Assessing Future Flood Risk oward a Sustainable City and Campus Stormwater and Landscape Ecology Plan: A Cambridge and MIT Case Study

In 2014, The MIT Office of Sustainability created a campus sustainability working group charged with the development of a Sustainable Campus Storm-water and Landscape Ecology Plan. The Plan will make recommendations and provide guidance for future projects that collectively work to slow and reduce storm-water runoff, reduce impervious surfaces and improve water quality, decrease thermal pollution of waterways, prevent or alleviate localized flooding, and contribute to improving the ecological health of the campus and the lower Charles River watershed. An integral part of these efforts has been the development of a comprehensive and detailed street/building level MIT Campus Storm-water Model (CSM) that has the ability to assess flood risks that may be substantially heightened under human-forced changes over the coming decades. In work thus far, we demonstrate through the use of design-storm numerical experiments that the CSM research identifies and corroborates the details that short-duration storms of heavy-to-extreme precipitation (inches of rainfall per hour), which typically occur during the summertime represent a salient threat from pluvial flooding (for example, the July 10, 2010 Kendall Square flood). These events are characteristically different from other meteorological and hydrologic events in that they typically occur with very little lead time to prepare and fortify structures - as well as protect students, staff, faculty, and the public-at-large. Most notably, however, it is the changes to the occurrence, shape, and intensities of these variety of storms that could significantly impact the future risks to students and staff, current and developed infrastructure as well as the research and education assets they contain and provide. Therefore, from a climate-science perspective, the underlying question is whether prediction methods can provide actionable piece of information to convey a changing threat or risk of these precipitation events that cause flooding. If so, the issue then becomes whether to fortify, upgrade, and/or upgrade existing infrastructure. The proposed work expands upon an “analogue method” to provide targeted estimates of shifts in these heavy-to-extreme events (i.e. summertime precipitation) – and for this particular effort – over the MIT campus and greater Cambridge watershed. Therefore, our overarching research question is: What is the expected change in summertime precipitation events that currently cause disruptive and/or damaging flooding conditions?

Typically, extreme events are evaluated based on one indicator variable in a univariate framework (e.g. drought based on deficit in precipitation, extreme temperature based on high quantiles of temperature data). A number of recent studies have emphasized that a multivariate framework is necessary for assessing risk of extreme events, especially in a warming climate. Regardless of traditional methods or multivariate copulas, the projections of univariate extremes future conditions rely on the quality of future climate trends simulated from climate models. While models have improved, challenges remain in correctly simulating the statistical properties of meteorological variables, and therefore their extremes, especially at regional to local scales. In our own analyses and relevant to this study's focus, we find a large scatter in the climate models’ climatology and future projections of heavy precipitation events – based on their simulated precipitation rates for a location or region of interest. However – an analogue method that we have recently developed which determines the “telltale signs” at the large meteorological atmosphere-scale for the occurrence of an extreme event at the local scale, has shown to improve upon the model performance and consensus of any changes in extreme event occurrences associated with climate warming. This method has also shown weak dependence on model resolution but present similar or better skill levels than current implementations of Regional Climate Models (RCMs). Also, the method is non-parametric in that it does not make any assumption about the distribution of the atmospheric patterns or the extreme event statistics.

Therefore, in conjunction with the CSM simulation framework, we present our results of both the evaluation and projection of change in summertime, flood-producing storm events for the MIT/Cambridge area of study. From a meteorological perspective, we consider a number of historical events that have occurred across the greater Boston metropolitan area - and these have been associated with the synoptic-scale conditions and phenomenon. Through the use of the MERRA2 reanalyses data as well as other corroborating reanalyses sources (e.g. ERA, NARR, JRA, etc.) we have determined a set of cogent and large-scale atmospheric features that are associated with these localized events. These patterns show consistent features in the synoptic environment associated with precursory and prevalent moisture fetch as well as vertical motion and convective instability. From these, we have employed machine-learning methods to train a predictive tool and demonstrate that when driven by information from the CMIP model collective, we can improve the skill and inter-model consensus of reproducing the occurrence of these events for the contemporary/historical climate. When apply this predictive approach to future projections of climate - that span the range of potential human-forced climate change under the Representative Concentration Pathways (RCPs) simulation framework - in order to quantify the changing risk of these storms to our CSM study area and design-storm assessment of flooding across the MIT campus and Cambridge.