The overall spread of the projections in 2100 includes scenario (forcing) uncertainty and structural uncertainty (model differences). It is also clear from comparing the historical period 1950–2000 and the projected time period 2000–2100 that some of the variability of the real system is only implicitly accounted for in this multimodel representation.

Figure 3.2 IPCC calibrated language: robustness versus agreement. This figure is taken from the IPCC guidance note on uncertainty [

Mastrandrea

, 2010] and depicts “evidence and agreement statements and their relationship to confidence. Confidence increases towards the top‐right corner as suggested by the increasing strength of shading. Generally, evidence is most robust when there are multiple, consistent independent lines of high‐quality evidence.” This quotation and the figure show the strength and weaknesses to define a common calibrated language, as the need to use a unified approach to confidence implies common definitions for a whole set of concepts such as agreement, evidence, and robustness.

Figure 3.3 Sources of explicit uncertainty on the way from the true system to our description and understanding. Between our description and understanding of the Earth system and the true system are several layers of explicit uncertainty, shown here as known sources of error. The truth itself is unattainable because not all errors are arbitrarily reducible and because of additional implicit uncertainties. Any addition of complexity in our description, corresponding to moving outward from the center, adds more layers of uncertainty. For simple or very focused descriptions, the types of uncertainty are model‐ or observation‐specific. But with increasing complexity, model and observation errors and thus uncertainties start to overlap and become intertwined. Moving away from the present state of the Earth system along the bent time axes also adds complexity and thus uncertainty. While the qualitative and quantitative contributions might differ for past and future, the overall uncertainties may be considered symmetric in both directions.

Figure 3.4 Different sources of uncertainty in the Earth system on the way to seamless environmental prediction. This schematic table shows basic relationships between the timescale of a prediction in the Earth system and involved sources of uncertainty and corresponding, technical terms for the ensemble creation in modeling.

Chapter 04

Figure 4.1 Conceptual overview of major factors influencing wildfire risk management. Boxes in light grey represent primary management options, and boxes in dark grey represent the primary components of wildfire risk.

Figure 4.2 Wildfire risk triangle, composed of the likelihood and intensity of wildfire along with the susceptibility of resources and assets to wildfire.

Figure 4.3 Primary sources of variability in burn probability modeling and their relation to the planning context.

Figure 4.4 Example FSPro burn (a) probability contours and (b) exposure of a select set of resources and assets.

Figure 4.5 Histogram of simulated final fire sizes output from FSPro.

Figure 4.6 Tabular exposure analysis results summarizing FSPro results intersected with spatial value layers; results are presented across burn probability zones as well as in terms of expected values.

Figure 4.7 Scatterplot of fire‐level

cNVC

versus fire size.

Figure 4.8 Exceedance probability curve for fire‐level

cNVC

, expressed in monetary units.

Chapter 05

Figure 5.1 Effusive versus explosive eruptions from an active volcano. (a) Visible and thermal images of the 27 June flow from Kilauea volcano taken on 1 October 2014. The thermal image viewpoint is represented in white box in the visible imagery; (b) the Eyjafjallajökull eruption on 17 April 2010 showing a heavy black plume engulfing the neighboring district of Eyjafjoll.

Figure 5.2 (a) An example volcanic ash advisory (VAA) from the Tokyo VAAC and (b) its accompanying volcanic ash graphic (VAG). This is for Karymsky volcano, Russia, on 16 October 2014 at 18:00 UTC or Z. Note in the VAA that satellite data were used as the source for the detection of the event. The VAG provides the satellite data at +0 hours and the forecasted locations at +6, +12, and +18 hr as reported by location points in the VAA. (The VAA is from

Tokyo VAAC

[2014a] and the VAG is from

Tokyo VAAC

[2014b].)

Figure 5.3 Suite of volcanic ash transport and dispersion models used for simulations of volcanic ash clouds. For full names of the models, see within

WMO

[2013].

Figure 5.4 Volcanic ash concentration maps produced as additional information to the standard VAA and VAG. Examples are from the (a) London VAAC for Grímsvötn eruption and (b) from the Toulouse VAAC for the Cordon Caulle eruptive event, both from 2011 (Grimsvötn image is from GVP of Smithsonian [

GVP

, 2014]).

Figure 5.5 (a) Maximum plume heights from Eyjafjallajökull eruption from 14 April 2010 at 12:00—18:00 UTC over varying time windows and (b) significance of temporal repeatability of plume height measurements on mass eruption rates.

Figure 5.6 Volcanic ash‐cloud products with probability of ash occurrence. Eruption from San Miguel volcano on 29 December 2013: (a) Volcanic brightness temperature split window, (b) retrieved cloud top height, (c) retrieved volcanic ash loading per pixel, and (d) ash probability of detection.

Chapter 06

Figure 6.1 (a) London VAAC’s VAA produced during Eyjafjallajökull eruption on 14 April 2010; (b) additional concentration product from the same date at 06:00 UTC; and (c) the progression to concentration thresholds during the Grimsvotn eruption on 25 May 2011.

Figure 6.2 Probabilistic modeling workflow, adapted from

Madankan et al.

[2014], using the Puff VATD model and coupled one‐dimensional plume rise model, BENT.

Figure 6.3 Real‐time processing routines from the probabilistic modeling of volcanic ash clouds. Results from these routines include mean ash mass loadings and ash concentrations at defined altitudes from all 161 ensemble members.

Figure 6.4 Cleveland Volcano Puff VATD model simulation for Probabilistic Simulation number 1. This is for start time on 3 December 2014 at 00:00 UTC with the particle locations and ash mass loading (mg/m

2

) at + 12 hr after the eruption start, or 12:00 UTC.

