Using AI (decision-trees) to assess vulnerability to climate change

Innovation methods

Data sources

Scaling

Cost

SDG(s)

Sustainable Development Goal(s)

10Reduced innequalities

11Sustainable cities and communities

13Climate action

Disclaimer:

Please be aware that the content herein has not been peer reviewed. It consists of personal reflections, insights, and learnings of the contributor(s). It may not be exhaustive, nor does it aim to be authoritative knowledge.

Overview

Prepared by (Name of the experimenter)

Javier Brolo and María Inés Castañeda

On date (Day/Month/Year)

May 28th, 2021

Current status of experimental activity

Completed

What portfolio does this activity correspond to? If any

Waste management

What is the frontier challenge does this activity responds to?

How to improve the collaboration between society and public institutions to increase resilience to climate change

What is the learning question(from your action learning plan) is this activity related to?

How to improve waste management?

Please categorize the type that best identifies this experimental activity:

Quasi Experimental (Analytical, observations, etc)

Which sector are you partnering with for this activity? Please select all that apply

United Nations agency, Public Sector

Please list the names of partners mentioned in the previous question:

Energy and Environmental programmatic area, Ministry of Environment and Natural Resources

Design

What is the specific learning intent of the activity?

We are interested in improving waste management. One of the arguments is that by separating organic and inorganic waste at home, we will increase resilience to climate change. However, this connection is not straightforward. Even though data is always insufficient, we know there are open sources of data available from administrative records. We want to learn if it's possible to leverage on the use of Artificial Intelligence techniques using this publicly available data to gain insights regarding the connection between waste management and vulnerability to climate change.

What is your hypothesis? IF... THEN....

If we use open data sources, then we can leverage the use of artificial intelligence to identify locations vulnerable to climate change

Does the activity use a control group for comparison?

Yes, a different group entirely

How is the intervention assigned to different groups in your experiment?

Non-random assignment

Describe which actions will you take to test your hypothesis:

We will process and aggregate open data from the 2018 Census in Guatemala. Then we will apply the Decision Tree algorithm using the open software R. We will consolidate the results in a report and present.

What is the unit of analysis of this experimental activity?

Municipalities

Please describe the data collection technique proposed

Official statistics

What is the timeline of the experimental activity? (Months/Days)

One week

What is the estimated sample size?

100-999

What is the total estimated monetary resources needed for this experiment?

Less than 1,000 USD

Quality Check

This activity is relevant to a CPD outcome, The hypothesis is clearly stated, This activity offers strong collaboration oportunities, This activity offers a high potential for scaling, This activity has a low risk

Please upload any supporting images or visuals for this experiment.

Please upload any supporting links

What are the estimated non- monetary resources required for this experiment? (time allocation from team, external resources, etc) If any.

We need access to internet, a computer, software R, time and data analysis skills. Note, knowing English is useful for troubleshooting.

Results

Was the original hypothesis (If.. then) proven or disproven?

Partially proven. We were able to use open data to implement decision trees, however, the link between waste management and vulnerability to climate change is more complex.

Do you have observations about the methodology chosen for the experiment? What would you change?

We need to make decisions regarding which variables to include in the model, which may have non trivial consequences in the process. It would be useful to contrast this approach with more standard econometric approaches.

From design to results, how long did this activity take? (Time in months)

One week.

What were the actual monetary resources invested in this activity? (Amount in USD)

US$0.00 additional resources

Does this activity have a follow up or a next stage? Please explain

We need to apply it to other topics, and transfer capacity to replicate.

Is this experiment planned to scale? How? With whom?

Within the country office projects, we want the methodology to be adopted. Also, with counterparts who have technical capacities.

Please include any supporting images that could be used to showcase this activity

Please add any supporting links that describe the planning, implementation, results of learning of this activity? For example a tweet, a blog, or a report.

Considering the outcomes of this experimental activity, which of the following best describe what happened after? (Please select all that apply)

This experiment did not scale yet

Learning

What do you know now about the action plan learning question that you did not know before? What were your main learnings during this experiment?

We know that people who are vulnerable to climate change (measured by access to drinking water) can be identified if we know they don't have access to waste collection. This means that if separating waste at home increases resilience to climate change, it does so in a group different than the one who separates waste. This is because people with vulnerabilities generally don't have access to waste collection. We also know that open data from the statistics office can be leveraged using AI methods.

What were the main obstacles and challenges you encountered during this activity?

In this case, there was a data management problem, where we needed to integrate data from different levels of analysis (individual, household, and building). We also needed to learn quickly how to use the tool. These results are often difficult to communicate (an algorithm to classify observations). Also, statistical approaches with a degree of error are highly questions, as practitioners have expectations to not make mistakes when classifying observations. Also, it would be better to have data desegregated beyond the municipal level.

Who at UNDP might benefit from the results of this experimental activity? Why?

All the programs that want to leverage the use of open data.

Who outside UNDP might benefit from the results of this experiment? and why?

In this case the Ministry of environment and natural resources and local governments are interested in using available data.

Did this experiment require iterations? If so, how many and what did you change/adjust along the way? and why?

We needed some filling with the code to process data and implement the procedure.

What advice would you give someone wanting to replicate this experimental activity?

Reach out if you'd like to replicate, so we can share code. Be aware of the variables included in the model. Also, thy to work out how the results will be communicated beforehand, and how the results should be used.

Can this experiment be replicated in another thematic area or other SDGs? If yes, what would need to be considered, if no, why not?

Certainly. We can use decision trees to classify other units of analysis with characteristics relevant to different SDG's.

How much the "sense" and "explore" phases of the learning cycle influenced/shaped this experiment? In hindsight, what would you have done differently with your fellow Solution Mapper and Explorer?

Sense has defined the topic. Exploration has pointed out the possibility of using open data and AI.

What surprised you?

It is surprising that access to waste collection services was such a strong heuristic to identify vulnerability to climate change. Over other aspects such as economic activity, how rural a place is, among others.