SEIRb Waning Immunity and Energy Coflow Modeling Insights
The discussion began with a recap of the SEIRb (Susceptible, Exposed, Infected, Recovered, with behavioral response) model, highlighting a couple of items not previously covered in detail.
Handling Waning Immunity: Natural vs. Vaccine-Based
A key challenge in SEIRb models is managing waning immunity when individuals possess both natural and vaccine-driven immunity. The question arises: how should such individuals be routed when their immunity wanes?
Several possibilities exist: - Wane natural immunity only: Send them to a susceptible but vaccinated group. - Wane both immunities: Send them to a completely unprotected susceptible group. - Wane vaccine immunity only: Send them to a recovered group without vaccine protection, with natural immunity waning later.
The critical insight is to avoid modeling two different state transitions simultaneously (e.g., a "double jump" from recovered-vaccinated to susceptible-unvaccinated). Instead, model single jumps, as both types of immunity can wane independently. This approach assumes the two immunity types are independent, each with its own waning rate. Relaxing this assumption would require a more complex model of their interdependence.
The proposed solution involves two independent waning flows: - Vaccine waning: Uses the same time constant as vaccine immunity waning for vaccinated-susceptible individuals returning to unvaccinated-susceptible. - Natural immunity waning: Uses the same natural immunity period as recovered individuals.
This approach ensures internal consistency and simplifies calculations, such as the average time a person remains in the "vaccinated and recovered" state (a harmonic average of the two time constants). Theoretically, vaccine immunity could also wane during the infected period, but this flow is often omitted due to the short duration and the immune system's active response during illness.
Dynamic Fraction of Population Willing to Vaccinate
Another important, albeit minor, accounting detail is the dynamic nature of the fraction of the population willing to vaccinate. While often considered a fixed number, this fraction can change over time. In a disease with a highly protective vaccine and high mortality, individuals unwilling to vaccinate are more likely to contract the disease and die at higher rates. This dynamically increases the proportion of the population willing to vaccinate. To capture this, the model needs to: - Accumulate vaccinated deaths. - Account for unvaccinated deaths. - Calculate the remaining fraction of the population willing to vaccinate, adjusting for these deaths.
Equilibrium Conditions
The equilibrium conditions for the SEIRb model were briefly touched upon. For simpler cases (waning natural immunity, no vaccination, no behavioral response), the setup is straightforward. Solving for cases with behavioral response and vaccination waning is more complex but doable. A shared paper provides guidance on solving equilibrium conditions with behavioral response. For vaccination and waning, it becomes a linear but messier problem, potentially solvable with AI tools like ChatGPT if the problem is clearly defined.
Energy Coflow Models
The main focus shifted to models for energy coflows, specifically tracking energy efficiency with vintage effects, retrofits, and new technology. The goal is to create robust formulations for overall energy consumption given various interventions.
Model Submissions and Feedback
The session reviewed several submitted models, focusing on different aspects of their conceptualization.
Ali's Model
- Vintage Tracking: Buildings are divided into three vintages (stages) with average lifetimes.
- Energy Consumption: Tracks total energy consumption for all buildings, initialized by vintage multiplied by initial consumption.
- Retrofitting: Simplistically assumes retrofits happen only to the oldest buildings. Retrofitted buildings are separated into a distinct stock. Energy consumption is reduced from the original vintage and added to the retrofitted stock with a new, lower consumption value.
- Assumptions:
- Retrofits happen only to the last vintage.
- Retrofitting gives buildings a "new life" (or 1/3 of a new life) with a new energy consumption level.
- Lifetime doesn't change due to retrofitting.
- Technology: New building energy consumption is based on new technology, which is exogenous. It assumes a weighted average with the maximum potential efficiency, accounting for older buildings not achieving new building efficiency.
- Critique: While the logic works with its assumptions, it's not the simplest formulation. A better approach would involve a coflow tracking the potential for improvements for each vintage, allowing that potential to change endogenously.
Jake's Model
- Vintage Tracking: Uses three vintages: new, mature, and old buildings, with exogenous construction rates.
- Energy Usage Coflows: Tracks energy usage for each vintage, similar to Ali's.
- Wear and Tear: Introduces a mechanism where energy consumption increases as buildings age (wear and tear). This is captured by adding "flaws" as buildings transition to the next stage.
- Critique of Wear and Tear: While a meaningful idea, the current implementation (adding all flaws at the beginning of a stage) is not robust to changes in time constants. A more robust approach would be to continuously add flaws per unit of time, creating a continuous flow into the energy usage coflow.
- Retrofit Benefits: Retrofit benefits are applied if the average efficiency of mature buildings is higher than a "lowest feasible efficiency" for that vintage.
- Subscripts: Uses cohort subscripts, creating copies of the model for different cohorts, which is a good way to capture cohort effects.