Figure 6.5 Probabilistic modeling outputs at 12:00 UTC, + 12 hr after eruption, for Cleveland 3 December 2014 model simulation. (a) Mean of the 161 simulation members showing ash mass loadings (mg/m

2

) and (b) probabilities (%) of ash mass loading exceeding predefined threshold.

Figure 6.6 Probabilistic modeling outputs at 12:00 UTC, + 12 hr after eruption, for Cleveland 3 December 2014 model simulation. Mean results from the 161 simulation members showing ash concentrations (mg/m

3

) at 2 (a), 10 (b), and 16 (c) km ASL.

Figure 6.7 Probabilistic modeling outputs at 12:00 UTC, + 12 hr after eruption, for the Cleveland 3 December 2014 model simulation. Probabilities (%) of ash concentration (mg/m

3

) exceedances at 2 (a), 10 (b), and 16 (c) km ASL.

Figure 6.8 Zhupanovsky Volcano Puff VATD model simulation for probabilistic run number 1. This is for start time on 29 December 2014 at 00:00 UTC with the particle locations and ash mass loading (mg/m

2

) at + 12 hr after the eruption start, or 12:00 UTC.

Figure 6.9 Probabilistic modeling outputs at 12:00 UTC, + 12 hr after eruption, for Zhupanovsky 29 December 2014 model simulation. (a) Mean of the 161 simulation members showing ash mass loadings (mg/m

2

) and (b) probabilities (%) of ash mass loading exceeding predefined threshold.

Figure 6.10 Probabilistic modeling outputs at 12:00 UTC, + 12 hr after eruption, for Zhupanovsky 29 December 2014 model simulation. Mean results from the 161 simulation members showing ash concentrations (mg/m

3

) at 2 (a), 10 (b), and 16 (c) km ASL.

Figure 6.11 Probabilistic modeling outputs at 12:00 UTC, + 12 hr after eruption, for Zhupanovsky 29 December 2014 model simulation. Probabilities (%) of ash concentration (mg/m

3

) exceedances at 2 (a), 10 (b), and 16 (c) km ASL.

Figure 6.12 Probabilities (%) of volcanic ash mass loading (mg/m

2

) exceeding a range of specific thresholds in the simulations for Zhupanovsky volcano, 12:00 UTC 29 December 2014. (a) 100 mg/m

3

, (b) 10 mg/m

3

, (c) 1 mg/m

3

, (d) 0.1 mg/m

3

, (e) 0.01 mg/m

3

, and (f) 0.001 mg/m

3

or 1 μ mg/m

3

.

Figure 6.13 Probabilities (%) of volcanic ash concentration (mg/m

3

) occurrence at 2 km ASL exceeding a range of specific thresholds in the simulations for Zhupanovsky volcano, 12:00 UTC 29 December 2014, when concentration threshold set at (a) 100 mg/m

3

, (b) 10 mg/m

3

, (c) 1 mg/m

3

, (d) 0.1 mg/m

3

, (e) 0.01 mg/m

3

, and (f) 0.001 mg/m

3

or 1 μ mg/m

3

.

Figure 6.14 Probabilities (%) of volcanic ash concentration (mg/m

3

) occurrence at 10 km ASL exceeding a range of specific thresholds in the simulations for Zhupanovsky volcano, 12:00 UTC 29 December 2014, when concentration threshold set at (a) 100 mg/m

3

, (b) 10 mg/m

3

, (c) 1 mg/m

3

, (d) 0.1 mg/m

3

, (e) 0.01 mg/m

3

, and (f) 0.001 mg/m

3

or 1 μ mg/m

3

.

Figure 6.15 Probabilities (%) at 2 km ASL from the simulations for Zhupanovsky volcano, 29 December 2014, when concentrations exceeding at 1 mg/m

3

, (a) 02:00, (b) 04:00, (c) 06:00, (d) 08:00, (e) 10:00, and (f) 12:00.

Figure 6.16 Probabilities (%) at 2 km ASL from the simulations for Zhupanovsky volcano, 29 December 2014, when concentrations exceeding at 0.001 mg/m

3

, (a) 02:00, (b) 04:00, (c) 06:00, (d) 08:00, (e) 10:00, and (f) 12:00.

Figure 6.17 Puff volcanic ash cloud simulations #51 and #160, for Zhupanovsky volcano on 29 December 2014 at 12:00 UTC showing Puff particle locations ([a] for #51; [d] for #160), mass loadings ([b] for #51; [e] for #160), and ash concentrations from 10 to 12 km ASL ([c] for #51; [f] for #160).

Figure 6.18 Polygons for simulation #51 and #160 for Zhupanovsky volcano on 29 December 2014 at 12:00 UTC. (a) Mass loadings for #51, yellow polygon, and #161, red polygon, as well as the mean from all simulation members, green polygon. (b) Ash concentrations at 10–12 km ASL for #51, yellow polygon, and #161, red polygon as well as the mean from all simulation members, green polygon.

Chapter 07

Figure 7.1 Coherence image examples of volcanoes in high‐latitude region during summer to early fall. Coherence images are overlaying on the corresponding radar intensity images and major volcanoes in the example sites are annotated: (a) Envisat data pair of Unimak Island (Alaska, USA) with a 35‐day time interval, (b) Radarsat‐1 data pair of Augustine volcano (Alaska, USA) with a 24‐day time interval, (c) Envisat data pair of Klyuchevskaya group of volcanos (Kamchatka, Russia) with a 35‐day time interval, (d) geographic locations of the coherence images shown in (a), (b), and (c).

