Combining Taxonomy, Satellite Data and AI to Speed Red Listing: A Classroom Hackathon

What if students could help conservation scientists decide which species most urgently need a Red List assessment? That is the idea behind this classroom biodiversity hackathon: teams combine open taxonomy records, satellite-derived habitat metrics, and simple machine learning to prioritize species for review. The goal is not to replace expert judgment. Instead, it is to create a practical, teachable workflow that mirrors how modern conservation prioritization is increasingly done at scale. For educators looking to connect ecology, data science, and environmental stewardship, this is a rare chance to build a project that feels real, relevant, and rigorous. If you want a broader framework for building classroom data projects, see our guide on data-driven content roadmaps and the lesson-focused approach in off-the-shelf market research.

This article is a definitive guide for planning, teaching, and running a biodiversity hackathon centered on Red List triage. It explains the scientific logic, the data sources, the classroom setup, the basic modeling steps, the ethics, and the judging criteria. It also gives you an adaptable format for middle school, high school, undergraduate, or teacher-training settings. Think of it as a turnkey blueprint: part ecology lesson, part data lab, part civic science project. The approach borrows from the same principles that power modern workflows in other fields, including scaling from pilot to operating model and workflow optimization through AI triage.

Why Red Listing Needs Faster Triage

The bottleneck is not just discovery

Many species remain unassessed long after they are known to science. Taxonomists, collection managers, and conservation assessors often work with incomplete records, uneven geographic coverage, and limited time. The IUCN Red List is a powerful system, but the pipeline from species discovery to formal assessment can be slow, especially for poorly studied groups such as insects, reef fish, fungi, and amphibians. A classroom hackathon can help students understand that conservation is not only about fieldwork; it is also about information flow, prioritization, and decision-making under uncertainty.

This matters because conservation resources are always limited. Assessors need to know which species are most likely to be threatened, which have the least data, and which habitats are disappearing fastest. That is where open biodiversity data and remote sensing become useful. Students can see that the same kind of prioritization logic used in breaking-news workflows can be adapted to conservation: when the situation is dynamic and the volume of information is large, you need a triage system before you can do deep analysis. For classroom planning that emphasizes usefulness and repeatability, the routine-building ideas in repeatable live routines are surprisingly relevant.

Taxonomy is the foundation

Taxonomy tells us what the species is, where it fits in the tree of life, and how records should be interpreted. Without a reliable taxonomic backbone, any AI model will be messy at best and misleading at worst. Students should learn that names change, synonyms exist, and some records are flagged as uncertain. This is not a weakness of science; it is science working as a correction system. In a hackathon, one of the most important tasks is cleaning and standardizing species names before analysis begins.

The classroom lesson is powerful: conservation decisions start with identity. If two databases disagree on a species name, students need to resolve the mismatch, document assumptions, and explain uncertainty. That exercise builds scientific literacy and data hygiene at the same time. It also mirrors responsible practices discussed in topics like creative control in the age of AI and the viral news checkpoint, where verification matters before action. In ecology, just as in media, trust starts with careful checking.

Satellite habitat data adds the missing spatial context

Open species occurrence points tell you where a species has been observed, but they do not show how its habitat is changing over time. Satellite-derived habitat metrics fill that gap. Students can use indicators such as forest loss, vegetation cover, land-use change, night-time lights, drought stress, fragmentation, distance to protected areas, or coastal modification depending on the species group. These variables make it possible to estimate whether known sites are becoming more degraded or more isolated, which is often a strong signal of extinction risk.

In other words, satellite data helps move the class from a static checklist to a living map. That is one reason this hackathon is such a good fit for environmental science education: students can literally watch habitat pressure accumulate across landscapes. They also learn that not all good data comes from the ground. The same analytical mindset shows up in other data-driven fields, like predictive analytics architectures and lab-to-market innovation pipelines. The lesson is universal: better context leads to better decisions.

What Students Build in the Hackathon

A species prioritization score

The core deliverable is a simple ranking model that estimates which species should be assessed first for Red Listing. Students can create a score based on three broad categories: data availability, habitat pressure, and taxonomic confidence. Data availability can include number of occurrence records, number of unique localities, and recency of observations. Habitat pressure can be measured using satellite indicators such as tree-cover loss, land conversion, or proximity to roads. Taxonomic confidence can be a manual or semi-automated flag that reflects whether the species name is stable or controversial.

