Since there are no currently active contests, we have switched Climate CoLab to read-only mode.
Learn more at https://climatecolab.org/page/readonly.
Skip navigation
Share via:

Pitch

Developing a Physics Guided Statistical Model to provide actionable, probabilistic climate forecasts to the public and private sectors


Description

Summary

Climate change is adversely impacting global ecosystems and a major concern for societies around the world. While many groups are doing their best to adapt to and mitigate the effects of climate change, often the measures taken “are not strong enough to avoid loss and damage when the situation is aggravated by an extreme weather event" according to a report released by United Nations University. This is because many climate change projections are not in a format usable by stakeholders (i.e., deterministic, decadal/centennial, averages instead of extremes). Yet stakeholders are still faced with making design and planning decisions in light of climate change.

To address this issue, our team, using a Physics Guided Statistical Model (PGSM), will translate climate change projections into concrete data that quantifies the climate-change threat today. This data will allow climate change risks to be embedded in the day-to-day decision making of stakeholders, allowing adaptation initiatives the credibility and urgency to compete for required resources. Measuring how climate-related risks are evolving today is the first step in building the adaptive capacity of organizations and communities. While this methodology is widely applicable, in this proposal we will demonstrate how the PGSM applied to extreme weather events is a novel, feasible and high impact approach to improve adaptation across the globe. In our action plan, we will demonstrate how tailoring these outputs for insurance companies will cause cascading effects across all sectors.


Category of the action

Mitigation/Adaptation, Changing public attitudes about climate change


What actions do you propose?

The Landscape Today

Businesses and public agencies rely predominantly on historical weather data when developing risk-mitigation strategies for climate adaptation purposes [Niehörster et al. 2013]. However, with growing evidence for climate change-driven increases in extreme weather risks events [Min et al. 2011; Emanuel 2013; Ganguly and Kodra 2014; Ganguly et al. 2009], historical data is becoming less useful in forecasting the future and the need for climate change informed extreme event forecasts is increasing. Incorporating global climate model (GCM) predictions into extreme event forecasting is difficult at best due to uncertainties like intrinsic variability, longstanding gaps in physics understanding, and differences in socioeconomic trajectories. Current methods for assessing climate change related risks do not provide distributions necessary for decision making.

Through an intersection of climate science domain knowledge and expertise in statistical modeling, our Physics Guided Statistical Modeling (PGSM) framework converts the multitude of data sources and complexities of climate change into probability distributions for weather extremes at decision-relevant scales. These distributions will be used by many stakeholders to make climate change informed decisions. For example, at risk communities can use these predictions to implement stronger measures against extreme climate events.Furthermore these communities could receive economic support from insurers who would be able to directly use the data in their actuarial models.

At the core of the PGSM framework, known physical relationships are used to guide the development of a Bayesian statistical learning architecture. Probabilistic outputs of the model are constrained by known physics and are designed to quantify uncertainty in future extreme events.

The PGSM framework has been found to perform accurately in multiple forecast tests against out-of-sample, real-world observed extreme precipitation events. Scientifically, this is a significant advance beyond modeling techniques available in peer-reviewed literature. The ability to utilize robust, probabilistic information on climate change and its effects on extremes will be a fundamental shift in risk assessment processes across many industries.

Actions Proposed

risQ will initially provide climate change informed inputs for insurance companies’ risk models. By demonstrating that insurance losses are occurring now and and in the near future due to climate change, insurance companies will be shown a direct link to climate change and their bottom line.

With competencies in risk management and finance, insurance companies are uniquely positioned to improve adaptations to climate change. They are also uniquely motivated: weather related damages have increased to nearly $130 billion per year and are considered a major threat to the industry. The Association of British Insurers (ABI) and other groups have released publications urging insurance groups to actively pursue climate change solutions. 

By providing data in a format that can be input into actuarial and catastrophe models, insurance companies will be able to provide new financial instruments that take into account climate change adaptations. For example, Munich Re provides micro-insurance to small farmers in developing nations. Better data would give other groups the ability to provide similar or even more innovative products.

Insurance companies can also cause cascading effects other than providing insurance. For example, insurance companies have played a large role in implementing building codes that minimize damage due to fires and earthquakes and could play a similar role in creating new building codes that take into account climate change associated risks. This could require infrastructure owners and civil engineering groups to develop and implement measures against climate change related disasters. Insurance agencies are also major players in the financial markets and have invested $6 billion in green investments. With better information about climate change impacts, insurance companies are also a potential source of investment in climate change adaptation. It is also of business interest for insurance companies to push for policies that reduce climate change and overall risk.