Figure 7.2 Example of one way LOS deformation images in a descending imaging geometry without/with measurement errors. The red marker is the location of Mogi source center. (a) True LOS inflation of Mogi source; (b) deformation contaminated by

ϵ

u

; (c) deformation contaminated by ϵ

a

; (d) deformation contaminated by ϵ

l

.

Figure 7.3 Example of probability distribution of estimated parameters where measurements have spatially uncorrelated errors

ϵ

u

. Red bars denote 2

σ

confidence bounds and black bars denote the true value of the Mogi source parameters (the same holds for Figures 7.4 and 7.5).

Figure 7.4 Probability distribution of estimated parameters when measurements have spatially correlated errors ϵ

a

.

Figure 7.5 Probability distribution of estimated parameters when measurements have spatially local correlated errors ϵ

l

.

Figure 7.6 Predicted LOS displacements for flat‐surface Mogi model (

u

f

) and elevation‐varying Mogi model (

u

T

): (a) 3D view of the terrain condition of test site, and (b) the spatial distribution of residual displacements of

in millimeters. The red cross indicates the horizontal location of the source center. (c) The relation between the residual of

and the local topography.

Figure 7.7 An example of Mogi model inversion with descending and ascending track. (a) Synthetic inflation map in ascending geometry; (b) tradeoffs between source depth and volume change. Grey crosses denote the result with ascending orbit data only under poor coherence condition; black crosses denote the result with two orbits under poor coherence condition; green dots denote the result with ascending orbit data only at full coherence; red dots denote the result with two orbits at full coherence.

Chapter 08

Figure 8.1 Ash source term inversion for the Grímsvötn 2011 eruption: (a) IASI satellite retrievals of ash for 23 May 2011 at 21–22 UTC; (b) a priori source term for fine ash derived from radar observations; (c) a posteriori source term constrained by IASI observations; (d) FLEXPART a priori transport simulation using the source term in (b); (e) the FLEXPART a posteriori simulation using the satellite‐constrained source term in (c) showing better agreement with the satellite observations than the a priori simulation. Both FLEXPART simulations are using ECMWF operational forecast and analysis data. The Grímsvötn volcano is marked by a black triangle.

Figure 8.2 Schematic showing the procedures for the SO

2

source term inversion and data assimilation (4D‐Var) for the Grímsvötn 2011 eruption.

Figure 8.3 SO

2

source term inversion and data assimilation (4D‐Var) for the Grímsvötn 2011 eruption: Modeled and retrieved SO

2

column densities (Dobson Unit: DU) for 25 May 2011. Top row: 24 hr forecasts using the source term inversion (left) and 4D‐Var initialization from 24 hr earlier (based on the observations available until initialization time) (right). Bottom row: 4D‐Var analysis field (left) updated with the up‐to‐date OMI satellite observations (right). The model fields are valid for 14 UTC, while the observations are combined from multiple overpasses occurring between 12 and 16 UTC.

Figure 8.4 Inversion and data assimilation (4D‐Var) for the Grímsvötn 2011 eruption: The spatial correlation coefficient between modeled and observed SO

2

columns as a function of forecast length. The lines and markers correspond to forecasts initialized at different analysis times (red lines). The source term inversions using 24, 48, or 72 hr of observations are shown with blue lines.

Figure 8.5 Multi‐input ensemble modeling for the Grímsvötn 2011 eruption: Ash concentrations simulated by FLEXPART run on the ECMWF operational forecast data (upper left), maximum ash concentration from the simulations run on the 50‐member ENS ensemble (upper right), probability of ash concentration above

(lower left; mean [max] is 9[36] %) and

(lower right; mean [max] is 3[6]%). All plots are valid for 23 May 2011 at 22 UTC for the 2–3 km a.g.l model layer and the simulations are using the source term from Figure 8.1c.

Figure 8.6 Multi‐input ensemble modeling (top) and model intercomparison (bottom) for the Grímsvötn 2011 eruption. Top row: Total column ash simulated by FLEXPART run on ECMWF operational meteorological input fields and 50 ENS member ensemble. For the ensemble, the mean, median, and maximum of the members are shown. Bottom row: Simulations run on ECMWF ERA‐Interim meteorological input data by the three models FLEXPART, SILAM, and WRF‐Chem. The maximum values obtained from the three models are also shown. All model simulations are using the a posteriori source term from Figure 8.1c. The last plot (bottom right) shows the IASI satellite observations. All panels are valid for 23 May 13 UTC.

Figure 8.7 Multi‐input ensemble modeling for the Grímsvötn 2011 eruption: Time series of the total amount of ash in the model domain (upper panel) and the total horizontal ash extent (lower panel) for the reference FLEXPART run driven by ECMWF ERA‐Interim reanalyses data (dashed black lines), the FLEXPART simulation run on operational ECMWF data (red lines), and the FLEXPART simulations run on the 50‐member ENS data (green lines/area). The mean, median, and max of the ensemble members are shown (green lines) and the full range of the member runs is highlighted within the envelope (shaded green area).

Figure 8.8 Model intercomparison for the Grímsvötn 2011 eruption: (a) Time series of the total amount of ash in the model domain and (b) the horizontal ash extent for the three models WRF‐Chem, FLEXPART, and SILAM, and for the three‐model maximum product. For the ash extent, a threshold value of

was used. All model simulations are using the a posteriori source term from Figure 8.1c and ERA‐Interim meteorological fields. Statistical measures (two bottom panels): (c) Fractional mean (FM) differences, and (d) root mean square error (RMSE) between the ash mass loadings from the three models, calculated over the eruptive phase + 48 hr for colocated ash grid cells and using the same threshold value as for the ash extent.