- Technology: Technological effectiveness of retrofits is subscripted by cohort and influences both new building efficiency and retrofit benefits.
- Critique of Technology: The model convolutes two concepts: how much retrofits can reduce energy efficiency and the benefit per unit of retrofit activity. It doesn't explicitly capture the effect on how far down efficiency can be brought (the lowest feasible efficiency).
- If-Then-Else: The use of
if-then-elsefor defining parameters based on subscripts is unnecessary; direct array-like definition is simpler.
Jason's Model
- Vintage Tracking: Standard three-vintage building stock.
- Energy Coflows: Tracks energy for new, mature, and old buildings.
- Energy Reduction: Introduces a "current energy efficiency factor" for new construction, which sets the maximum achievable energy efficiency for mature and old buildings with a delay.
- Assumptions: New building technology drives energy efficiency improvements, and retrofit technologies follow with a delay.
- Critique: This formulation ties retrofit potential to new building technology, which is an assumption. A more flexible and robust approach would be a coflow that tracks the underlying potential for improvement for each building vintage, allowing it to change based on current technologies.
- Retrofit Decision: Retrofits are modeled by closing the gap between current and minimum energy efficiency over a time constant.
- Critique of Retrofit Decision: This simplifies the decision-making process, omitting considerations of limited resources, cost-effectiveness, and lifecycle benefits.
Gerardo's Model
- Subscripts: Uses subscripts for "building type" (technological segment) and "energy class" (energy efficiency level).
- Retrofitting: Modeled as moving buildings between energy classes (e.g., Class C to B to A), representing discrete steps of improvement.
- Cost and Budget: Includes cost for each improvement step and a budget, influencing the rate of improvement.
- Critique:
- If-Then-Else for Draining: Uses
if-then-elseto prevent stocks from going below zero, which is problematic. The standard solution is to use a first-order limit on the draining rate (minimum of desired rate and stock/minimum time). - Parameter Proliferation: Uses two parameters (cost and sensitivity) where one might suffice.
- Missing Outflows: Demolition rates are zero, meaning buildings accumulate indefinitely, which is unrealistic.
- Cumulative Consumption: Calculates total energy consumption by summing buildings multiplied by energy use per building.
- Technology: Technology advancement exogenously reduces the cost of efficiency improvements, making retrofits cheaper.
- Critique of Technology: This model focuses on reducing retrofit costs but doesn't capture the ability to achieve lower energy consumption levels (e.g., Class A buildings becoming even more efficient due to new technology). It also assumes all buildings can eventually reach Class A, which might not be feasible for older vintages.
- If-Then-Else for Draining: Uses
Hassam's Model
- Subscripts for Vintages: Uses subscript ranges for vintages (V1, V2, V3), allowing for easy expansion to more vintages.
- Retrofitting: Allows retrofitting across all vintages, moving buildings into a "retrofitted area" stock.
- Technology for Retrofits: Assumes new building technology sets the "gold standard" for retrofitting. The new retrofit level is a weighted average between current efficiency and the best achievable.
- Critique: The current formulation for retrofit technology hardcodes the limit, not allowing dynamic changes based on new technologies. A third coflow for underlying potential would be beneficial. The subscripting for vintages is a good approach for scalability, though the specific implementation could be more generalized using Vensim's subrange mapping features.
Anne's Model
- Subscripts for Vintages: Uses "year" as a subscript, which can be confusing and potentially lead to Vensim errors.
- Single Building Stock: Uses one stock for buildings, with vintages captured by subscripts.
- Energy Inefficiency: Uses "energy inefficiency" as a variable, which is conceptually the inverse of efficiency and can be confusing. Units are also inconsistent.
- Retrofit Potential: Retrofit potential is derived from "maximum possible efficiency" (determined by technology) and the gap between current and maximum efficiency, closing the gap over a time constant (similar to Jason's).
- Vintage-Specific Potential: Captures different retrofit potentials for different vintages through parameters, which is a good feature.
- Critique: The confusion between efficiency and consumption, along with unit inconsistencies, can lead to weird behaviors. The model lacks multiple life stages for buildings, which is a simplification.
General Feedback on Energy Coflow Models
- Underlying Potential for Efficiency: Most models are missing a separate coflow to track the underlying potential for enhancing energy efficiency for each vintage. This coflow would dynamically change with investments in technology, allowing for a more robust and flexible representation of how far efficiency can be improved.
- Subscripting for Scalability: Using subscripts effectively (especially with Vensim's subrange mapping) allows for easily adding or removing cohorts/vintages without rewriting equations.
- Investment and Allocation: Many models simplify retrofit decisions to time-constant-based gap closure, omitting crucial aspects of resource allocation, cost-effectiveness, and lifecycle benefits. A more rational formulation would consider these factors.
