International Conference on Applications of Statistics and Probability in Civil Engineering (ICASP) (12th : 2015)

Assessing and managing natural risks at the Panama Canal Alfaro, Luis D.; Baecher, Gregory B.; Guerra, Fernando; Patev, Robert C. Jul 31, 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
53032-Paper_122_Alfaro.pdf [ 4.51MB ]
Metadata
JSON: 53032-1.0076030.json
JSON-LD: 53032-1.0076030-ld.json
RDF/XML (Pretty): 53032-1.0076030-rdf.xml
RDF/JSON: 53032-1.0076030-rdf.json
Turtle: 53032-1.0076030-turtle.txt
N-Triples: 53032-1.0076030-rdf-ntriples.txt
Original Record: 53032-1.0076030-source.json
Full Text
53032-1.0076030-fulltext.txt
Citation
53032-1.0076030.ris

Full Text

12th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  1 Assessing and Managing Natural Risks at the Panama Canal Luis D. Alfaro Autoridad del Canal de Panamá, Panama Gregory B. Baecher University of Maryland, College Park, Maryland, USA Fernando Guerra Autoridad del Canal de Panamá, Panama Robert C. Patev  University of Maryland, College Park, Maryland, USA ABSTRACT: The Panama Canal Authority (ACP) has undertaken a comprehensive assessment of nat-ural and chronic risks to improve planning and to optimize its engineering safeguards. The risk assess-ment program began with an all-inclusive risk register. The register lists somewhat more than 500 items divided among the many categories of facilities constituting the Canal (dams, locks, cuts, gates, power stations, water plants, and others) and among the various hazards facing the Canal (seismic, hy-drologic, meteorological, operational). Scientific and operations data for the Canal have been compiled to characterize risk, while modern reliability models have been developed to translate those data into actionable assessments of reliability and consequence. Risks were categorized as catastrophic, signifi-cant, or moderate. The first set has been engineered in detail; the others have been approached opera-tionally. The resulting probabilities and consequences are tracked in acceptable risk charts in FN for-mat better to understand where risk remediation is called for. This comprehensive risk assessment is al-lowing ACP to reduce risk while meaningly keeping costs under control. The Panama Canal, commissioned in 1914, is one of the world’s iconic engineering projects. The Canal provides passage to 18,000 vessels a year, and carries more than five percent of inter-national maritime trade. In the early 1900’s, the Panama site, unlike Nicaragua, was thought free of natural hazards and was favored in part be-cause of this. History has changed that appraisal and it is now understood that seismic, hydro-logic, and meteorological hazards do affect the Canal. In addition to natural hazards, an engi-neered system of this scope must also grapple with chronic risks due to aging and maintenance.  1. PROJECT PHASES Beginning in 2011, a systematic risk analysis was undertaken to assess the state of natural risks facing the Canal. The project was divided into phases: Phase I focused on developing a basis for the risk analysis. This included expanding the ex-isting risk register for natural risks and building an inventory of existing ACP infrastructure. This inventory includes, but is not limited to, dams, spillways, locks, navigation channels, power plants, water intakes, communications systems, bridges, and other significant structures (Figure 1). Phase I also included a failure and effects analysis (FMEA). This work was undertaken by the Engineering Division of La Autoridad del Canal de Panamá (ACP). Phase II focused on engineering and sys-tems reliability. This involved assessing annual probabilities associated with natural hazards af-fecting the Canal, and the corresponding fragili-ties of the infrastructure. Life cycle analyses were performed of maintenance repair and re-place strategies. 12th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  2   Figure 1. Critical infrastructure components of the Panama Canal.  Phase III developed a probabilistic risk analysis methodology and implemented this for a series of modeling approaches to the various in-dividual classes of structures. Phase IV identified potential consequences of adverse behaviors and failure on financial costs to ACP and economic costs to the Nation. Potential loss of life was considered negligible. These consequences are visualized in frequency-magnitude (complementary cumulative distribu-tion) curves for the purposes of comparison with acceptable risk guidelines, and for communi-cating with stakeholders.	   Phase V built on the assessment of risks and consequences and their sources to lay the foun-dation for a risk management strategy.   2. QUALITATIVE RISK  The initial step was the development of a sys-tematic risk register. The risk register is a list of hazardous events, facilities and facility compo-nents, and possible consequences if the hazards occur. The risk register provides the platform for the risk analysis, and is thus a critical step.  