Figure 8.9 Model intercomparison for the Grímsvötn 2011 eruption: Measured and modeled PM10 time series at a number of locations in Scandinavia. A background value of

was added to the model output to compensate for the measured background values at these locations.

Chapter 09

Figure 9.1 The energy line concept. PDC generation and propagation are conditioned by the initial height of collapse (

H

0

) and the angle between the energy line and the horizontal (

ϕ

), the latter representing a proxy for PDC mobility. Note that, in the diagram,

H

0

coincides with the top of the gas‐thrust region,

H

0

 ≈ 0.1

H

T

[

Wilson et al.,

1978]; not to scale

Figure 9.2 Geographical setting of the study area. (Top left) Map of south‐central Europe with Italy located in the middle. The yellow square denotes boundaries of the main image where different key locations are identified. The city of Napoli stands on or is surrounded by two principal volcanic systems: Campi Flegrei to the west, and Mount Vesuvius, our target volcano, to the east. The Campanian plain is the tectonic basin where the two volcanic systems originated. Some geomorphological highs, which limit the Campanian plain, are the Nola‐Sarno mountains (top right of map) and the Sorrento Peninsula (bottom right of map). Northward from Vesuvius, a remnant from previous edifice collapses is present: Mount Somma. A 3D view from the southwest (Torre del Greco is in the bottom‐center of the caption) of Mount Vesuvius cone and Mount Somma rim is given on the bottom left of the figure

Figure 9.3 Schematic representation of the origin and way of addressing the different types of uncertainty (aleatory and epistemic) quantified in this chapter. Aleatory uncertainty (solid lines) is described through probability density functions (PDFs) of the Energy Cone Model (ECM) parameters: collapse height,

H

0

, and PDC mobility,

ϕ

(note that

H

0

values are not to scale and

ϕ

values might seem greater than the actual values used in the chapter). Every simulation provides a value of area of PDC invasion (not shown here) and maximum runout (MR; bottom right of the cartoon). Input uncertainty is explored by running the ECM over Digital Elevation Models (DEMs) with diverse spatial resolutions (also notice that only the horizontal position of the 20, 40, and 80 m DEM grid points has to be considered). Parametric uncertainty is characterized by means of alternative choices for the PDFs (e.g., Tukey or linear PDFs; dotted‐dashed lines). Theoretical uncertainty arises from the fact that possible relationships between

H

0

and

ϕ

are not known. Finally, structural uncertainty derives from all the simplifications adopted by simulating the real phenomenon, PDCs, via the ECM (see text for more details).

Figure 9.4 Subsampling of the Gaussian‐Exponential, 40 m DEM configuration (aleatory uncertainty, AU) to explore theoretical uncertainty. Each graph delineates the ECM parameter space according to three eruption sizes: small (top), medium (middle), and large (bottom). Black dots indicate pairs of

ϕ

and

H

0

as sampled from the aleatory‐uncertainty configuration. Red open circles denote the inverse‐pattern subsample and purple open circles denote the direct‐pattern subsample. Magenta circles represent points shared by both inverse and direct subsamples.

Figure 9.5 Epistemic uncertainty description taking as example the case of theoretical uncertainty and area of PDC invasion. Aleatory uncertainty (AU) corresponds to the solid red line (independent pattern). Epistemic uncertainty is defined as the pale red area between the three empirical cumulative distribution functions (ECDFs). Horizontal distances between the curves demark possible ranges of the output variable considering both aleatory and epistemic uncertainties (see text for more details).

Figure 9.6 Structural uncertainty (SU) description. (Top) Misfit distributions of area of invasion (top left) and maximum runout (top right) according to three different eruption sizes at Mount Vesuvius (Italy): small (green), medium (purple), and large (red). (Bottom) Medium‐sized structural‐uncertainty representation in the form of SU ECDFs (dashed cyan lines) computed from aleatory uncertainty (AU ECDF, solid purple line) by adding distinct values of areal misfit,

M

A

(bottom left), and maximum‐runout misfit,

M

MR

(bottom right), picked up from the graphs on top (see text for details). Note that the horizontal distance between the minimum‐maximum SU ECDFs in the bottom graphs corresponds to the domain of

M

A

and

M

MR

(top graphs) (i.e., around 50 km

2

and 3 km, respectively, in the medium eruption size).

Figure 9.7 Aleatory uncertainty description, in terms of output empirical cumulative distribution functions (ECDF) for area of invasion and maximum runout (

A

and

MR

) of pyroclastic density currents simulated with ECM at Mount Vesuvius (Italy), according to three different eruption sizes: small (green), medium (purple), and large (red).

Figure 9.8 Comprehensive epistemic uncertainty description for the area of invasion (

A

) and maximum runout (

MR

) of pyroclastic density currents at Mount Vesuvius (Italy) according to three different eruption sizes: small (green), medium (blue), and large (red). Aleatory uncertainty (Fig. 9.7) lies inside the band of input uncertainty. Note how the contribution of each type of epistemic uncertainty to the total uncertainty changes along the graphs.