This ranking does not produce a formal Red List category. Instead, it helps answer a practical question: which species are likely to be both important and urgent? That is a valuable outcome in itself because it lets limited expert review time go farther. Students can compare their score against known threatened species, then test whether the model tends to capture species that conservationists already worry about. For a parallel example of how scoring helps focus limited resources, see scoring decisions under constraints and test-based prioritization.

A map and a shortlist

To make the project visually compelling, each team should create two outputs: a map showing occurrence points and habitat change, and a shortlist of candidate species for assessment. The map can be built in a spreadsheet, a GIS platform, or a notebook environment depending on skill level. The shortlist should include a short rationale for each species, such as “high habitat loss,” “few recent records,” or “restricted range with rapid forest conversion.” Students should be encouraged to present their findings as if they were briefing a conservation NGO or museum collections team.

This is where the hackathon becomes an education event rather than a coding exercise. Students are not just making charts; they are practicing decision support. If you want to compare how different evidence types influence decisions, the structure in reading market signals is a good analogy: a single data point rarely decides the outcome, but multiple weak signals can become persuasive when combined thoughtfully. That is exactly what the Red List workflow looks like in real life.

A reproducible workflow notebook

Every team should document their process in a shared notebook or slide deck so others can reproduce it. This should include data sources, cleaning steps, model features, assumptions, and limitations. Reproducibility is not just a technical detail; it is the backbone of trust. Even a simple classroom model should be understandable enough that another team could rerun it and get the same result.

Teachers can frame this as scientific accountability. Students should ask: where did this record come from, how was it filtered, and why was this variable included? That mindset is aligned with responsible digital practice in other domains, such as responsible engagement design and human-vs-AI evaluation practices. In conservation, transparency is not optional because the consequences are ecological, not merely technical.

Data Sources Students Can Actually Use

Open biodiversity records

Start with accessible occurrence data from global biodiversity platforms and museum records. These records give students specimen-based or observation-based evidence of species locations. Depending on the platform and species group, records may include date, collector, coordinates, and uncertainty radius. Teachers should emphasize that occurrence data can be biased toward accessible places, charismatic organisms, and regions with strong sampling networks.

A strong hackathon uses this bias as a teaching point. Students can discover that a species may appear common simply because it has been sampled more intensively, while others look rare because they are under-recorded. That distinction is central to conservation prioritization. For background on how public data and classification systems are changing scientific work, the source article on marine conservation, taxonomy, and Red Listing provides useful context: technological advances and open biodiversity platforms are accelerating collaboration across the species discovery pipeline.

Satellite-derived habitat metrics

Students do not need advanced remote sensing training to use satellite habitat indicators. Many classroom-friendly workflows rely on preprocessed datasets or simple layers summarizing forest cover change, vegetation health, urban expansion, or fire history. The teacher’s job is to choose one or two metrics that are understandable and relevant to the organisms being studied. For example, forest-dwelling birds may be sensitive to tree-cover loss, while coastal species may be more affected by shoreline change, turbidity, or nearby development.

The key is to connect the metric to ecology in plain language. Students should explain why a habitat trend matters to the species’ life history, not just plot a pretty map. That explanation step is where real understanding emerges. If you are building a broader data-rich classroom sequence, the practical framing in data-driven planning and pilot-to-scale thinking can help teachers organize complexity into manageable stages.

Species traits and taxonomic metadata

Where possible, add species traits that may improve prioritization, such as body size, endemism, habitat specificity, elevation range, or reproductive rate. Even a few trait columns can make students think more like ecologists. Taxonomic metadata also helps prevent confusion when names differ across sources. A strong classroom routine is to keep a “taxonomy notes” column with accepted name, synonym, source, and confidence level.

That exercise creates a direct bridge between biodiversity science and information management. It also reflects the same discipline used in project planning and source validation in other fields, such as source checking and measured reporting. If students can manage taxonomic uncertainty, they are learning a transferable research skill.

How the Simple Machine Learning Model Works

Start with a transparent baseline

A good classroom hackathon should begin with an interpretable model, not a black box. Logistic regression, decision trees, or a basic random forest are ideal because students can understand feature importance and thresholds. The first version of the model might predict whether a species is likely to need urgent review based on known threatened species versus lower-priority species. From there, students can apply the model to unassessed species and rank them by predicted urgency.