The purpose of the risk register is to identify as many significant risks to the Canal infrastruc-ture as possible, and to rank order those risks for further analysis (Figure 2). This rank ordering categorized risks into three sets: (1) those risks which required further analysis and possibly modeling to obtain quantitative assessments (red), (2) those risks that were significant and needed to be monitored but were not deserving of detailed analysis (yellow), and (3) those risks 12th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  3 that were not deserving of special attention but could be managed as part of normal operations (green).  A qualitative risk assessment was used to rank order the structures and components within the portfolio. Development of qualitative risk as-sessment protocol required a number of working sessions to develop categories of hazards and an inventory of critical infrastructure in the ACP portfolio. ACP utilized its own subject matter experts (SMEs) in multidisciplinary teams to as-sess likelihoods and consequence and the parti-tion and ordering of the risk register.    Figure 2. Example section from the risk register, showing risks for part of the Madden Dam spillway section. Each row (specific risk) is characterized by the name of the structure, type of the hazard, magni-tude of the hazard, component affected, and cause of the adverse outcome. The probability and cost of each outcome are ranked from one to four, and an overall risk ranking based on a risk matrix.  In the qualitative phase of the risk analysis, simple ordinal scales or rankings were assigned to probabilities and consequences. That is, the probability of the hazard was ordinal-scaled from one to four. The severity of consequence was or-dinal-scaled from one to four. This provided a starting point, but ordinal nature of the scales made cross comparisons within the register diffi-cult. So a semi-quantitative set of scales was needed. To make cross comparisons more reliable, semi-quantitative scales were developed for haz-ard probability (Table 1) and for consequence (Table 2). An attempt was made to anchor these semi-quantitative scales to events and outcomes that were intuitively familiar to ACP’s subject matter experts. In this qualitative phase, the probabilities and costs in the risk register were based on the judgment of the SME’s Table 1. Semi-quantitative scale of probability 	   DESCRIPTION	   PROBABILITY	   COMPARISON	  1	   Very	  likely	  	   P>0.1	  	   Small	  land-­‐slide,	  draft	  	  restrictions	  2	   Likely	  	   P=0.1	  to	  	  0.01	  	   Landslides	  without	  	  controls	  3	   Unlikely	  	   P=0.01	  to	  0.001	  	   La	  Purisima	  2010	  Flood	  4	   Very	  unlikely	  	   P<0.001	  	   Large	  	  earthquake	   Table 2. Semi-quantitative scale of consequence 	   VERBAL	  DESCRIPTION	   EXAMPLE	  OF	  LOSS	  1	   Complete	  Loss	  of	  navi-­‐gation	  operations	  for	  more	  than	  a	  year	   Loss	  of	  Gatun	  or	  Mad-­‐den	  Dam.	  2	   Impede	  operations	  for	  long	  period	  (>1	  year)	  or	  create	  major	  direct	  or	  indirect	  economic	  cost	  More	  than	  $1b	  loss	  of	  toll	  revenue.	  Seriously	  compromise	  reliability	  of	  important	  compo-­‐nents	  3	   Impede	  operations	  for	  short	  period	  (<1	  year)	  or	  moderate	  direct	  or	  indirect	  economic	  cost	  More	  than	  $500m	  loss	  of	  toll	  revenue.	  Direct	  repair	  costs	  greater	  than	  $500m	  4	   Damages	  that	  affect	  canal	  capacity	  and	  rev-­‐enues	  More	  than	  $100m	  loss	  of	  toll	  revenue.	  Direct	  repair	  costs	  greater	  than	  $100m	  5	   Economic	  damages	  but	  canal	  continues	  to	  op-­‐erate	  Less	  than	  $100m	  loss	  of	  toll	  revenue.	  Direct	  re-­‐pair	  costs	  Less	  than	  $100m.	  No	  impact	  to	  ACP	  reputation	  3. QUANTITATIVE RISK  The quantitative risk analysis was conducted us-ing current hazard-vulnerability-consequence  Structure Hazard Magnitude Component Comment Cause Pr Cost RankMadden7Spillway Floods PMF Drum7gates Steel7Structure7and7all7castings7concrete Debris7blockage,7undermaintenance7or7repair 4 3 0Impact7by7barge7or7lare7object 3 4 0Conrete7piers Previously7damaged7form7other7even 4 4 0Concrete7main7section High7water7velocity 2 4 0Uplift7forces,7overload,7cavitation 4 2 0Overload 4 4 0Roadway7bridge Impact7by7large7objects 4 4 012th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  4 methods similar to those adopted in the USACE Interagency Performance Evaluation Taskforce study following Hurricane Katrina (IPET 2008) and the California Delta Risk Management Study (DRMS) of seismic and hydrological risk to the California Delta (URS/JBA 2007).  The approach separates the components of risk into three parts: Hazards, the natural or an-thropogenic threats posing potential loads on the system; system response, fragility of the engi-neered system or its components to the loads posed by the hazards; and consequences, the po-tential outcomes in financial cost, mortality and morbidity, environmental impacts, or other fac-tors caused by adverse performance of the sys-tem under hazard loads (Figure 3). 	  