Figure 9.9 Conditional‐probability (

CP

) maps of PDC arrival (given the occurrence of an eruption of a specific size) computed by running the Energy Cone Model at Mount Vesuvius, Italy, according to three different eruption sizes: small (green text), medium (blue text), and large (red text). Leftmost‐side maps are actual conditional‐probabilities of PDC arrival (expressed between 0 and 1) while the other three maps in each eruption size show the differences in conditional probability, at each grid point, between alternate configurations (Gaussian‐Linear, 40 m DEM, independent pattern ‐second column‐; Gaussian‐Exponential, 40 m, direct pattern ‐3rd column‐; and Gaussian‐Exponential, 40 m, inverse pattern ‐4th column‐) and the aleatory‐uncertainty (AU) configuration (see text for more details). Colored zones in aleatory‐uncertainty maps indicate conditional‐probabilities greater than 0.05 and the solid red line displays the limit of

CP

> 0 (notice there is no red line in the large aleatory‐uncertainty map). The white star indicates the location of the city of Napoli (1: Somma Vesuviana; 2: Torre del Greco; 3: Scafati; 4: hospital in Massa di Somma; 5: Napoli‐Capodichino airport).

Figure 9.10 South‐north profile (A‐A’) of the study area (map on the left). (Top) Altitude profile where some regional topographic highs are identified. The three other profiles, from top to bottom, show the conditional probabilities of PDC invasion computed from ECM simulations at Mount Vesuvius and taking into consideration three eruption sizes: small (green), medium (purple), and large (red). Solid lines indicate the value of conditional probability as calculated from aleatory‐uncertainty simulations. Dashed lines denote the minimum and maximum values of conditional probability taking into account all sources of epistemic uncertainty but the structural uncertainty (see text for more details). The link between topography and PDC invasion can be visualized in all three eruption sizes.

Chapter 10

Figure 10.1 Point density analysis of (a) historic earthquakes with a magnitude greater than 3.5 and (b) historic earthquakes with a magnitude greater than 5.0. The Gyeongju area in the southeast of the Korea peninsula is marked by thick solid line.

Figure 10.2 Point density analysis of instrument‐detected earthquakes between 1978 and 2013 with a magnitude greater than 2.0.

Figure 10.3 Site classification map of the study area. Epicenter of the magnitude 6.7 scenario earthquake is shown by a black star.

Figure 10.4 Distribution of (a) population and (b) residential buildings in 2005.

Figure 10.5 Estimates of damage to residential buildings: (a) Distribution of residential buildings with moderate or severe damage and (b) damage to residential buildings by jurisdictions.

Figure 10.6 Hospital functionality at (a) day 1 and (b) day 7 after the earthquake.

Figure 10.7 Functionality of essential facilities: (a) schools, (b) police stations, and (c) fire stations.

Figure 10.8 Distribution of people seeking temporary shelter.

Figure 10.9 Estimated casualties and fatalities for an earthquake occurring at 2 a.m., 2 p.m., and 5 p.m. from left to right. Estimates are also grouped by jurisdiction.

Figure 10.10 Comparison of moderate or severe damage to residential buildings with and without the use of a site classification map by jurisdiction.

Chapter 11

Figure 11.1 The three digital elevation models tested with their respective flow paths for five lahar initiation points on the San Vicente volcano, El Salvador.

Figure 11.2 Comparisons of observed and simulated debris flows for the three digital elevation models tested (shown with the ArcGIS Spatial Analyst hydrological preprocessing technique, except where noted).

Figure 11.3 Debris flow simulations for the upper and lower limits for a 95% confidence interval for the three digital elevation models tested (shown with the ArcGIS Spatial Analyst hydrological preprocessing technique).

Chapter 12

Figure 12.1 (a) Localization of Colima Volcanic Complex and Popocatépetl volcano in the Trans Mexican Volcanic Belt; (b) localization map and digital elevation model of Colima volcano and Montegrande ravine; (c) map of Popocatépetl volcano and Huiloac gorge.

Figure 12.2 (a) Geophone raw data (blue lines and dots). Red dotted line represents the enveloping outline of the geophone signal used as the hydrograph. Peak discharge from

Vazquez et al.

[2014] data. (b) Hypothetical hydrograph constructed.

Figure 12.3 Results of Lahar Patrio simulation and comparison with the observed flow: (a) Differences between observed (purple plus yellow areas) and simulated lahar; (b) transversal sections comparing measured deposits and flow depths estimated by FLO2D at the same location.

Figure 12.4 Results of flow velocity calculations for lahar simulation of Lahar Patrio.

Figure 12.5 Simulation results using alternative input parameters: (a, b) Results of flow depth and flow velocity for the hypothetical hydrograph; (c, d) Simulation results with varying Manning‐

n

value for flow depth and flow velocity. Black line represents aerial distribution of lahar simulation using original input parameters.

Figure 12.6 Simulation results using different rheologic coefficients for Popocatépetl volcano [modified from

Caballero and Capra

, 2014]: (a) Areas of inundation, (b) differences in flow depths.

Chapter 13

Figure 13.1 A simplified model of factors driving the number, extent, and intensity of wildland fires in the natural environment. (Note that, for example, weather is a source of ignitions, and that ignitions are affected by the landscape as well, since an ignition must land on a receptive fuel in order to be viable. For the sake of simplicity, however, these relationships are not shown.)

Figure 13.2 A simplified model of factors driving the number, extent, and intensity of wildland fires and how management actions interact with wildfire in a coupled human‐natural system.

Figure 13.3 A simplified model of factors driving the number, extent, and intensity of wildland fires and how management actions interact with wildfire in a coupled human‐natural system, including temporal dynamics. Feedback loops are presented with dashed arrows.

Figure 13.4 Representation of the three dimensions of uncertainty (nature, location, level).

