Crop Cycle Models: Cyclical, Perennial, and Rolling Continuous

The canonical vocabulary and comparison logic for how different crops cycle: rice’s Main and Off Seasons, oil palm’s Production Year and tree age cohorts, rubber’s tapping panels, and pineapple’s Planting Batches.

A Season in the platform is one crop cycle bound to a Field. But “one cycle” means different things for different crops. Rice runs two short cycles per year. Oil palm runs one continuous 25-year producing life. Pineapple runs rolling batches that never share a boundary. If we treat all of them the same way, the analytics, rule cards, and rollups produce nonsense.

This page is the canonical reference for how each crop cycles, what a Season means for that crop, and which vocabulary the UI uses. Every other page that touches Season inherits from this one.

Why cycle model matters

Three things depend on getting the cycle right:

1. Comparability. Cycle Analysis can only compare like-for-like. Rice Main Season vs Rice Off Season is apples to oranges (different water regime, different pest pressure, different yield ceiling). Same crop, same Season Type, aligned by phenology week is the honest comparison.
2. Rule cards. The Risk Model phenology clock only makes sense inside a cycle. A rule card with stage_window: booting needs a defined planting date to know when booting is. Some rules are also Season-Type specific; bacterial leaf blight pressure is higher in the Off Season, and a rule card can filter for that.
3. Verification and rollups. One Season is usually one Verification bundle boundary for cyclical crops. For perennials the bundle boundary is a calendar period (month or quarter). Aggregation rollups can slice by Season Type so estate managers can compare Main to Main across years.

The three cycle families

Every supported crop maps to exactly one of these families.

Cyclical

Discrete cycles with a planting date and a harvest date. Multiple cycles can happen per calendar year. Each cycle is one Season.

Examples: rice, corn, groundnut, most vegetables.

Perennial

One long producing life (years to decades). No planting-to-harvest cycle in the annual sense. Instead, the year is the container, and shorter internal cycles (flowering cohorts, tapping panels) live inside it.

Examples: oil palm, rubber, coconut, cocoa.

Rolling continuous

Planting happens continuously through the year. Each batch has its own multi-month cycle. Batches at different ages coexist on the same Field or Estate.

Examples: pineapple, banana, papaya, nursery operations.

The cycle family is set at the crop level and inherited by every Field of that crop. It drives the Season Type vocabulary the UI offers, the default comparison mode in Cycle Analysis, and the Verification cadence.

Cycle vocabulary by crop

The Season Type dropdown in the Fields Workspace changes by crop. Use these tables as the source of truth for what each term means.

Rice (cyclical)

Malaysian rice runs on a two-season calendar dictated by the monsoon and by irrigation authority schedules (MADA in Kedah, KADA in Kelantan). A single Field runs two Seasons per year in most granaries.

Season Type	Local term	Typical window (Peninsular Malaysia)	Notes
Main Season	Musim Utama	Plant Aug-Oct, harvest Jan-Mar	Primary cycle. Higher yields, reliable water, cooler nights.
Off Season	Musim Luar	Plant Feb-Apr, harvest Jun-Aug	Secondary cycle. 15-25% lower yields are normal, higher pest pressure, tighter water.
Dry Sowing	-	Sub-type of either	Seeds broadcast on dry ground before flooding.
Wet Sowing	-	Sub-type of either	Pre-germinated seeds broadcast onto puddled paddy.
Ratoon	-	60-90 d after main harvest	Second harvest from the same crop stubble. Uncommon in Malaysia but supported.

Comparison rule: compare Main to Main and Off to Off across years. Never compare Main to Off directly; the yield delta is structural, not agronomic.

Season label pattern: Main Season 1 · 2026, Off Season 2 · 2026. The number represents the cycle index within the year (1 = first, 2 = second), which matters when an unusual year has three cycles.

Oil Palm (perennial)

Oil palm has no traditional season. A single planting yields for 20-25 years, harvested every 10-14 days on a rotation. Fresh Fruit Bunches (FFB) are the harvested unit; downstream mills convert FFB to CPO (crude palm oil) at 20-23% Oil Extraction Rate (OER).

Season Type	Meaning	Notes
Production Year	Calendar or fiscal year of FFB output	The primary Season boundary. Compare Production Year to Production Year.
Immature Phase	Year 0-3 from planting	No harvest. Growth-only phase.
Young Mature	Year 4-7	15-22 t/ha/yr, yield climbing.
Prime	Year 8-18	25-30 t/ha/yr, peak production.
Declining	Year 19-25	18-22 t/ha/yr, yield falling.
Replant	Year 25+	Fell and replant. Field returns to Immature Phase.
Flowering Cohort	~5-week cycle of bunches from one flowering event	Sub-cycle within Production Year. Used for short-window analytics.

Comparison rule: compare Production Year to Production Year at the same Tree Age Cohort. A 2018-planted block at year 7 compares to a 2019-planted block at year 7, not to itself last year (age differs by one year, which materially changes yield potential).

Season label pattern: Production Year 2026, with a Tree Age Cohort chip like Planted 2018 · Age 8 for context.