Figure 3. Hazard-vulnerability-consequence method of natural hazard risk analysis, using seismic accel-eration (PGA) as an example. AEP=annual exceed-ance probability. Adapted from Grossi and Kunreu-ther (2005). 4. HAZARDS The hazards to which the Canal is exposed are divided into three categories: Natural hazards, operational and maintenance, and malicious an-thropogenic hazards. Three natural hazards were addressed: seismological, hydrological, and me-teorological. A variety of others—hurricane, tsu-namis, tornedo, and sedimentation—were re-viewed but none reached the catastrophic level. Operational and maintenance hazards are those that arise internal to the operations of the Canal and those due to aging and maintenance. These include navigation incidents, dredging and tug erosion, and time-related deterioration. Mali-cious anthropogenic hazards are those caused by purposeful acts typically by agents external to the Canal operations. These include acts of ter-rorism but may also include acts by aggrieved personnel. The present study focused on natural and operational hazards. Malicious anthropogen-ic hazards were the subject of a separate, inde-pendent study sponsored by the ACP Protection Division.    Figure 4. Seismic hazard curves for the New Atlantic Locks as PGA vs. AEP. The mean curve shown in red, median in green; also shown are the 5%, 15%, 85%, and 95% fractiles (URS 2008).  A typical probabilistic seismic hazard curve is shown in Figure 4. Similar hazard curves were developed for hydrological and meteorological hazards for each major component of the Canal infrastructure. 5. FRAGILITY  The performance of individual components was summarized in one of two ways: (1) As a fragili-ty curve expressing the conditional probability of failure of the structure for various levels of haz-ard loading, or (2) as a systems response curve expressing the conditional probability of levels of an engineering performance indicator (e.g., displacement, deformation, factor of safety). HazardVulnerabity(Fragility)ConsequencesAssetsLoadAEPLoadPfConsequencesAEPP(Load) P(Failure|Load) P(Consequences)12th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  5 The former is used when the modeling leads to a discrete failure vs. no-failure outcome, while the latter is used when the modeling leads to a gradient of possible severity of the outcome. In the former case the consequence of failure is a fixed although perhaps uncertainty value, while in the latter case the consequence of failure is a variable dependent on the level of the perfor-mance indicator (and possibly also uncertain).   Figure 5. Gatun Dam from the above, showing the spillway at center, Gatun Powerhouse center.  Figure 6. General forces diagram for Gatun Spill-way: (1) monolith, (2) pier, (3) water behind the spillway gates, (4) gate, (5) lake, (6) soil in front of monolith, (7) rock, (8) downstream, (9) key, (10) up-lift.  In either of these cases, the corresponding fragility or system response is analyzed using structural and geotechnical reliability methods of the sorts described in Ditlevsen (1996) or Baecher and Christian (2003). Depending on the system and failure mode, these methods ranged from simple Monte Carlo simulation to stochas-tic finite element method. For example, Figure 5 shows Gatun Dam, spillway, and powerhouse. Seismic forces were generated using a design spectrum from a seismic report prepared by URS (2008). For each site, response spectra were de-veloped with different damping coefficients. Us-ing Chopra’s simplified method (Tan and Chopra 1996), the lateral earthquake forces were esti-mated from the earthquake design spectrum (Figure 6). The effects of lake interaction and water compressibility, dam-foundation rock in-teraction, and the absorption of hydrodynamic pressure waves were considered in the reservoir bottom sediments and in the underlying founda-tion rock.  6. EPISTEMIC UNCERTAINTY A logic tree approach provides a numerical way of handling parameter uncertainty and propagat-ing its effects to uncertainties in (or confidence bounds on) the output predictions. In the system simulations, the epistemic uncertainties are gath-ered into their own uncertainty tree—a so-called logic tree—ahead of the simulation representing aleatory uncertainties (Bommer and Scherbaum 2008).  The aleatory simulation is calculated condi-tional on the value of the epistemic parameters generated in the logic tree for hazard, fragility, and consequence. These probabilities scan be combined to yield probabilities of the end leaves of the epistemic analysis . Using the Monte Carlo method, the logic tree approach decomposes the numerical model-ing into a two-step, nested process: (1) A simula-tion is made of many values, N, of the epistemic uncertainties leading to a large number of reali-zations, and (2) a set of M iterations of the HVC model is made for each realization of the epis-temic uncertainties simulated in step one (Figure 7). 