PACIFIC DATAVIZ CHALLENGE 2026
When Water Changes, Everything Changes
Samoa cannot plan for an average water year. In the official 1979–2025 record, 2020 has the highest rainfall anomaly and 2022 the second-lowest.
My father is from Vaiafai, Iva, in Savai'i, and much of our aiga still lives there. I grew up in Aotearoa with a Māori mum and a Samoan dad. For me, Samoa is family, not an abstract region on a chart. This is a personal reading of public data, not a claim to speak for every Samoan community.
This reversal shows observed climate variability. It does not prove climate change caused these particular years. The records below do not describe Savai'i village by village, but they show why an average year is a poor plan.
THE EVIDENCE BOUNDARY
What this water story can and cannot show
They can
- show water conditions national plans should be tested against;
- make reversals hidden by averages easier to see;
- document one history of adaptation after a major crop shock.
They cannot
- prove that ocean temperature caused a rainfall or crop outcome;
- describe a particular Savai'i village, catchment, farm or household;
- choose a locally appropriate response without local evidence and consent.
ACT 1 · WHEN WATER CHANGES
Water has a rhythm. The records show why it cannot be taken for granted.
Samoa's official annual rainfall indicator moved from above its reference average in 2020 to below it in 2021 and 2022. The EEZ sea-surface-temperature series sits beside it as context, not as an explanation for the reversal.
Samoa-wide official indicator
A wet-to-dry reversal in three annual readings
1991 to 2020 baseline
Read this first: +27.3 mm in 2020 became −18.1 mm in 2021 and −20.7 mm in 2022. These are annual Samoa-wide anomalies, not local drought declarations.
Samoa EEZ aggregate
The warming ocean is the background condition
1971 to 2000 baseline
2024: +0.8°C in the EEZ annual series. This is not a coastal or inshore temperature record and is not evidence that SST caused a rainfall outcome.
ACT 2 · WHEN RAIN ARRIVES ON DIFFERENT TERMS
A seasonal pattern can be familiar without making a particular month predictable.
The near-Upolu grid point has a wet and dry rhythm across the year. Individual months can still sit far from that pattern. Start with an episode, then use the controls to test another year and month.
Coarse MERRA-2 grid-cell estimate near central Upolu
A within-model monthly average gives orientation, not a local forecast
1981 to 2025 calendar-month record mean
This coarse modelled estimate is used to compare months within the same record. It is not used as local rainfall, a rain-gauge reading or a Samoa-wide estimate.
WATER STRESS TEST
Choose a historical water condition worth testing
Each preset explains why the date matters. Use the year and month controls to explore another condition.
SELECTED EPISODE
April 2016: unusually wet month
What happened
Evidence scope
Planning question
What it cannot determine
Official Samoa-wide or EEZ indicators
National and ocean context
Samoa rainfall anomaly
Not reportedAnnual, Samoa-wide, 1991 to 2020 baselineSamoa EEZ SST anomaly
Not reportedAnnual, EEZ aggregate, 1971 to 2000 baselineModelled grid point near central Upolu
April in 2016
Modelled monthly departure
Not reportedRelative to the same calendar month’s 1981–2025 record meanFood-resilience history
Taro yield in 2016
Reported taro yield
Not reportedAnnual national/country series, descriptive onlyCoarse MERRA-2 grid-cell estimate near central Upolu
Every month compared with its own record mean
−211 mm drier · 0 · +358 mm wetter
On mobile, use the year and month controls above. The heat map remains available as an overview.
April 2016 is the wettest departure in this 1981 to 2025 grid-point record (+358 mm); February 1998 is the driest (−211 mm). Those are historical modelled values at one grid point, not local impact reports.
ACT 3 · FOOD NEEDS OPTIONS
When conditions shift, food systems need more than one way to respond.
The 1993 taro-leaf-blight outbreak was a disease crisis, not evidence of a climate effect. FAO documents a response built around a broader genetic base, regional exchange and on-farm farmer evaluation. That makes the history relevant to adaptation, even though this yield series cannot assign causes by itself.
Food-resilience history
The documented outbreak, the 1994 low and a partial recovery
Annual taro yield · kg/ha
Recorded series: 3,409.1 kg/ha in 1993; 1,500 kg/ha in 1994; 5,000 kg/ha in 2000; 5,574.6 kg/ha in 2024. The series was already falling before the documented outbreak, and 2024 remains below earlier peaks. It describes yield, not causes.
A CURRENT POLICY CONNECTION
More options remain part of Samoa’s adaptation agenda.
Samoa’s NDC 3.0 sets a 2035 target to develop and distribute at least three climate-resilient crop varieties adapted to Samoa’s agro-ecological zones. It also states that adoption requires consent from farmers and commercial and community-based farms.
Read Samoa’s NDC 3.0ACT 4 · PLAN FOR THE RANGE, DECIDE IN PLACE
National data can identify the tests. Communities determine the response.
Samoa's NDC 3.0 says Community Integrated Management plans enable each of Samoa's 368 villages to identify locally relevant adaptation measures and draw on local knowledge. This story can frame questions. It cannot make local decisions.
Stress-test national plans
Test wet years, dry years and unusual months, not one average scenario.
Match evidence to the decision
National, EEZ and grid-point evidence can frame questions. It cannot replace local observations or judgement.
Build adaptive options
Crop diversity, farmer-led trials, community planning, local knowledge and consent are resilience infrastructure.
My family connects me to a Samoa larger than any chart. Water supports the ordinary things that hold aiga together: food, travel, gathering and care. The data cannot tell a village which choice belongs in its place. It can show the conditions our plans must be ready to face.
When water changes, everything changes. Planning starts with the range of conditions we may face, then continues with the people who live with them.
APPENDIXMethods, data and limitations
Pacific climate context
Pacific climate contextNiño 3.4 is retained here as context only. It is not a Samoa rainfall or taro-causation label.
Forecast-method back-test
The held-out comparison is methodological evidence, not a current forecast.
Monthly distribution mechanics
Each box contains 45 observations for the same calendar month. Box = middle 50%; whiskers = 1.5×IQR; dot = median.
Geographic scope
The locator shows only the disclosed MERRA-2/POWER sample coordinate. It is not a risk map.
Data treatment and limitations
- Missing values remain missing; they are never converted to zero.
- Rainfall, SST, taro yield and Niño 3.4 retain separate units, baselines and spatial scopes.
- No displayed alignment is treated as causal evidence.
- The seasonal model comparison is retrospective and does not produce an operational outlook.
- Raw source captures and deterministic transformations are retained in the project repository.
Sources
- SPC/Pacific Data Hub: Samoa rainfall anomaly
- SPC/Pacific Data Hub: Samoa EEZ SST anomaly
- SPC/Pacific Data Hub: Samoa taro yield
- NASA POWER/MERRA-2: monthly precipitation
- NOAA PSL: Niño 3.4 monthly SST
- FAO: The taro improvement programme in Samoa
- Government of Samoa: NDC 3.0
- Pacific Dataviz Challenge 2026 rules
Accessed and frozen for this review build in August 2026. Dataset-specific reuse terms remain a final submission check.
Development process
This review build uses AI-assisted coding and drafting. The competition rules permit supportive AI use, while requiring the participant’s own creative and analytical judgement. Final narrative, cultural framing, analysis and submission decisions therefore remain explicit human review gates.