AI Climate.

In 2018, the World Bank estimated that almost 30% of the world’s population lived in informal settlements. These communities are expected to expand rapidly in the Global South (World Bank 2019), especially in secondary cities and regions, which are the gateways for rural-to-urban migration and displaced populations. As a result, these urban areas, where about 75% of the world population lives (Cities Alliance, 2018) are undergoing fast, informal-settlement formation. 

Concurrently, the escalation of climate change impact is generating more extreme risks for these expanding communities (Intergovernmental Panel on Climate Change, 2021). These combined forces demand more proactive and efficient resiliency strategies to prevent climate hazards from becoming disasters. Yet secondary cities and their regions lack the resources for this task.

Ai Climate targets four problems:

First:  Parallel-to-climate-change, cities in the global-south are dealing with rapid-urbanization and expanding-informal-settlements. 

Second:  National laws and policies for climate change adaptation are necessary but not sufficient to change the situation on the ground. The link to the ground are land-policies and fiscal-policies, concentrated-on-the poor and that use land-as-a-financial-mechanism for resilience and-resources-building.

Third:   Many cities, particularly secondary-ones-that-are-gateways-to-rapid-urbanization, lack the resources and local-data-granularity to shape-growth and build-resilience.

Fourth: Current-planning-tools-and-data-collection-methods are slow and costly. By combining remote-sensing-with-machine-learning, we can accelerate data-collection and reduce-costs.

Addressing rapid-urbanization and climate-uncertainties, requires efficient-land-management-tools.  We were able to work in urban areas with medium and higher resolution images for better local data granularity. All models’ ready-to-use layers are integrated in one public platform and integrate information in order to formulate integrated policies and plans. 

We have identified ML/CV as our opportunity to address the lack of efficacy for resilience-planning and the slow pace of adaptation to rising climate risk in secondary cities and regions.

Ai Climate is a land-management and decision-making tool for building-climate-resilience and spatial-inclusion. It uses machine-learning to map, monitor, and predict-exposure of informal-settlements to weather-impacts. 

Technically AIClimate, is an open-source-learning-platform, hosted in the public domain,  that uses geospatial images and georeferenced datasets to derive analytics-ready-layers of impacts associated with climate change. 

We developed AI Climate in Honduras, applying it in Tegucigalpa and the San Pedro Sula Region(11 municipalities). Honduras was chosen due to its high climate change risk, rapid urbanization, and expanding informal settlements. Tegucigalpa is prone to landslides, and the Sula Valley to extreme flooding.

Our design framework for AI Climate is based on five main points:

1.  “Simplify as much as possible.” 

2.  Fast and easy to update

3.  "Perfect is the enemy of good" – prioritizing low-cost-solutions

4.  Viewer is publicly-accessible and open-source

5.  Vertically-integrated-process, linking technical-partners to informal-communities.

We developed 4 models in Honduras, Central-America: M1Informal-Settlements, M2Flooding Susceptibilities, M3Landslides Susceptibilities and M4PoC Land-values-differentials; we integrated them in-one-public-visor.

Building on this experience we plan to:

1-Improve existing and expand to 6 models in Honduras.

2-Cultivate an AIClimate mindset in Honduras among public officials and communities in the Sula Valley through a small regional learning center for AI/ML for capacity-building of local and regional technicians for technology transfer.

3-Scale the model for informal settlements and flooding to three countries in Central America: Honduras, Guatemala, El Salvador. Afterwards, complete Central America. 

4- Develop/scale in two countries one in West and other in East Africa. We have secured local partnerships.

Total proposed Models under this grant are: improve existing, M1-M4-  and develop new: M5-SolidWaste; M6 Land Degradation/Heat Island

Paradigm changes: 

*Reduce urban planning time and cost, leading to effective risk reduction update frequency that does not exist at present and which is crucial at a time of rapid urban-growth and rapid change in climate-pattern. 

*Able to work in urban areas, that lack local data,  using free sentinel 2 satellite images or better if we choose Planet labs discounted price, offering the cities quick a working assesment of growth and climate impacts for climate action and inclusion strategies.

The vertical integration  process:

Our international-team strongly believes that vertical-integration is a powerful-tool for inclusion, information-re-distribution, and community-empowerment. In Honduras we have gone through this specific process:

During the vertical integration process, our technical-partners-Planet-labs-and-Dymaxion-Labs, initiated the modelling process, while local partners such as Instituto Hondureno de Ciencia de laTierra, Habitat for Humanity-Honduras, and Goal Honduras provided local-ground-truth and expert-validation. Then Informal-communities validated the results, which were used to retrain the machine. I2UD created and managed the project, receiving funding and technical-support from the MacGovern Foundation Accelerator Grant. 

Ultimately, to-date AI Climate combines four models enabling planners, stakeholders and communities:  

• Action on local-climate-impacts and protect IS from weather impacts . 

• Identification of safe-land for guiding-growth or land-banking strategies. 

• Monitoring, assessing  and measuring adaptation-strategies outcomes.  

• Quick estimation of disaster-costs-and-land-value-creation (or value capture to fund resilience strategies). 