12th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  6 7. CONSEQUENCES Consequences are divided into three compo-nents:  1. Direct cost of damages, 2. Lost direct revenues due to blockage, and 3. Implications on the national economy.  Due to the nature of the Canal, the economic im-pact on the Canal or the nation is far greater than the cost of direct damages. Thus, consequences are presented with this segregation to allow dif-ferent types of risk analysis. Three states of di-rect cost were considered: Severe (closure of Ca-nal for six months or more, direct cost more than USD1b), moderate (temporary closure of Canal, direct cost up to USD1b), and light (no closure of Canal, direct cost less than USD100m). Only monetary losses were considered.  Figure 7. Schematic drawing of nested probability calculation. Epistemic uncertainties are represented in a logic-tree, which captures probability distribu-tions on model and parameter uncertainty. Instances of the epistemic parameters are then used to charac-terize one iteration each of the aleatory event tree, which is repeated many times. 8. COMPARISON OF RISKS The annual risks across the various failures modes of Canal infrastructure were compiled. Given the magnitude of these risks, management considered two questions: (1) Is a particular risk acceptable, and if not, (2) how much must it be reduced or how can it be managed?  To help answer these questions, ACP has long turned to frequency-magnitude (FN) curves of the sort introduced in Whitman (1984) (Figure 8). An early example was the use of such FN curves in developing ACP’s instrumentation program in the Gaillard Cut (Alfaro 1988). 8.1. Catastrophic risks In addition to the simple calculation of expected consequences (i.e., risk = probability × cost), ACP also judged the acceptability of risk by comparison to risks accepted at other facilities and in other contexts. These are risks now being accepted; they may or may not be "acceptable" in the context of any one operation, but they pro-vide a background for informing decisions and for communicating with management.   Figure 8. A generic FN chart (adapted from Whitman 1984). Variants of this chart as shown in Figure 8 were adapted to judge the acceptability of various risks in the risk register.  Figure 8 shows annual frequencies of failure of various types of constructed facilities and es-timates of the consequences of failure from Whitman. The data shown come from empirical studies of past failures, from insurance industry statistics, and from published risk analyses that were performed during design or operation (Note, the UK Canvey Island studies in part in-formed HSE’s (2001) tolerable risk guidelines for loss of life). More detailed discussion of FN Engineered(SystemLoads(&(DisturbancesPerformanceProper9es(&ParametersAleatory SimulationEpistemic Simulation12th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  7 curve approaches is provided in Bedford and Cooke (2001). Two envelopes are sketched in Figure 8. The first is an approximate upper bound for risks generally agreed to be acceptable for failure of constructed facilities that threaten the general public. The second is an approximate upper bound for risks appearing to be marginally ac-ceptable for failure of facilities that give no threat to the general public. These data are im-precise and incomplete. The envelopes are at most first-order approximations.    Figure 9. Slope failures in the Gaillard Cut. Dia-monds are “routine failures” due to rainfall and ero-sion. Crosses are a Poisson-lognormal approxima-tion to historical failures. Triangles are the binned-approximation. Red squares are calculations of seis-mically induced failure probabilities due to ground acceleration.  Typical results for slope failures in the Gail-lard Cut are shown in Figure 9. The low proba-bility-high consequence risks to the right-hand side are associated with large earthquakes on the Pedro Miguel and Limon fault system. These are catastrophic risk in that many km of slopes may slide and the associated cost would be in the hundreds of millions of USD.  To the left-hand side are historical failures due to rainfall and channel erosion. These are not catastrophic in that they occur nearly every year and are managed by observation and mainte-nance as mentioned above. The decision was made to separate catastrophic from routine risks at a cost of USD 10m. Similar results were gen-erated for each high-risk entry in the risk regis-ter. 8.2. Non-catastrophic risks The “acceptable risk” curve adopted for evaluat-ing risk items apply only to catastrophic risks, that is, major failures, not to routine accidents. “Yellow” and “green” risks in the risk register are mostly of this latter type. These risks may plot above the acceptable risk line and still be satisfactory. From comparison and presentation purposes, routine risks are plotted in fN space (i.e., the derivative of FN space) and compared with lines of constant expected annual value, i.e., 45-degree lines (Figure 10).  	  