All outcomes
Skills

Climate and Environmental Open Data Analysis

8 weeks · 0 milestones

Analyse a real named publicly available climate or environmental dataset: NOAA Global Surface Temperature, NASA GISS Surface Temperature Analysis, Copernicus Climate Change Service ERA5 reanalysis, EPA Air Quality System historical data, or equivalent. The proof is a documented analysis report: data source and version, preprocessing decisions with rationale, statistical analysis (trend detection, anomaly identification, spatial or temporal comparison), uncertainty quantification, and written conclusions with specific physical interpretation. All sources listed are free and publicly accessible. Real climate datasets contain real measurement uncertainty, coverage gaps, and confounds that require genuine scientific reasoning to navigate — rigorous analysis produces real artifacts of scientific value. Reviewed by an environmental scientist who challenges the statistical methodology — specifically asking whether the identified trend is robust to a specific alternative specification — and asks you to interpret a specific dataset feature you did not discuss in your report.

Milestone map

Milestone map

3 milestones

Define research question and identify open dataset

1–2 weeks

Formulate a specific, answerable climate research question and identify the open dataset(s) that contain the relevant data. Open climate datasets are the natural route for this outcome — no physical data collection is required, and the field's most important datasets (ERA5, NOAA Global Surface, NASA GISS, PRISM) are freely available. The quality of the research question determines the quality of the analysis: 'Has [specific variable] changed in [specific region] over [specific time period]?' is answerable; 'Is climate change real?' is not.

Proof required

Submit your research question (one specific sentence), the dataset(s) you have selected with their source URL and access confirmation, a brief description of the dataset's spatial and temporal coverage and resolution, and the dataset's known limitations relevant to your question.

What gets checked

  • Research question is specific enough to be answered with a yes, no, or quantified trend from the identified dataset — not a broad descriptive question
  • Dataset limitations are described in terms relevant to the analysis question — not a generic 'data may have errors' but the specific spatial resolution, temporal gaps, or measurement biases that affect this question
  • Dataset access is confirmed — not just the URL cited but evidence of having downloaded or accessed the data (file size, number of records, date range confirmed)

Common mistakes

  • Choosing a research question that cannot be answered by any freely available dataset — some highly specific questions require data that is not openly available
  • Selecting a dataset without checking whether it covers the time period and spatial extent required by the question

Resources

Foundationstart here

Depthgo deeper

Masteryfor the dedicated

What a verifier looks for

  • Ask the submitter to describe the specific limitation of their chosen dataset most relevant to their question — can they name the limitation and explain its effect on their analysis?
  • Ask why this dataset rather than an alternative for the same question — confirming they evaluated options demonstrates genuine methodological thinking.
  • Ask what temporal or spatial resolution the dataset provides and whether that is adequate for their question — a dataset with annual means cannot answer a question about seasonal variability.

Execute analysis and produce documented visualisations

2–4 weeks depending on dataset size and analysis complexity

Perform the statistical analysis and generate visualisations. All analysis must be reproducible — code and data access instructions are part of the proof. Statistical choices must be justified: if you apply a trend test, name the specific test and why it is appropriate for climate time series (autocorrelation is a common issue in climate data that affects standard significance tests).

Proof required

Submit your analysis code (R or Python, documented with comments explaining key decisions), the output visualisations (at minimum a time series or spatial plot and a statistical test result), and a written account of the statistical method used and why it is appropriate for your data.

What gets checked

  • Code is submitted and runs to produce the submitted figures — not just described
  • Statistical test choice is justified — not just 'I ran a regression' but why that particular test is appropriate given the data structure (temporal autocorrelation, non-stationarity, sample size)
  • Visualisations include axis labels, units, source attribution, and appropriate uncertainty display (confidence intervals, standard error, or significance level)

Common mistakes

  • Producing visualisations without a statistical test — visual trend identification is not a scientific result; it requires quantification and a significance assessment
  • Ignoring autocorrelation in climate time series — standard linear regression assumes independent observations; climate data is serially correlated, and not accounting for this inflates significance

Resources

Foundationstart here

Depthgo deeper

What a verifier looks for

  • Ask the submitter to explain their choice of statistical test — can they name why the test is appropriate for their specific data structure?
  • Ask about temporal autocorrelation — do they know what it is and whether they addressed it in their analysis?
  • Ask them to reproduce one of their figures in front of you from the raw code — confirms the code actually runs and they understand it.

Write report and present findings with Q&A

1–2 weeks to write and schedule review

Complete a research report communicating the findings, their limitations, and their connection to the broader climate science context. Present to a climate scientist, environmental scientist, or quantitative researcher for a Q&A that challenges methodology, interpretation, and the scope of conclusions drawn.

Proof required

Submit your complete analysis report (2500–4000 words: introduction with research question, data and methods, results with figures, discussion of findings and limitations, and conclusion) plus a Q&A record showing specific methodological or interpretive challenges from the reviewer and your responses. The reviewer must be named and their quantitative or climate science background stated.

What gets checked

  • Conclusion is proportionate to the evidence — does not extrapolate beyond the spatial or temporal scope of the dataset analysed
  • Limitations section names specific data quality or methodological limitations — not generic 'more data needed' but the specific ways this analysis could be wrong or incomplete
  • Q&A record shows at least two methodological challenges and substantive responses

Common mistakes

  • A conclusion that claims global or long-term trends from a local or short-period dataset — a 30-station 20-year analysis does not support claims about global climate change
  • A reviewer without quantitative analysis experience who cannot challenge the statistical choices — the Q&A must probe analytical decisions, not just content familiarity

Resources

Depthgo deeper

What a verifier looks for

  • Ask the submitter to bound their conclusion — what spatial or temporal scope can this analysis actually support, and what would require additional data?
  • Ask whether the pattern they found could be explained by a factor other than the one they emphasised in their conclusion — tests whether they have considered confounds.
  • Verify the reviewer has experience with quantitative data analysis — a qualitative climate policy researcher may not be able to challenge statistical method choices.

We use analytics to improve Powstik. No ads, ever.