Finally, AI Climate aims to promote climate resilience, inclusion, and sustainability, ultimately-protecting-lives-and-ecosystems.

Ai Climate serves specifically Informal settlements (IS) in cities that do not have data granularity, no budget/ resources,  especially those that are undergoing both climate impacts and rapid urbanization, including informal settlements formation. Ai Climate is a tool that aids public officials and communities at the same time.

Today, mapping Informal settlements  without ML, requires months to update information and on-the-ground-personnel/volunteers for all AOI (area of Interest) having limited access to key areas inaccessible due to budgetary, geographic, or security constraints.and security issues. 

Flood/landslide planning requires expensive lab-modeling. Land Values need rigorous statistics and calculations. Consequences include outdated GIS information and undisclosed and disjointed information for planning. Thus no plan and no action. 

There is no affordable integrated tool that allows for a quick visualization of  both moving targets: informal settlements and their growth;  and Climate patterns and it's changes. 

This tool identifies local climate impacts seeking to  protect informal settlements communities.  It also identifies safe land for guiding growth and land value capture strategies for resilience building funding. I can also monitor assess and measure the development of resilience and adaptation strategies.   Additionally, it supports planners with swift calculations of disaster costs and land-value-creation estimation ex-ante interventions.  

We know we can work effectively as a team because we already have. We co-created AIClimate Platform through vertical integration of all partners, coordinated by I2UD through structured meetings, project tasks, and clearly defined roles. Equally important, each member has demonstrated the collaborative spirit required to get past challenges in complex projects.

The organizational structure of the project integrates a multidisciplinary team along a vertical integration of co-production, from the sourcing of images to validation of results with local communities. This process knits together a variety of organizations, from suppliers of satellite images to local communities. We will continue to build AI Climate in collaboration with a variety of field-partners and technical experts that provide the required data and knowledge of field conditions, validate results, and to whom we can devolve the information gathered in support of their own strategic initiatives.

Our international-team strongly believes that vertical-integration is a powerful-tool for inclusion, information-re-distribution, and community-empowerment. In Honduras we have gone through this specific process:

Recently-in-collaboration-with GOAL-Honduras,-USAID-and I2UD, Habitat-for- Humanity-Honduras hosted a forum to present-AI-Climate to local-mayors and-stakeholders of-the-Sula Valley.  (11 Municipalities just in the Sula Valley plus the Tegucigalpa greater Region)

At the forum, we agreed to explore-the-creation-of-a-regional-AI-learning-center and enhance AiClimate adoption and technology-transfer.

We are aware of how important it is that the Municipalities adopt and grow AI Climate - This is our main goal.  We can't accept the idea of AIClimate staying locked in a "filing cabinet."

The Sula Valley and the Tegucigalpa Region has about 2 million people. Central America 180 million of which 77.2 % of the population is urban.

We are applying to Solve's Global Challenges program because we believe it offers the comprehensive support, resources, and network needed to truly make a difference to AI Climate.We are eager to contribute to and learn from this vibrant community, and to use this opportunity to create significant, sustainable change.

We encounter several challenges, one of which is market-related. As a non-profit working with other non-profits and municipalities, our solution is built to create significant societal impact, not profits. However, we envision a cross-financing model where profit-oriented entities like micro-insurance companies or banks could support the non-profit aspects of our AI Climate work. We believe Solve's robust network can help us form these strategic partnerships, increasing our solution's reach and visibility.

Culturally, our solution necessitates a paradigm shift in how municipalities finance city growth and resilience. We are confident that Solve's network and conference exposure can help facilitate this cultural shift by raising awareness about our solution and the problems it addresses.

On the financial front, we acknowledge that Solve isn't merely a fundraising platform. Nevertheless, the financial assistance it offers in the form of grants and investments is pivotal to our solution's growth and impact.

Finally, navigating legal complexities often slows our progress. Access to legal services from Solve's supporters would be invaluable in overcoming these legal hurdles, allowing us to focus on our solution's core mission.

Through Solve's comprehensive support, we anticipate overcoming these barriers, thereby improving our solution and maximizing its impact.

1-The technology used is Supervised, Machine learning, CNN, UNet applied to Informal Settlements, Flooding, Ladles and Land Value differentials and a series of datasets including datasets derived from geospatial information and ground truth provided by the local partners.

2- The integration of information (IS, Flooding, LandsLides, Land Value differentials) in a public domain, on a cesium platform viewer, for Municipalities, Informal Settlements and other stakeholders to use to their benefit.

Population reached and transformational actions if our primary goal. 

When possible, we will track  our impacts, specifically for SDG 11.5 by measuring reduction of the number people affected by water-related disasters in IS; and for SDG 11.b we will measure the number of cities adopting the AI Climate platform to promote integrated policies and plans towards inclusion, resource efficacy and climate resiliency.

Qualitative metrics to evaluate progress of AIClimate for next five years:

Stage 1: Consolidating Honduras: extent to which AI Climate is ingrained in their planning system. 

Stage 2: Scaling in Central America extent to how much the algorithm and process can be transferred and how much we will have to correct. 