This approach is pedagogically valuable because it teaches that AI is a tool for pattern detection, not magic. Students should compare model predictions to expert-labeled examples and ask why certain species score high. If a species has few records, shrinking habitat, and narrow range, the model may mark it as high priority. If another species has abundant records, stable habitat, and broad range, it may be lower priority. That logic can be summarized in one sentence: risk is a combination of scarcity, pressure, and biological vulnerability.

Feature engineering matters more than fancy algorithms

For beginners, the value of the project lies in choosing meaningful variables, not in chasing the newest model. Students should test how many records a species has, how recent those records are, whether habitat loss has increased near known sites, and whether the species occupies a small range. Those inputs often explain more than an advanced model with poorly chosen features. Teachers can make this point explicit by comparing a “simple and sensible” model with a “more complex but less interpretable” one.

That lesson echoes the tradeoffs discussed in other practical decision guides, like performance versus practicality and build vs buy decisions. In conservation, as in consumer choice, the best solution is not always the most sophisticated one; it is the one that fits the task and can be defended clearly. A model that students can explain is more educationally useful than one they cannot.

Use AI as a ranking assistant, not a replacement for experts

Students should understand the proper role of AI in this workflow. It is there to narrow the field, not to declare final threat status. The Red List process depends on expert assessment, published evidence, and standardized criteria. The classroom model can suggest which species should be reviewed first, which is incredibly useful when assessors face large backlogs. But it should always be presented as a triage aid, not a conservation verdict.

That boundary is crucial for trust. It matches the careful framing seen in discussions of AI risk review frameworks and enterprise AI rollout: if you do not define the human decision point, the system can overpromise and underdeliver. For educators, this is a teachable moment about the limits of automation.

Hackathon Design: A Step-by-Step Classroom Blueprint

Before the event: prepare the data

Teachers should preselect a study region, taxonomic group, and 20 to 100 species depending on class level. Good beginner topics include amphibians, orchids, reef fish, birds, or butterflies because there is often enough open data to support analysis. Prepare a cleaned occurrence table, a habitat metric layer or summary, and a simple reference sheet for taxonomy issues. If students are new to data work, give them a starter notebook or spreadsheet template so the event focuses on reasoning rather than setup friction.

You can also assign roles in advance: taxonomy lead, data cleaning lead, mapping lead, model lead, and presenter. Role assignment helps every student contribute, even if they are less comfortable coding. This is similar to how organizers in fast-moving environments structure teams for resilience, a concept also useful in skills-based team design and mentorship pipeline building. A good hackathon is inclusive by design.

During the event: sequence the work into four sprints

Run the hackathon in four sprints: question, clean, model, and communicate. In the first sprint, students define what “priority” means and why their species group matters. In the second, they harmonize species names, remove duplicate records, and inspect missing values. In the third, they create the prioritization score or train the baseline model. In the fourth, they convert results into a concise presentation, poster, or class briefing.

This structure keeps the event from becoming chaotic. It also mirrors how professional teams operate when dealing with volatile information streams, much like the planning ideas in volatile-beat reporting and repeatable audience routines. For teachers, the lesson is simple: a sequence reduces overwhelm and creates momentum.

After the event: compare human and model judgment

End by asking teams to explain where the model was convincing and where it was weak. Did it over-prioritize well-studied species because they had more data? Did it miss species with tiny ranges but limited records? Did it confuse taxonomic uncertainty with conservation urgency? Those questions are the heart of data literacy. Students should leave understanding that a model is a hypothesis generator, not a final answer.

To make the debrief feel real, invite students to compare their ranking with published threatened species lists or with a teacher-curated “expert expectation” list. This allows meaningful reflection on false positives and false negatives. It also reinforces the conservation principle that uncertainty is not failure; it is a reason to ask better questions. In that sense, the hackathon teaches scientific humility as well as technical skill.

Comparison Table: Data Options, Difficulty, and Classroom Use

The table below helps teachers choose the right combination of data sources and methods for their students. The best option depends on age, time, and technical comfort. Start simple if the class is new to GIS or coding, then increase complexity in later versions of the event.

Evaluation, Ethics, and Trustworthiness

Do not confuse accessibility with simplicity

An accessible hackathon should still respect scientific rigor. Students should be taught to document data sources, note missingness, and explain why some species could not be assessed confidently. The most important ethical lesson is that incomplete data does not mean unimportant species. In fact, some species are both poorly known and highly threatened, which is exactly why prioritization is needed.

