The objective of this project was to develop guidance that outlines best practices and methodologies that can be used to formulate scenario-dependent probabilistic estimates of future extreme sea levels under a changing climate and apply it to select Department of Defense (DoD) sites in the Pacific Islands. The proof-of-concept products can be used to support decision-making ranging from area-wide vulnerability assessment related to climate adaptation planning and disaster risk reduction to site-specific analysis related to design and maintenance of facilities and infrastructure at DoD sites.
Analysis of Pacific Island tide station records has shown that extreme water levels primarily result from a combination of global and regional changes in mean sea level, El Niño Southern Oscillation (ENSO) and other modes of natural variability, tropical and extratropical storms, and unusually high tides. This suggests that in response to a changing climate, and in addition to increasing global sea level, alterations to natural patterns of sea level variability and storminess may contribute to the more frequent occurrence of extreme water level events. Analyses also show that the relative importance of the various contributors to extreme water levels varies from location to location. This suggests that changes in the frequency of occurrence of extreme water level events due to a changing climate may be highly localized. Potential limitations related to the sparseness of data (e.g., if the tide station record is of limited duration or if no station exists near a location of interest) represent another set of issues to account for.
A sequence of analyses using families of distributions – the Generalized Pareto Distribution (GPD) and the Generalized Extreme Value (GEV) – was conducted. The focus was on a form of non-stationary extreme value analysis (EVA) using the GEV distribution fit to monthly maximum water levels. The team allowed the GEV location and scale parameters to vary in time, as temporal functions (linear, quadratic, exponential, and periodic) or covariates. Applied to tide station records at select sites, this approach enabled the extreme component of the water level signal to be decomposed into various components (e.g., tidal versus non-tidal, patterns versus trends) and be recombined in a way that more accurately reflects how the various contributors can combine to determine extreme water levels at a specific place and time. These analyses were reviewed at a technical workshop held in Honolulu, Hawaii in November 2013, along with a number of other considerations related to best practices and methodologies that can be used to formulate probabilistic estimates of extreme water levels.
Methods were developed to create innovative proof-of-concept products that account for patterns of sea level variability and storminess as well as global and regional trends at specific locations in the Pacific Islands, and that can be used directly to support decision-making. The results confirm that considerable variation occurs from location to location with respect to which and to what extent various contributors are expressed. For most locations, the seasonal cycle and long-term trend account for the bulk of the low frequency time-varying changes in extreme distribution (location parameter). For some locations, low frequency climate variability (i.e., elevated sea level anomalies) also was significant.
Influences of extra-tropical storms are significant seasonally (in scale parameter) at many locations and reveal long-term changes at a few locations. Tropical cyclone events are significant at some locations in terms of the overall characterization of the distribution reflective of its “shape” parameter. Methods similar to those applied to the formulation of the hindcast products created have been applied elsewhere, but not in the Pacific Islands. Their application to the formulation of the forecast products created by this project is particularly novel. Further work is needed to complete the location-specific diagnoses. Specifically isolating impacts that result from prolonged sea level anomalies (location parameter) and changes in typical (extratropical) storm track (scale parameter) need to be explored as does the applicability of customized climate indices.
Development of these methods, along with a review of previous work and expert input on best practices, formed the basis for the creation of guidance that can be used to formulate probabilistic estimates of extreme events under a changing climate for specific locations in the Pacific Islands and beyond. This work suggests that for a given setting the “best” approach to the formulation of probabilistic estimates of extreme events under a changing climate depends on a number of factors. Requirements framing, data reconnaissance, and data assembly are part of the scoping that needs to be carried out prior to data treatment and analysis. Depending on the outcome of the scoping process, one among a set of possible EVA method “cases” may be most applicable for determination of the current extreme event probability. In all cases, a sea level trend and climate sensitivity analysis also needs to be conducted. The results of the various analyses are combined to create estimates of future extreme event likelihood in forms amenable to decision-making.
This work also has revealed that a tendency has been to focus on rare (high magnitude/low probability) inundation events. The behavior of more common (i.e., low magnitude/high probability) inundation events have only recently begun to receive attention. Because these “lesser” extremes (i.e., defined as <5 year event probabilities) are likely to have the greatest cumulative impacts over the coming decades, delineating their expression in a changing climate represents an area where future research and follow-on applications development is warranted.
This project has advanced the practical application of extreme value analysis to inform decision- and policy-making as well as our basic understanding of the factors affecting sea level rise and coastal inundation. DoD can apply the results to improve its understanding of which components of DoD infrastructure are potentially vulnerable to sea level rise/coastal inundation and how they could be affected, as well as how species and ecosystems associated with DoD lands and waters will respond to sea level rise/coastal inundation. The results can (and have been) incorporated into location- and region-specific tools and models. In this regard, efforts have been carried out to transfer these results to other ongoing SERDP projects. This project has expanded the capacity to communicate the risks of sea level rise/coastal inundation, and more broadly the impacts of climate change and climate variability. The results, both the location-specific products and the more general best practices guidance, have broad applicability within the region and the nation. This project also has served to identify issues and opportunities for future work that will further enhance DoD’s ability to understand and predict the potential impacts of sea level rise and coastal inundation under a changing climate.