Stage 3: Scaling in Sierra Leone and Tanzania,  then West and East Africa.  We will measure by population reached. We expect that technically it will require adjustment in algorithms given the change of IS patterns and colors in satellite images. 

We anticipate satisfactory progress in all 3 stages.

See previous question please-

Additionally, because Ai Climate is fast and cheap to update, one of AIClimate uses is that of measuring spatial strategic interventions including assessing urban patterns, informal settlements formation and climate resiliency strategies.

Long-term Goal: AI Climate seeks to create a more sustainable and resilient world by aiding municipalities, non-profits and vulnerable communities  in making data-driven decisions on climate change and growth through the use of AI and advanced analytics.

Preconditions or Steps Needed:

1. Development of AI Tools: Building robust AI tools and models that accurately predict climate patterns, risks, and provide actionable insights.

2. Partnerships: Forming strategic partnerships with municipalities, local experts and non-profits;  additionally profit-oriented entities like micro-insurance companies or banks for cross-financing.

3. Training and Education: Offering necessary training for non-profits and municipal technicians to understand and apply AI tools for climate change mitigation, spatial inclusion and growth; and eventually  serving regionally to Central America.

4. Awareness and Cultural Shift: Generating awareness about the importance of data-driven decision-making in climate change mitigation and creating a cultural shift in municipalities' approach towards growth and resilience financing.

5. Policy Influence: Using data and insights to influence policy for climate change mitigation at the local, regional, and national levels.

6. Sustainability: Ensuring the AI Climate model's sustainability and scalability to expand its reach and make a broader impact.

The Theory of Change for AI Climate centers around the belief that with the right tools, partnerships, and cultural shift, municipalities and non-profits can significantly mitigate climate change and promote spatial inclusion and sustainable growth through data-driven decisions.

Technically speaking, AI Climate is a:

• remote sensing, open source  platform 
• processes geospatial images 
• georeferenced datasets 
• derives analytics-ready layers 
• of informal settlements and impacts associated with climate change.

We developed four models: M1-InformalSettlements, M2-Flooding, M3-Landslides, and M4-LandValueDifferential.   All models were developed with two image resolutions: public satellite images Sentinel 2 (12m) and basemaps (5m) from our partner Planet.

For Informal Settlements, we detected 3 categories in Tegucigalpa: new, in development and consolidated. We detected one in the Sula Valley.

Our methodology entailed, preprocessing satellite images, Ground Truth prepared by partners, and other datasets; we trained them and generated predictions using CNN with Unet architecture. We then refined and formatted results during the post-processing stage. Then comes the validation by local experts and communities whose inputs are used to retrain the algorithms.

We used sentinel 2 (10 m resolution)for Global and Planet 5m for Local  

In Tegucigalpa we prepared an extra band with information in GIS mapping vulnerability (to complement the ground truth).

 a) Preprocessing: We generate the dataset that uses a model with images and masks that limit the region of interest.

b) Training: The model learns the patterns that define the object of interest.

c) Prediction: the model already trained process new images and identifies the patterns defined during training.  For each image it returns another image where each pixel represents a probability.

d) Post processing: one applies a threshold to the probability- Filter and formats results 

-For Flooding susceptibilities, we detected one flooding category in Tegucigalpa and two categories extreme and moderate in the Sula Valley.

We used datasets-images from a variety of sources  including GIS information related to the IOTA/ ETA 2020 and Precipitation event OCT 2008 event, from Unosat as Ground Truth.  We also used derived factors from DEM Alos-palsar:

For example curvature affects the acceleration and deceleration of flow and topographic wetness index refers to moisture in the soil

-For Landslides susceptibility, we detected one category in Tegucigalpa. We used eight datasets  including lithological and soil maps. 

-For land values differentials. We used 8 datasets, incorporating previous models, and correlated them with distance to points of interest, existing corridor bands, and VIIRS nighttime lights as a density measure.

R: distance to centers of interest

G: L2 prediction results

B: L3 prediction results

R: existing corridor band

G: distance to centers of interest

B: L1 prediction results

R: VIIRS nighttime lights

G: L1 prediction results

B: LULC

Overall. we generated 65 layers, mapping 340,000 km² in Tegucigalpa and the Sula Valley.

Our team is composed of practitioners, academics and public officials of diverse ages, genders, countries (six), cultures, and disciplines including data scientists, urban and regional planners, economists, and AI engineers.

Our structured development process integrates advanced technology with local expertise and community validation knowledge. This is how we avoid falling into the trap of magnifying the information and knowledge asymmetries and gaps between those developing, owning and controlling AI and ML technologies, and those who lack access to them.

The I2UD AI Climate platform incorporates local partner and informal settlement communities’ knowledge of their own conditions to ensure that community organizations can make sense of the platform’s findings and use those outcomes for their own benefit. The combination of Planet’s higher resolution images, AI technology, and ground truth and validation processes provided by local partners and informal settlements creates a powerful tool for inclusion and effective city resiliency co-production.

We need help with our business model - We are a non profit and we are in the process of traditional fundraising to continue with Ai Climate.  However, we know there's a  revenue stream, such as insurance companies and banks, to be able to cross finance and scale our impact.