With the ability to accurately assess and predict damages caused by climate change, insurance companies will be able to create new innovative insurance policies and push for regulatory policies that will cause a cascading financial incentive to implement climate change adaptations.

What is the difference between what risQ is doing and current methodology? Is there a practical difference?

  • Currently there are many climate models each projecting a different outcome. For decision makers, this is a serious problem: which model’s forecast should be used? Or would an average be better?  And even more pressing, many decision makers (particularly in the insurance industry) require probabilistic data. If they select a model (or average several) what are the chances of the model being wrong and how wrong can it be? Without this data it’s impossible to incorporate climate change predictions into the tools of many industries.

  • In the image above different model projections are represented in red and the observed data is in black. risQ's model is able to take these different projections and combine them into a probability band (represented in blue). While risQ's forecasts more accurate, more important is the fact that we can generate probability data at all. 

  • risQ’s methodology is the only one that can take global climate models -- all of which project different things about the future - and consolidate those different projections into a concise and accurate probability distribution. These distributions are incredibly important for decision making. We have received letters of support from AIG, Munich Re, AIR Worldwide, the City of Boston and other groups who recognize the importance and utility of the data risQ is providing.

Why target the insurance industry?

  • risQ is ambitious. We want to enact systemic changes that will address climate change issues. By providing the insurance companies the knowledge and tools to understand the costs of climate change, we believe that the full financial and political power of the industry will be brought to bear on solving climate issues. Insurance companies are already taking actions toward climate change. In 2011, companies invested an estimated $23B into climate mitigation (projects such as clean energy).

How would this impact the developing world?

  • We agree that this data would not be directly usable by the developing world. However as previously mentioned, by pushing the insurance industry to directly address climate change we believe there will be cascading impacts across the global.

    One clear example is microinsurance. Microinsurance is currently provided to ~135 million people in the developing world, an estimated 5% of the total coverable. Companies like Partner Re, Swiss Re and Munich Re are becoming more and more active in the market (Munich Re is one of the insurance companies that has come out in support of risQ’s data). Furthermore in an insight published by Lloyd’s, many companies are also moving to “consider the related issues of adaptation, recognising that the effects of climate change related catastrophes may be better addressed through adaptation."

What else can be done with this data?

  • Directly selling this data is only a small portion of risQ’s goal of combating climate change. We hope to develop a suite of products using our methodology. One that is planned for the next 2-5 years is using the probabilistic data to generate flood projections that take into account climate change.


Who will take these actions?

We have been actively engaged with:

  • City of Boston

  • Boston Resilience Advisory Group

  • DHS

  • Public/private sector organizations

who will be able to use the publications and software developed to significantly advance best practices in climate change adaptation. 

Evan Kodra is an applied statistician with a core focus on developing best-of-class quantitative approaches in climate change, weather extremes, and resilience that can be used by practitioners. He holds a Ph.D. in Interdisciplinary Engineering from Northeastern University’s Sustainability and Data Sciences (SDS) Lab as well as a B.S. and M.S. in Statistics from the University of Tennessee. His research on climate and weather extremes and natural hazards has been published in journals including Nature Publishing Group’s Scientific Reports [Kodra and Ganguly 2014], Geophysical Research Letters [Kodra et al. 2011], PLOS ONE [Bhatia et al. 2015], Computers and Geosciences [Parish et al. 2012], Nonlinear Processes in Geophysics [Ganguly et al. 2014], and Environmental Science & Technology [Kodra et al. 2015]. 

Auroop Ganguly will assist in future development of risQ's products. He is the Principal Investigator of the SDS Lab and Associate Professor of Civil and Environmental Engineering at Northeastern University. Ganguly has a PhD from the Massachusetts Institute of Technology as well as 17 years of experience spanning academia, a government-owned national research laboratory, and the private sector. He is the Co-Chair of the Societal Dimensions Working Group for the Community Earth System Model.

Colin Sullivan's experience in business development and technology transfer will assist in the careful tailoring of products to meet stakeholder needs. Sullivan leads the teams stakeholder outreach and engagement. With a BS in Computer Engineering from Northeastern University and years of experience in technology transfer at both Northeastern and Yale, Sullivan has experience to translate research into technical software products.


Where will these actions be taken?

While the initial focus is in the United States, the data products should be applicable worldwide. We have been actively engaged with many European private sector groups and in India.


What are other key benefits?