Rubber (perennial)

Rubber is tapped daily or every-other-day; a shallow spiral cut in the bark releases latex overnight into a cup. Bark is organized into panels (A, B, C, D); each panel is tapped for ~5 years then rotated so the tree can regenerate.

Season Type	Meaning	Notes
Production Year	Calendar year of dry rubber output	Primary Season boundary. Measured in kg/ha/yr DRC (Dry Rubber Content).
Immature Phase	Year 0-6	No tapping. Trees grow to tappable girth.
Tapping Panel	A, B, C, or D	The bark section currently being tapped. Rotated ~every 5 years.
Tapping Cycle	d/2, d/3, d/4	Tap every 2, 3, or 4 days. d/2 is intensive, d/4 conservative.
Wintering	Feb-Apr in Malaysia	Annual leaf fall. Tapping paused; yield drops to near zero.
Stimulation Round	Ethephon application cycle	Chemical stimulation to raise yield short-term.

Comparison rule: compare Production Year to Production Year on the same Tapping Panel and same Tapping Cycle. Comparing d/2 tapped Panel B to d/3 tapped Panel C is meaningless.

Season label pattern: Production Year 2026 · Panel B · d/3.

Pineapple (rolling continuous)

Pineapple is planted continuously through the year. Each Planting Batch has its own 14-24 month cycle. Multiple batches at different ages coexist on the same Field or Estate. Flowering is induced (forcing) with ethephon or calcium carbide.

Season Type	Meaning	Notes
Planting Batch	A group planted together on one date	The primary Season boundary. Batches are the comparison unit.
Plant Crop	First fruiting, ~14-18 months after planting	The main harvest of a batch.
Ratoon Crop	Second (and sometimes third) fruiting from the same plant	~10-12 months after Plant Crop. Lower yield than plant crop, common in Malaysia.
Forcing	Induced flowering event	Marks the transition from vegetative growth to fruit development. Sets the harvest date.

Comparison rule: compare batches at the same age from planting, not the same calendar date. Batch A planted January vs Batch B planted April both at their 12-month mark tells you if the field is trending; comparing them on the same calendar day compares different growth stages.

Season label pattern: Batch 2026-A · Plant Crop or Batch 2025-B · Ratoon 1.

How cycle model interacts with phenology

Cycle model and phenology work together but describe different things.

	Cycle model	Phenology
Scope	The whole Season boundary	Position inside a Season
Unit	Cyclical: cycle. Perennial: year. Rolling: batch.	Stage (emergence, tillering, booting, fruit development, ripening)
Duration	Months to decades	Days to weeks
Who declares it	User at planting; inherited from crop	Computed by phenology clock from planting date + GDD
Used for	Grouping, comparison, verification cadence	Rule card stage_window, yield impact layering, mitigation window

The X-axis fix. When Cycle Analysis compares two Seasons, it aligns them by phenology week (days since planting expressed in weeks), not calendar week. Week 8 of Main 2024 is booting, and week 8 of Main 2025 is booting, always. Comparing calendar week 24 across years compares different biological stages and is misleading.

How cycle model interacts with rule cards

All Risk Model rule cards run inside a Season. Three cycle-related fields matter:

rule_card:
  id: rice_blb
  stage_window: [tillering, booting]        # Phenology inside the cycle.
  season_type_filter:                        # Optional. Cyclical crops only.
    include: ["Off Season"]                  # Only fire in Off Season.
  cycle_model_filter:                        # Optional. Rare.
    include: ["cyclical"]                    # Skip on perennials entirely.

When to set season_type_filter: the hazard behaves differently across Season Types. Bacterial leaf blight is a good example; humid warm nights of the Off Season make it substantially more likely than in Main Season, so a rule card can gate on that.

When to skip both filters: the hazard is season-independent (most drought, most nutrient stress). Leave them unset; the rule fires wherever phenology matches.

How cycle model interacts with Verification and rollups

Verification bundle boundary:

Cycle family	Bundle boundary	Reason
Cyclical	One Season = one bundle	Subsidy and takaful claims are per-cycle.
Perennial	One month or one quarter = one bundle	Continuous harvest; no cycle boundary. Estate closes books monthly.
Rolling continuous	One Batch = one bundle	Each batch is its own claim unit at harvest.

See Verification for the full evidence bundle spec.

Aggregation rollups:

• Cyclical crops roll up per Season Type per year (Main 2026, Off 2026 as separate slices).
• Perennials roll up per Production Year, with an optional Tree Age Cohort filter.
• Rolling continuous rolls up per Batch cohort (e.g., all Q1-2026 batches).

See Aggregation Model for how confidence and k-anonymity apply to these slices.

Related

• Field Data Model - the Season object and the four cycle-related fields.
• Cycle Analysis - the three comparison modes that use this vocabulary.
• Risk Model - phenology clock and season_type_filter.
• Verification - bundle boundaries per cycle family.
• Aggregation Model - how Season Type slices roll up.
• Fields Workspace - where the Season pill exposes this vocabulary to operators.