Teachers can reinforce this by asking students to create a “limitations” slide. That slide should name sampling bias, taxonomic uncertainty, spatial uncertainty, and the possibility that satellite data may miss local ecological variation. This habit prevents overclaiming and builds trust. It also mirrors the caution required in topics like information verification and AI output evaluation.

Think about bias in the data pipeline

Biodiversity data are not neutral. They reflect who sampled, where they sampled, what they could identify, and which species attracted attention. Satellite layers also carry bias because they compress complex landscapes into generalized pixels. A good teacher explicitly names these biases so students do not mistake map precision for ecological truth. The point of the hackathon is not perfection; it is informed prioritization.

This is where the event can become an excellent lesson in systems thinking. Just as rebuilding local information systems requires understanding audience gaps, conservation prioritization requires understanding data gaps. Students begin to see that missingness itself is a signal. In conservation, absence of evidence is not evidence of absence.

Ask what would change the ranking

A trustworthy model should be sensitive to new evidence. Students should identify which additional data would most improve the ranking: more occurrence records, higher-resolution habitat imagery, trait data, or expert annotations. That reflection teaches that models are provisional tools. If a species suddenly appears in a protected area, or if forest loss accelerates near its range, the ranking should change.

This is the essence of adaptive science. It resembles the iterative logic behind scaling from pilots and integrating decision support into workflows. For students, that mindset is empowering: they learn that science is a living process, not a static textbook answer.

How to Adapt the Hackathon for Different Age Groups

Middle school: focus on mapping and storytelling

For younger learners, reduce the coding and emphasize maps, species profiles, and conservation stories. Students can sort records, compare habitat images, and manually rank species using a rubric. The final product might be a poster or oral presentation rather than a machine-learning model. The educational win here is curiosity: students learn that ecology is about places, patterns, and decisions.

High school: add scoring and basic model testing

At the high school level, students can compute simple scores and test whether those scores match known threatened species. They can use spreadsheets or beginner-friendly notebooks to calculate averages, weights, or categories. Teachers can ask them to justify which variables mattered most and why. This level is ideal for connecting science, statistics, and environmental policy.

Undergraduate or teacher-training: add validation and interpretation

More advanced groups can build a training set from assessed species and validate whether their model recovers expected priorities. They can discuss precision, recall, class imbalance, and threshold choice in plain language. This version is ideal for future teachers because it shows how to translate technical ideas into classroom-ready language. It also reinforces that open data literacy is increasingly central to environmental education.

Conclusion: Why This Hackathon Matters

A hackathon that blends taxonomy, satellite data, and AI does more than teach technical skills. It shows students how modern conservation works when time, money, and expertise are limited. By turning open biodiversity records into a prioritization workflow, students learn to think like ecologists, data analysts, and science communicators at the same time. They also gain a realistic appreciation for the role of the Red List in conservation planning.

For educators, this is a high-value classroom event because it is authentic, flexible, and interdisciplinary. It can be run in a single day or stretched into a week-long project. It can be adapted for geography, biology, data science, or environmental studies. Most importantly, it gives students a concrete way to contribute to conservation thinking while learning how to handle uncertainty responsibly. If you want more inspiration for event planning, the frameworks in deadline-driven project design and event experience design can help shape the final presentation and judging format.

Pro Tip: The best student teams do not just rank species. They explain why their ranking is useful, what data changed the score, and how an expert would use the result as a starting point for review rather than a final answer.

Related Reading

• From Pilot to Operating Model: A Leader's Playbook for Scaling AI Across the Enterprise - Useful for designing a classroom workflow that can grow from a simple pilot into a repeatable event.
• When AI Features Go Sideways: A Risk Review Framework for Browser and Device Vendors - A practical reminder to build safeguards into any AI-supported student project.
• The Viral News Checkpoint: 7 Questions to Ask Before You Share Anything - A strong model for teaching source verification and evidence checks.
• Rebuilding Local Reach: Programmatic Strategies to Replace Fading Local News Audiences - Helpful for thinking about data gaps, community needs, and communication strategy.
• Human-Written vs AI-Written Content: What Actually Ranks in 2026 - Relevant for evaluating AI outputs carefully and keeping human judgment in the loop.