Engineering firms must develop climate-informed engineering design parameters and building codes due to increased urbanization, coastal inhabitancy, and climate impacts. Intensity-Duration-Frequency (IDF) characteristics of extreme rainfall have long been used for water management and civil engineering planning. Reliable IDF statistics are also crucial as key drivers of flooding and storm surge. Organizations including the Massachusetts Department of Transportation and the National Oceanic and Atmospheric Administration currently provide IDF data - but only based on historical precipitation data. We will partner with engineering design firms who can translate our IDF data to engineering design parameters that reflect current and future climatic risks so as to help their clients adapt to climate change.


What are the proposal’s costs?

risQ plans to sell data products to stakeholders in order to continue development of its data products. R&D costs to bring an preliminary product to market are expected to be $225,000 and the team is currently in the process of securing these funds.


Time line

To date we have we have achieved major milestones in this vision:

  • Used PGSM to develop probability distributions for regional precipitation extremes for the United State’s HUC4 regions

  • Developed relationships with stakeholders in insurance and other areas such as catastrophe modeling and critical infrastructure

In the near future (1-2 years):

  • Complete development of data inputs to insurance stakeholders to allow them to create new and innovative risk management instruments

  • Publish whitepapers on the impacts of climate change to target markets. These white papers will prove useful for industry, as well as to researchers within academia, in assessing how to factor climate change in their respective practices

  • Develop a statistical downscaling model that will generate Intensity-Duration-Frequency (IDF) curves for local weather stations

  • Work closely with stakeholders to develop new data products that will embed climate change in day-to-day decision making.


Related proposals


References

Warner et al (2012). Evidence from the frontlines of climate change: Loss and damage to communities despite coping and adaptation. Loss and Damage in Vulnerable Countries Initiative. Policy Report. Report No. 9. Bonn: United Nations University Institute for Environment and Human Security (UNU-EHS)

Mills, E.  (2007). Responding to climate change – THE INSURANCE INDUSTRY PERSPECTIVE.

Lloyd’s 360 Risk Insights (2014). Insurance in developing countries: Exploring opportunities in microinsurance.

Ganguly, A. R., Kumar, D., Ganguli, P., Short, G., & Klausner, J. (2015). Climate Adaptation Informatics: Water Stress on Power Production. Computing in Science & Engineering, 17(6), 53-60.

Ganguly, A. R., et. al (2013). Computational data sciences for actionable insights on climate extremes and uncertainty. Computational Intelligent Data Analysis for Sustainable Development, 1127.

Ganguly, A. R., Kodra, E. A., Agrawal, A., Banerjee, A., Boriah, S., Chatterjee, S., ... & Wuebbles, D. (2014). Toward enhanced understanding and projections of climate extremes using physics-guided data mining techniques. Nonlinear Processes in Geophysics, 21(4), 777-795.

Geneva Association. (2014). Warming of the oceans and implications for the (Re)insurance industry. The Geneva Reports, 5(1), 109-146.

Giorgi, F., & Mearns, L. O. (2002). Calculation of average, uncertainty range, and reliability of regional climate changes from AOGCM simulations via the “reliability ensemble averaging”(REA) method. Journal of Climate, 15(10), 1141-1158.

Gleckler, P. J., Taylor, K. E., & Doutriaux, C. (2008). Performance metrics for climate models. Journal of Geophysical Research: Atmospheres (1984–2012),113(D6).

Katz, R. W., Craigmile, P. F., Guttorp, P., Haran, M., Sansó, B., & Stein, M. L. (2013). Uncertainty analysis in climate change assessments. Nature Climate Change, 3(9), 769-771.

Knutti, R., Furrer, R., Tebaldi, C., Cermak, J., & Meehl, G. A. (2010). Challenges in combining projections from multiple climate models. Journal of Climate, 23(10), 2739-2758.

Knutti, R., & Sedláček, J. (2013). Robustness and uncertainties in the new CMIP5 climate model projections. Nature Climate Change, 3(4), 369-373.

Kodra, E., Chatterjee, S., & Ganguly, A.R. (2011). Challenges and opportunities toward improved data-guided handling of global climate model ensembles for regional climate change assessments. ICML Workshop 2011, Seattle, WA, https://sites.google.com/site/mlforglobalchallenges/overview.

Kodra, E., Ghosh, S., & Ganguly, A. R. (2012). Evaluation of global climate models for Indian monsoon climatology. Environmental Research Letters, 7(1), 014012.

Kodra, E. (2014). Addressing gaps in climate science: hypothesis and physics guided statistical approaches. Northeastern University, Ph.D. Dissertation

Kumar, D., Kodra, E., & Ganguly, A. R. (2014). Regional and seasonal intercomparison of CMIP3 and CMIP5 climate model ensembles for temperature and precipitation. Climate Dynamics, 43(9-10), 2491-2518.