- Perfect Mixing Assumption: All single-stock models inherently assume perfect mixing. To introduce heterogeneity, disaggregation into multiple stocks is necessary. Higher-order delays disaggregate implicitly but limit direct manipulation of intermediate stocks.
Predator-Prey Model with Carrying Capacity (Next Week's Assignment)
The next assignment involves building a predator-prey model that includes carrying capacity for the prey, representing a more complex version of the traditional Lotka-Volterra model.
Problem Definition
- Policy Application: Understanding ecosystem sustainability, managing pests (e.g., deer eating crops, causing accidents), or threats to wildlife. This model helps understand ecosystem dynamics before interventions.
- Reference Mode-Based: Explaining oscillatory patterns in population sizes observed in real-world data (e.g., fish populations).
- Mechanism-Driven: Investigating behaviors arising from specific mechanisms: two populations interacting (one eating the other), their birth/death rates, and the prey's food source (vegetation) regenerating itself. This approach aims to explore various reference modes (oscillations, overshoot and collapse) generated by these mechanisms.
Stock Variables
- Predators: One or maximum two stocks (e.g., to include age groups for additional delays).
- Prey: One or maximum two stocks.
- Vegetation/Resource Capacity: One or maximum two stocks (e.g., to represent different growth stages of trees). Starting with one stock is recommended for simplicity.
Performance and Robustness Tests
- Recreate Expected Reference Modes: Test if the model can generate oscillations, overshoot and collapse, etc., by introducing shocks.
- Inputs to Change:
- Reproduction rates of each species.
- Initial states of populations.
- Hunting (policy lever or shock).
- Changing land availability/environmental encroachment (shock).
- Disease (exogenous shock).
- Outputs to Track:
- Populations of all species.
- Carrying capacity.
- Robustness Conditions:
- Extreme conditions of population (e.g., taking out or adding a significant portion of a population).
- Making carrying capacity a non-binding constraint.
- Adding one species at the beginning of the simulation (initial states).
Modeling Approach
- Iterative Development: Build the model in iterations:
- Model the prey population alone (infinite resources).
- Add the predator, creating a two-species model without carrying capacity.
- Add the carrying capacity (vegetation) for the prey.
- Testing at Each Step: Ensure the model behaves correctly at each iteration.
- Unit Consistency: Maintain unit consistency throughout the model.
- Analysis: Analyze one interesting scenario from the final model, setting parameters to generate a specific behavior (e.g., oscillations). Understand the underlying structures driving that behavior at different points in time.
The carrying capacity for the deer refers to the vegetation they eat. This food source is not fixed but rejuvenates itself, requiring a stock variable (or two) to model its dynamics.
Takeaways
- The SEIRb model should treat natural and vaccine immunity waning as separate, single‑step flows to avoid double‑jump transitions and keep the system internally consistent.
- The fraction of the population willing to vaccinate is dynamic; deaths among the unvaccinated increase the willing proportion, so the model must track vaccinated and unvaccinated deaths to update this fraction over time.
- Energy‑coflow models for building stocks generally lack a dedicated coflow for the underlying efficiency potential, and adding such a coflow improves flexibility by allowing technology investments to dynamically shift achievable efficiency for each vintage.
- Robust retrofit modeling should incorporate cost, budget limits, and resource allocation rather than relying solely on time‑constant gap‑closing, and subscripting vintages enables scalable, cohort‑specific analysis.
Frequently Asked Questions
Why does the SEIRb model avoid a double‑jump transition when immunity wanes?
The model avoids a double‑jump because allowing both natural and vaccine immunity to wane simultaneously would create an unrealistic instantaneous transition from a recovered‑vaccinated state to fully susceptible, breaking the assumption of independent waning processes. Modeling each immunity type with its own single‑step flow preserves mathematical consistency and makes the system easier to analyze.
How does tracking a separate underlying efficiency potential coflow improve building energy models?
Tracking a separate underlying efficiency potential coflow lets the model capture how technology investments and retrofits expand the maximum achievable efficiency for each building vintage, rather than fixing a static efficiency ceiling. This dynamic coflow updates the potential over time, enabling more realistic simulation of retrofits, vintage effects, and policy impacts.
Who is MIT OpenCourseWare on YouTube?
MIT OpenCourseWare is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.
Does this page include the full transcript of the video?
Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.
arises: how should such individuals be routed when their immunity wanes? Several possibilities exist: - **Wane natural immunity only:** Send them to
susceptible but vaccinated group. - Wane both immunities: Send them to a completely unprotected susceptible group. - Wane vaccine immunity only: Send them to a recovered group without vaccine protection, with natural immunity waning later.
Helpful resources related to this video
If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.
Links may be affiliate links. We only include resources that are genuinely relevant to the topic.