Figure 13.5 Location of uncertainties in wildfire model architecture at the near‐term (incident) planning horizon [format after

Warmink et al

., 2010].

Figure 13.6 Location of uncertainties in wildfire model architecture in the midterm (1–10 yr) planning horizon [format after

Warmink et al

., 2010]. While the sources of model technical, structural, and context uncertainties remain the same compared to the incident context (Fig. 13.5), additional uncertainties in model inputs are in play and different model outputs are available.

Figure 13.7 Location of uncertainties in wildfire model architecture in the long‐term (10–50 yr) context [format after

Warmink et al

., 2010]. Additional sources of input uncertainty have come into play in four of the five domains (ignitions, weather, landscape, and management). Uncertainties in long‐term planning inputs are of greater magnitude, including shifts in vegetation type and composition, changes in wildfire policy, and possible no‐analog fuel conditions.

Figure 13.8 Compounding uncertainty across planning levels. As modeling frameworks move from shorter to longer‐term planning contexts, additional sources of uncertainty come into play, and existing sources of uncertainty grow in magnitude.

Chapter 14

Figure 14.1 Effect of scan angle on MODIS observation of the Station Fire in Pasadena, California, on 30 August 2009. Fire detections near nadir (bottom left) show pixels to be almost square shaped at 1 × 1 km resolution, whereas near scan edge (bottom right) pixels are much fewer, individually stretched almost up to 4 × 2.5 km resolution, duplicated, and overlapping one another, and total FRP is underestimated.

Figure 14.2 Analysis of the effect of scan angle on collocated MODIS fire observation from Terra and Aqua, within 20 min of each other globally for 2003–2009. This collocation is only possible within the high latitudes (>55°N) where there is significant overlap of MODIS swaths between the two satellites. MxD14 represents the official MODIS fire products from Terra (MOD14) and Aqua (MYD14). There were 11,295 pairs of Terra/Aqua fire observations. (Top) Relative percentages of off‐nadir single pixel fire detection from one satellite and corresponding number of near‐nadir pixels of the same fire from the other satellite. (Bottom) Ratios of the FRP value of single off‐nadir pixels to total FRP value of the corresponding near‐nadir fire pixels, expressed both in terms of FRP (as in MODIS collection 5) and FRP per unit area (as in MODIS collection 4). The point values are the means of such ratios for bins of 1° off‐nadir observations starting at 25° scan angle, whereas error bars are the corresponding standard deviations of the FRP ratios.

Figure 14.3 Evaluation of uncertainty in aerosol optical depth (AOD) generated from WRF‐Chem model based on FEERv1 aerosol emissions, by comparison to satellite‐observed AOD over northern sub‐Saharan Africa (NSSA) during January–February 2010: (a) Fire locations and associated FRP values from MODIS on Terra and Aqua; (b) composited Terra and Aqua MODIS mean AOD for 5 February 2010, showing boxes where AOD comparisons are made; (c) WRF‐Chem simulation of only smoke aerosol AOD for 5 February 2010, also showing the sampling box locations. AOD values increase from blue to red. Notice the difference in AOD value ranges as indicated by the color scales between (b) and (c). Boxed areas are selected to avoid the main dust trajectory (as indicated by the dark‐red thick aerosol plume in [b]), such that the sampled AOD may be mainly smoke aerosols; (d) daily MODIS average AOD at Terra and Aqua overpass times (colored curves) and corresponding WRF‐Chem simulations (black curves) for 12–1 p.m. local time, which coincides approximately with the average of the local times of Terra and Aqua overpasses. The size of the circles on the satellite‐AOD curves indicate the extent of spatial coverage of the satellite retrievals within the sample boxes, as gaps do occur due to cloud or other factors that can cause AOD retrieval to fail (as seen in [b]).

Chapter 15

Figure 15.1 (a) Uncertainty associated with data is at a minimum in the present or near present time, with uncertainty in data increasing into the past and future timeframes. (b) If processes are correctly represented then model structure uncertainty is expected to decrease with model complexity. In the laboratory the associated increase in data uncertainty is flat, if the model is applied to data directly measured in the field for current conditions then the data uncertainty is higher but still relatively low and can accommodate a fairly complex model. If the data are from future projected conditions then the data uncertainty is expected to increase steeply with increasing model complexity.

Figure 15.2 Function shapes and parameters for (a) fuel load, (b) fuel moisture, (c) wind, and (d) slope.

Figure 15.3 (a) Tripod fire study area; (b) Tripod fire perimeter within the study area, overlaid on the digital elevation model; (c) example simulated perimeter from search 1; (d) example simulated perimeter from search 2.

Figure 15.4 Distribution of parameter values not rejected for each search. A broad distribution with no obvious modes indicates an identifiability problem for that parameter.

Figure 15.5 Sensitivity of WMFire_beta equations to the shape parameters for each component used to calculate the probability of fire spread is indicated by a steep relationship between the parameter value and the response, in this case threshold values at which the component spread probability crosses 0.9. If the threshold value changes rapidly with increasing parameter value then the model is judged sensitive, whereas if the threshold value shows little change with increasing parameter value then sensitivity is low. Sensitivity to the shape parameter decreases (becomes less steep) as the parameter value increases.

Chapter 16

Figure 16.1 Wildfire characteristics in the study region: (a) Mean monthly area burned over the period 1966–2010; (b) monthly area burned over the period of record (1966–2010) and the 96th sample quantile (horizontal red line, equal to 4047 ha or ~10,000 acres). Also shown is the study area in the context of the historic range of the longleaf pine (inset in [a]) and the histogram of observations in the inset in (b).