Figure 10.  f-N curve of  less-than-catastrophic (yel-low) risks. The constant f-N curves indicate how much could be reasonably spent annually on reduc-ing the risks. These risk items can reasonably plot above the “acceptable risk” curve as they are non-catastrophic.  9. CONCLUSIONS An orderly Risk Management System (RMS) for the Panama Canal infrastructure subjected to physical hazards has been established over the past five years, in compliance with the Canal’s 2010 “Integration Program for the Expanded Ca-nal”. The RMS integrates the various existing, but separate risk mitigation programs at the Panama Canal, and benefits from the experience of each.  1.E$04'1.E$03'1.E$02'1.E$01'1.E+00''0.01'' '0.10'' '1.00'' '10.00'' '100.00'' '1,000.00''Annual&Exceedance&Probability&Cost&(mUSD)&F$N'Chart'Historical' Total'Cost' FN'Seismic' MC'Total'Cost'O&M risks Catastrophic risksLines of constant f*NAcceptableRisk CurveConsequence, NMarginalProbability, f12th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP12 Vancouver, Canada, July 12-15, 2015  8 The RMS has the capability for integrating dif-ferent types of physical risks: catastrophic risks, chronic risks and human risks. This segregation relates to their respective mitigation strategies. Catastrophic risks are mitigated through analysis, reinforcement, and mitigation before the hazard-ous event. Chronic risks are mitigated by sys-tematic inspections and maintenance practices. Human risks are managed by strategies imple-mented by the Canal Protection Division.  The risk analysis has demonstrated that the dominant risks facing the Canal concern the three large dams retaining Lake Gatun and its water supply (Gatun, Madden, and Miraflores Dams), and the Gaillard Cut slopes. For each of these, the hazard of greatest importance is seis-mic. The risk analysis also indicated the need for more advanced and detailed analysis of the struc-tural and geotechnical reliability of these struc-tures and components. The RMS enables an objective comparison of risks and a systematic comparison of costs and potential consequences, to develop a rational strategy for prioritizing capital investments relat-ed to risk mitigation. The RMS permits continuous upgrading and maintaining so its relevance is maintained as the Canal ages and changes. The output from this ef-fort is being directly incorporated into the ACP’s overall Enterprise Risk Management Program. In this way, physical risks are viewed in the same light as other types of risks affecting the Organi-zation. REFERENCES  Alfaro, L. D. (1988). The risk of landslides in Gail-lard Cut. Autoridad del Canal de Panamá, Balboa Heights, Panama, 78. Baecher, G., and Christian, J. (2003). Reliability and Statistics in Geotechnical Engineering. Wiley, Chichester, West Sussex, England  ; Hoboken, NJ. Bedford, T., and Cooke, R. M. (2001). Probabilistic risk analysis  : foundations and methods. Cambridge University Press, Cambridge, UK  ; New York, NY, USA. Bommer, J. J., and Scherbaum, F. (2008). “The Use and Misuse of Logic Trees in Probabilistic Seismic Hazard Analysis.” Earthquake Spec-tra, 24(4), 997–1009. Chopra, A., and Tan, H. (1989). Simplified Earth-quake Analysis of Gated Spillway Monoliths of Concrete Gravity Dams. Report to USACE WES, University of California, Berkeley. Ditlevsen, O. (1996). Structural reliability methods. Wiley, Chichester  ; New York. Grossi, P., and Kunreuther, H. (2005). Catastrophe Modeling: A New Approach to Managing Risk. Springer. HSE. (2001). Reducing Risks, Protecting People – HSE’s Decision Making Process. UK Health and Safety Executive, London: HMSO. IPET. (2008). Performance Evaluation of the New Orleans and Southeast Louisiana Hurricane Protection System, Final Report, v.8 – Engi-neering and Operational Risk and Reliability Analysis. Interagency Performance Evalua-tion Taskforce, Washington, DC. Tan, H., and Chopra, A. (1996). “Dam-Foundation Rock Interaction Effects in Earthquake Re-sponse of Arch Dams.” Journal of Structural Engineering, 122(5), 528–538. URS. (2008). ACP Geotechnical Services Contract Task Orders 1 and 5: Seismic Design Crite-ria for ACP Critical Structures. Autoridad del Canal de Panamá, Balboa Heights, Pana-ma. URS/JBA. (2007). Delta Risk Management Strategy (DRMS) Phase 1 Risk Analysis Report. Pre-pared by URS Corporation/Jack R. Benjamin & Associates, Inc. for the California Depart-ment of Water Resources, Sacraemento. Whitman, R. V. (1984). “Evaluating the calculated risk in geotechnical engineering.” Journal of the Geotechnical Engineering Division, ASCE, 110(2), 145–188.   

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.53032.1-0076030/manifest

Comment

Related Items

Admin Tools

To re-ingest this item use button below, on average re-ingesting will take 5 minutes per item.

Reingest

To clear this item from the cache, please use the button below;

Clear Item cache