-
-
Control of Hypertropical Cyclone Formation (note the wind and pressure vortex)
-
Cloud Seeded Intervention of Cyclone Formation (minimal to no formation of vortex)
-
Control vs Cloud Seeded Drought Intervention (note that this uses electric ionization seeding and gave a %16 precipitation increase)
-
Comparison of 4 Different Types of Interventions in Same Drought Prone Area (electric ionization vs glaciogenic vs hygroscopic vs laser)
In the past 5 years, climate adaptation has skyrocketed from the side show of mitigation to an integral part of society’s climate survival plan. Geoengineering, the process of purposefully manipulating climate systems, has been a source of much controversy due to its immense potential to change the way the world views our control of climate, and its theoretical downsides or cascading effects.
While long-term, global solutions like Solar Radiation Management (SRM) are the subject of much debate, we see much more value in techniques that are being scientifically validated on a local scale such as rain/cloud seeding, used currently to clear smog and disperse fog. These technologies, including electric ionization (high voltage electrons shot into clouds), glaciogenic (silver iodide for bergeron processes in supercooled clouds), hygroscopic (large salt particles as warm cloud condensation nuclei), and laser induced condensation (femtosecond lasers to create hygroscopic oxidative byproducts) seeding, have immense potential to mitigate droughts and even reduce the intensity of tropical cyclones.
However, current players in the space are hyperfocused on modeling to prevent negative impacts, with extremely limited research and governmental cloud seeding missions.
Morro Atmosphere solves this by, for the first time, creating a comprehensive optimization system for geoengineering aimed at prevention. A governance-embedded climate intervention simulator, we integrate multiple different AI weather models and types of geoengineering to predict and prevent natural disasters before they start to harm societies.
Our initial MVP focuses on two main disasters: droughts/heat domes and tropical cyclones. Our inputs into the two weather models we selected, Google Deepmind’s Graphcast and Nvidia’s Earth2Studio, are ERA5 state vectors or a snapshot of earth’s atmosphere with 40+ environmental variables across pressure (vertical) levels on a 1 degree latitude/longitude grid. Both AI weather models are trained on 40+ years of data and each have a mesh of 40K+ nodes. Each takes 2 initial timestamps (current time and 6 hours before) to run a prediction that regressively predicts for 10 days ahead.
For each specific type of geoengineering, we create a mask based on all the verified scientific research available on the specific method of cloud/rain seeding. The mask is an additive perturbation that is applied to the initial state of either autoregressive model, which then cascades through the timesteps and predicts the impact of this intervention. Each mask is calibrated to the dose of the intervention, as well as the current atmospheric conditions, and generates perturbations based on the thermodynamics identities of atmospheric physics.
Additionally, we developed Random Forest Classifiers that are able to predict drought/heatdome severity (using the Composite Drought Severity Index (DSI)) or the existence of a tropical depression (using pressure) at a given timestamp. This analysis lets us know where to apply geoengineering methods in concert, and we generate predictions for all of the possible interventions, eventually returning those that are most effective while minimizing negative externalities.
In terms of societal impact, Morro can cause a cyclone to break 2-3 miles earlier or dissipate before reaching the coastline, potentially mitigating millions of losses in infrastructure and property damages while simultaneously protecting human life. Similarly, Morro can spot the preconceptions of droughts or heat domes and alleviate the environmental and human impacts of these events.
Why tackle this challenge now of all times? Well, we face a few challenges with this technology currently: 1) the best forecasting models are only accurate up to 10 days in advance (which is a technical limitation), 2) there is no training data on how to teach a model to cloud seed (another technical limitation), and 3) that cloud seeding is an incredibly controversial topic in environmental sciences right now. Our project addresses these three issues by using novel physics-based model techniques built on top of existing forecasting models to accurate predict, mitigate, then demonstrate how cloud seeding (when done correctly) can prevent weather-related disasters.
We are at a unique period in the field of research on cloud seeding in the sense that there are preliminary studies that demonstrate how cloud seeding the region within or around a forming hurricane or cyclone can reduce its wind speeds and lifespan or in some extreme cases can even stop it from forming all together (as our model prediction showed). However, governance is an extreme limitation on this technology as the only two places that have mostly adopted the potential positives of it seem to be the UAE and China (yet they mostly remove it for smog removal). Our technology and software isn't meant for right now, it is a prediction based on a 5-10 year timeline in which we anticipate governments realizing how critical it could be economically and ecologically to manipulate our atmospheres to protect agriculture and to stop potential urban destruction via hurricane mitigation.
Morro is designed for decision makers: drought agencies, multinational meteorological bodies, and environmental regulators that are charged with safe-real world application of these technologies. We provide critical intelligence throughout the adoption pathway, from simple simulation sandboxes to scientific advisory reviews for initial research collaborations to testing of different regulatory scenarios to large-scale deployment of a suite of geoengineering tools. Morro's models are also aimed at evaluating negative cross-border impacts, choosing localized interventions that are not only best suited for the task at hand, but that also minimize risk throughout the region. This allows for environmental justice concerns to be considered with every decision, and helps build up the data needed to establish liability frameworks.
In terms of future improvements to this project, there is one major thing. Right now current models for cloud seeding and optimized and trained to minimize harmful effects, however ours is trained to maximize positive effects. This has both its advantages and disadvantages, ideally with more data and stronger forecasting models (which are coming out at high frequency now as research accelerates) we will be able to solve both the negative consequences of improperly implemented cloud seeding while optimizing even more powerfully the potential environment saving possibilities.
While we had the advantage of implementing open-source, high quality weather models, we faced many challenges regarding how to incorporate modeling of extremely novel geoengineering techniques and are particularly proud of how our solution accounts for the many extraneous factors that can impact effectiveness while delivering a solution with strong reduction in damaging climate conditions. Overall, we see immense potential to expand our MVP to cover more dangerous climate phenomena and incorporate the most advanced research in geoengineering, creating a framework that is not only ready for deployment by companies and governments trying to take back control from rampant climate change, but also is effective in pushing forward legislation that allows for the advancement of geoengineering to meet the challenges of a warming world. Natural disasters are only getting more frequent, more intense, and more unpredictable, but Morro Atmosphere is ready to meet the challenge.
Log in or sign up for Devpost to join the conversation.