Figure 16.2 (a) Histogram of 10,000 simulations from the five‐fold cross‐validated ecological model and the verifying observation (vertical dashed line) of the number of EBA months (>4047 ha) for the period 1980–2009. (b) BMA PDF of the predicted number of EBA months for the period 1980–2009 for the joint ecological‐climate model. Verifying observation is the dashed vertical line, fiftieth percentile is the solid thick vertical line, thin vertical lines cover the 95% interval, and grey lines are individual GCM PDFs. (c) BMA PDF of projected difference between the number of EBA months in 2070–2099 versus 1980–2009 for the b1 emission scenario. (d) Same as (c) but for the a1b emission scenario. (e) Same as (c) but for the a2 emission scenario.

Chapter 17

Figure 17.1 Climate projections under two Representative Pathways (RCP 4.5 and RCP 8.5) for the western United States. For the historical period (1895–2010), the thick black line corresponds to PRISM data. From 2011 forward, each climate model projection has been drawn and the thick black line corresponds to the ensemble mean.

Figure 17.2 Simulated vegetation distribution (mode for 1971–2000) and projected vegetation distribution for the end of the 21st century (mode for 2071–2100) by 20 climate models from CMIP5 under RCP8.5.

Figure 17.3 Changes between mid‐21st century (2036–2065) or the end of the 21st century (2071–2100) and the historical baseline (1971–2000) for (a) average annual temperature and (b) average annual precipitation under RCP 4.5 and 8.5 across 20 CMIP5 climate models.

Figure 17.4 Level III ecoregions in the contiguous United States. Central Valley: 7; Willamette Valley: 3; Sierra Nevada: 7, and Cascades: 4.

Figure 17.5 Carbon stocks and fluxes dynamics as well as fraction of area burned by wildfires over the western United States from 1895 to 2100 using climate projections from 20 CMIP5 climate models under RCP 4.5 and 8.5. From 2011 forward, each climate model‐driven projection has been drawn and the thick black line corresponds to the ensemble mean. NEP stands for net ecosystem production (net primary production minus heterotrophic respiration).

Figure 17.6 Ecosystem carbon density simulated for baseline conditions (1971–2000) and simulated change under future conditions for the middle and end of the 21st century relative to baseline.

Figure 17.7 Temporal dynamics of the live vegetation carbon pool, soil and litter carbon pools, and area burned for the Willamette Valley of Oregon and the central valley of California as simulated by the MC2 dynamic vegetation model. From 2011 forward, results are shown for 20 of the CMIP5 climate futures under RCP 8.5 and the thick black line corresponds to the ensemble mean.

Figure 17.8 Temporal dynamics of the live vegetation carbon pool, soil and litter carbon pools, and area burned for the Sierra Nevada of California and the Cascades of Oregon as simulated by the MC2 dynamic vegetation model. From 2011 forward, results are shown for 20 of the CMIP5 climate futures under RCP 8.5 and the thick black line corresponds to the ensemble mean.

Chapter 18

Figure 18.1 Overview of the HESFIRE model: (a) Model diagram, (b) optimization procedure, (c) comparison of average annual burned area over 1997–2011 between HESFIRE and observation‐derived data (Global Fire Emission Database, GFED [

van der Werf et al.

, 2010]). Note that agricultural and deforestation fires were excluded from the GFED data, consistent with the representation of fires on natural ecosystems only in the model.

Figure 18.2 Examples of discrepancies in the input data used to run HESFIRE: (a) land use (cropland + urban areas) patterns in Globcover, (b) difference with the MODIS land cover product, (c) average annual relative humidity in NCEP, and (d) difference with the ERA data.

Figure 18.3 Dominant parameter‐sensitivity of average annual burned area in HESFIRE.

Figure 18.4 1997–2011 average annual burned area in HESFIRE in the four runs with different input data.

Chapter 19

Figure 19.1 The postfire debris flow hazard cascade with the sequence of preconditions and processes leading to hazardous conditions. The down arrows reinforce the linear nature of this conceptual model where the generation of hazardous conditions is contingent upon the preceding component steps. This hazard cascade offers logic for future development of a comprehensive postfire debris flow hazard prediction framework.

Figure 19.2 Conceptualization of forest hillslope structure before and after high‐severity wildfire where all biomass is consumed, emphasizing the role of vegetation disturbance in postfire erosion processes and debris flow generation. Loss of vegetative cover including canopy, litter, and duff increases effective rainfall. Though coniferous trees are illustrated, the processes are similar in deciduous forests and shrublands. The presence of a soil hydrophobic layer is uncertain as fire may increase, decrease, or destroy water‐repellent conditions [

Shakesby and Doerr

, 2006]. While vegetation disturbance is recognized as exerting significant influence over postfire erosion and debris flow processes, research has substantially focused on the role of soil changes [

Moody et al.,

2013;

Shakesby,

2011] with limited study of effects of vegetation disturbance, leaving significant knowledge gaps.

Figure 19.3 Conceptualization of values‐at‐risk downstream of burned areas, illustrating structures and infrastructure that can be damaged or destroyed, reflecting two phases of postfire debris flow hazards: direct impacts from a debris flow mass, and impaired water quality from sediments deposited into streams by debris flows. Sediments may compromise intended water use in reservoirs and at irrigation, commercial, and water supply intakes. While wetland habitat may be enhanced, it can also be disturbed beyond ecosystem tolerances depending on the resilience of the system, debris flow magnitude, and history of prior disturbances.

