How it works

Inside the Algorithm

From observation to probability.

Inside the algorithm

How NatureCasting thinks

Different parts of nature operate on different clocks. Fast biological behaviour can contain information over hours and days. Phenology, soils and migration integrate conditions over weeks. Oceans and seasonal biological development move more slowly and become increasingly important in the longest outlook. NatureCast therefore changes the weight of each observation according to the forecast horizon.

7NatureCast-7

Behavioural signals dominate: birds, insects, plant closure, soil-surface activity and coastal changes. Designed to detect changes already propagating through the environment.

30NatureCast-30

Timing and accumulated conditions matter more: soil temperature, soil moisture, vegetation, migration, insect lifecycles and marine signals.

90NatureCast-90

Long-range output is probabilistic. Marine conditions, phenology, soil heat and seasonal anomalies carry more weight than immediate animal behaviour.

The natural science

Five independent families of evidence.

Click each section to see why the observations are potentially useful. Agreement across unrelated biological systems is more valuable than one spectacular observation.

BBirdsSwallow feeding height · feeding intensity · seabird movement · migration timing+

Birds respond quickly to pressure, wind, temperature and especially the movement of the insects they feed on. Low-flying swallows, changes in feeding intensity, inland seabird movement and unusually early or late migration can therefore act as biological indicators. Short-lived behaviour is most useful for the 7-day model; migration timing becomes more useful at 30–90 days.

IInsects & invertebratesBee activity · insect height · spider webs · emergence timing+

Insects are highly sensitive to humidity, temperature, wind and soil moisture. Bee foraging, flying-insect height, spider web construction, ant activity and butterfly emergence can move before people notice a broader change. NatureCast treats behavioural signals as short-range evidence and lifecycle timing as a seasonal anomaly.

PPlantsPine cones · flower closure · bud burst · flowering & fruiting+

Plants integrate weather over days, weeks and seasons. Pine cones, flower closure and leaf behaviour can respond to moisture in the short term, while bud burst, flowering, fruiting and leaf-fall dates reveal whether the biological season is running ahead of or behind its local norm.

SSoil & fungiSoil moisture · soil temperature · earthworms · fungal activity+

Soil stores both heat and moisture, so it changes more slowly than the atmosphere. Soil temperature, soil moisture, earthworm activity and fungal emergence therefore provide a bridge between immediate weather and longer seasonal conditions. At one to three months, soil anomalies become much more important.

CCoastal & marineSea state · seabirds · shoreline change · marine-life timing+

The ocean has greater thermal memory than the atmosphere. Coastal bird behaviour, sea state, shoreline changes and marine-life timing can therefore provide useful regional context. For the 90-day outlook, marine and ocean-linked signals receive some of the highest long-range weightings.

From observation to probability.

Each observation is normalised against what is expected for that place and season, then weighted by horizon, historical reliability and local relevance.

Ŷₓ,ₜ₊ₕ⁽ᵏ⁾ = Bₓ,ₜ,ₕ + λᴴ(Hₓ,d,τ − H̄ₓ,d) + Lₓ,c⁽ᵏ⁾ + Clamp[ Σⱼ IⱼQⱼTⱼDⱼHⱼSⱼ⁽ᵏ⁾UⱼAⱼFₕRⱼ ] + M⁽ᵏ⁾(X)The complete NatureCasting forecast equation

01 ObserveVolunteer records structured natural signals.

02 NormaliseCompare against region, date and seasonal norm.

03 WeightApply horizon, reliability and regional relevance.

04 EnsembleResample thousands of combinations to create probabilities.