Service Behavior & Expectations

This guide describes the expected runtime behavior of FHI after an enrollment is submitted: when imagery becomes available, what behavior is normal versus exceptional, and what FHI guarantees versus what depends on external satellite providers.
It complements the FHI Overview and API Reference, which cover concepts and request/response shapes.

How FHI works at runtime

An enrollment defines what imagery should be generated, for which fields, and over what monitoring window — it does not guarantee instant availability. Actual imagery availability depends on external satellite imagery providers:

1. The enrollment is submitted and validated.
2. FHI tracks every requested field-date combination as an independent unit of work.
3. Imagery is generated only once the imagery provider publishes data for that date.
4. Outputs are written automatically to storage as soon as imagery becomes available.

note

A completed enrollment status means the GeoJSON and monitoring parameters were successfully ingested and scheduled — it does not guarantee imagery is available for every requested field-date. See Common scenarios below.

Updating a field boundary

Field geometries are keyed by field_id. If you submit an enrollment reusing a previously used field_id but with a different boundary, FHI reuses the existing geometry and ignores the new boundary.

caution

To change a field's boundary, submit a new, different field_id. Re-submitting the same field_id will never update the stored geometry.

Expected imagery availability timelines

Availability depends on whether imagery is historical (already captured) or newly captured:

• Historical imagery (already captured by the provider): usually available within about 12 hours of enrollment.
• Newly captured imagery: may take up to about 72 hours after capture to become available via the provider. For example, imagery captured on June 13 will typically be available by June 16.

These delays are expected behavior and outside of FHI's direct control — they reflect the publication timeline of the upstream imagery provider. Once imagery is published, FHI automatically processes and delivers it with no further action needed from you.

Common scenarios and what they mean

"My enrollment is completed, but I don't see imagery yet." This is usually expected. It can mean:

• The imagery has not yet been published by the provider.
• The requested date range does not contain available images for that field.
• Cloud cover or an acquisition gap prevented usable imagery for that date.

"Some dates are missing imagery." This is normal. Satellite acquisition is not guaranteed for every field on every date.

"When should I contact support?" Reach out if:

• Imagery has not appeared after the expected availability window described above.
• You observe inconsistencies across many fields or dates.
• You are preparing for a production launch and want validation ahead of time.

What FHI guarantees vs. what's external

FHI guarantees:

• Deterministic handling of all valid enrollments.
• Automatic processing as soon as imagery becomes available.
• Stable and predictable storage paths.
• No data loss once imagery has been published by the provider.

External dependencies (outside FHI's control):

• Satellite acquisition timing.
• Imagery publication delays.
• Scene availability and quality (e.g., cloud cover).

Image quality and daily selection

Satellite imagery quality varies with acquisition conditions (e.g., cloud cover, viewing geometry). FHI evaluates quality before processing and may intentionally skip images that are not suitable for analysis — a missing date often reflects insufficient quality rather than a processing error.

Attribute	Description	Default threshold
area_covered_percent	Percentage of the field geometry covered by the satellite scene	≥ 70%
clear_percent	Percentage of the field area free of haze, cloud, snow, and shadow	≥ 80%
clear_confidence_percent	Average confidence for pixels classified as clear	≥ 95%
view_angle	Sensor viewing angle relative to nadir (lower values generally preferred)	—

These attributes are used in two ways:

• Filtering — imagery that doesn't meet the minimum thresholds may be skipped entirely, to avoid generating misleading or low-value outputs.
• Selection — when multiple candidate scenes exist for the same field and date, these attributes deterministically select a single representative image, used consistently to generate every requested layer for that field-date.

tip

Only one image per field per day is used to produce all requested outputs, so analytical and visual layers for the same date always come from the same source scene.

Scale and practical limits

FHI is designed to support thousands of fields per day. For high-volume or production launches, we recommend sharing a representative enrollment sample ahead of time so capacity and performance can be validated collaboratively.

Payload limits (see API Reference for the authoritative, up-to-date values):

• Maximum GeoJSON file size: 50 MB.
• Maximum number of fields per enrollment: 250 (adjustable for specific use cases).