Chapter 20

Figure 20.1 Study area with the selected watersheds identified. The top panel shows a zoomed in detailed view of the watershed outlets into the adjacent urban interface, and identifies key water‐infrastructure located in the vicinity. In some cases, there is urban development within the watershed itself (i.e., watershed 1). The bottom left panel shows complete delineation of the watersheds and their more general location within the Sandia Mountain range. The color of each watershed corresponds to the colors used to distinguish between watersheds in the results figures. The bottom right panel shows the location of the watersheds with respect to the greater landscape used in modeling the fire inputs.

Figure 20.2 Overview of modeling framework, showing linkages between the various models (named in bold), which provides the ultimate outputs characterizing watershed‐level risks (debris‐flow probability and volume). The two grey boxes represent inputs modeled using fire‐behavior and growth models.

Figure 20.3 The depiction of the final step in the methods that produces the necessary fire inputs for the debris‐flow model. The spatial intersection of the FlamMap severity output and the FSim fire‐perimeter output with watershed boundaries jointly determine the distribution of watershed area and percent burned at MHS.

Figure 20.4 Debris‐flow probability versus debris‐flow volume for the six storm recurrence intervals for each watershed. The individual dots represent the individual fire events associated with each watershed and the range of variability surrounding the potential fire events. Each dot may represent more than one fire event because of the limited spatial combinations of percent area burned at MHS.

Figure 20.5 Frequency histograms showing distributions of the postwildfire debris‐flow volume for the set of fire events. The histograms allow comparison of watershed hazard by frequency of event, which can distinguish watersheds with low‐frequency large‐volume events from those with high‐frequency small‐volume events, which can result in similar combined hazard metrics.

Figure 20.6 The summary statistics for the wildfire events by watershed. The top left panel shows the annual burn probability as a percentage for each watershed. The top right panel shows the percentage of each watershed that would burn at MHS as modeled in this study, if the entire watershed burned. The probability of debris flow (center panels) and volume of debris flow (bottom panels) show the maximum and mode (left, same value for each watershed) and mean (right) values, given a wildfire event. The colors represent the same value between the mean and the mode event for the two variables, allowing for a direct visual comparison between the modal statistics.

Chapter 21

Figure 21.1 Map showing the area of Upper Adige river draining at Trento. Shades of color show terrain elevation. Triangles show locations of available rain gauges in the region. Dot symbols correspond to the locations of the 82 debris flows analyzed. Location of the Macaion weather radar with the range circle at 60 km is also shown.

Figure 21.2 Maps of rainfall accumulation for the 10 rainfall events (a–j). Dotted circles represent the location of triggered debris flows, black triangles show the location of rain gauges. Events are presented in chronological order: (a) 1 August 2005; (b) 20–21 June 2007; (c) 26 June 2008; (d) 20 July 2008; (e) 29 July 2008; (f) 6 August 2008; (g) 16–17 July 2009; (h) 30 July 2009; (i) 4 September 2009; (j) 14–15th August 2010. Location of each rainfall map relative to the basin is shown using a black rectangular box superimposed on a minimap of the basin.

Figure 21.3 Rainfall duration (

D

, in hours) and mean rainfall intensity (

I

, in mm h

−1

) of debris flow triggering rainfall events, according to radar observations. Black line is the estimated

ID

rainfall threshold at 10% exceedance level.

Figure 21.4 Relative error in rainfall estimation for different interpolation methods and gauge densities examined. Note that square symbols correspond to mean values and vertical bars extend from 5th to 95th percentile.

Figure 21.5 Distribution of

ID

thresholds derived according to different estimation methods and for different gauge densities. Color area corresponds to range of distribution of resulted

ID

thresholds. Reference

ID

threshold (black line), based on radar, is superimposed. Panels from top to bottom correspond to gauge densities of one station over 10, 20, 50, and 100 km

2

, respectively.

Chapter 22

Figure 22.1 High level flow of the modeling approach used in this analysis. Input data in the form of landslide records and weather data are transformed into 1‐dimensional feature vectors, each of which contains descriptive variables for a single landslide event. Supervised learning models are built and evaluated using training and test subsets of the data, respectively. Model parameters and feature definition are adjusted based on measured model performance, and the analytical cycle is repeated until predictive performance has been optimized, after which final landslide predictions are generated.

Figure 22.2 Composites of all Swiss winter slide dates. The colors in panel (a) correspond to the value of the standardized anomaly of 500‐hpa geopotential height (Z500) and the arrows show vertical integral atmospheric vapor flux. The colors of panel (b) correspond to the level of total precipitable water (TPW) and the contour lines are 500‐hpa vertical pressure velocity (ω500).

Guide

Cover

Table of Contents

Begin Reading

Pages

ii

iii

iv

vii

viii

ix

x

1

2

3

4

5

6

7

8

9

11

12

13

14

15

16

17

18

19

21

22

23

24

25

26

27

28

29

31

32

33

34

35

36

37

38

39

40

41

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

167

168

169

170

171

172

173

174

175

176

177

179

180

181

182

183

184

185

186

187

188

189

191

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

215

216

217

218

219

220

221

222

223

224

225

226

227

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

277

278

279

280

281

282

283

284

285

287

288

289

290

291

292

293

294

295

296

297

298

299

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

319

320

321

322

323

324

325

326

327

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

Geophysical Monograph Series

A Continental Plate Boundary: Tectonics at South Island, New Zealand

David Okaya, Tim Stem, and Fred Davey (Eds.)

Exploring Venus as a Terrestrial Planet