Watering Strategy

What We Are Trying to Learn

The aim is to recognize each container's normal dry-down, notice when that pattern changes, and water when the plant and root zone are ready. More data should eventually make care easier. As a pattern becomes reliable, fewer well-timed readings should be needed.

The scale measures the entire weighing assembly: plant, roots, medium, pot, labels, and any accessories included on the scale. A falling number records a change in that assembly's mass. It does not identify exactly where moisture remains, prove that roots are healthy, or measure a watering dose.

For the usual cactus and succulent group, the care starting point is a thorough watering in a container with working drainage, followed by drying between waterings. Adjust that decision for the actual plant and growing conditions. University of Minnesota Extension supports thorough watering, drainage, and reduced watering during low-light rest; it does not establish a universal number of days between waterings. University of Minnesota Extension

Three Different Decisions

The logger's Water date* is conditional; the asterisk matters. Its Dry-check window describes uncertainty around approaching a previous mass reference. It is not the next weighing appointment. Forecast basis explains which observations support the forecast; it is not a separate care action.

The daily report separates Water Today, Dry reference reached — waiting for plateau, plant-specific Check only exceptions, and useful weighing. Its main Water list requires a confirmed sustained plateau and the current validity/manual guards; a reached dry reference alone stays out of that list. The owner performs the usual moisture/readiness check before watering, so an unlogged routine check alone does not exclude an otherwise supported candidate. Known moisture, root, setup, new-watering, and species restrictions still take precedence. A separately labeled AI assessment can explain an exception without changing the main list or claiming an unsupported plateau is confirmed. Follow the current report policy rather than a generic calendar.

What Wet and Dry References Mean

New Weigh observations are stored as Routine. Derived views infer their place in a watering cycle without rewriting the original History rows:

1. An explicitly logged Water starts a watering cycle.
2. A weight saved with Water supplies the Wet reference. If no weight accompanied that save, the first eligible positive reading within five days can supply it.
3. The final eligible non-Wet weight before the following Water becomes a completed Dry endpoint.
4. Low readings in the still-open cycle remain Routine. A new minimum alone does not become a completed dry reference.

The five-day allowance is a recovery rule for missing data, not a recommendation to wait five days to weigh after watering. For a useful wet anchor, weigh after the actual watering has finished draining, using the normal weighing assembly. A delayed reading can miss the earliest mass loss.

The completed Dry endpoint means “the last eligible measured weight before the next watering.” It does not certify a water-free pot or establish that the previous watering decision was ideal. Growth, pruning, spilled medium, changed accessories, or a different watering practice can also make an older reference less representative.

Reading the Current Weight Difference

Current weight difference = latest eligible weight − current setup's completed dry reference

For an illustrative dry reference of 350 g, a current weight of 368 g gives +18 g; 346 g gives −4 g. Positive means above that historical reference, zero means at it, and negative means below it. Neither sign independently establishes moisture or readiness. The difference is unavailable when a valid comparable reference is missing; blank is not zero.

The wet-to-dry difference describes the observed mass range between references. A “remaining” percentage based on that range is a relative position on the recorded curve, not a measured percentage of water in the substrate.

How the Curve Supports a Decision

The model uses eligible observations for the same plant and pot setup. Its forecast can begin with compatible completed cycles and then adapt to the current cycle. The weighing strategy explains the training requirements and forecast labels.

An initially faster decline followed by slower loss can help describe a dry-down. The actual observations, their timing, and the whole cycle matter. There is no universal 1 g/day watering threshold: the same mass change can mean very different things in a small pot and a large shared planter. Reaching zero daily loss is not a care objective.

The current implementation has two signals that can bring a readiness inspection forward:

• Previous weight reference reached: an eligible current weight has reached or crossed the completed dry reference.
• Sustained plateau: the recent tail is small relative to that pot's earlier decline, even if it is still above the old reference.

The current plateau detector uses four readings at least 12 hours apart over 2–10 days, with no seven-day lockout. The earlier measured decline must start at least 24 hours after Water, span at least a day, and exceed 4 g; total observed loss must be at least 10 g. Elapsed-time regression rates compare the tail with that earlier decline: tail loss must be at most 20% of the earlier rate, and its spread at most 5% of total loss, with a minimum 2 g allowance. A net tail gain up to 2 g is tolerated as noise; renewed loss exceeding 2 g and 40% of the earlier rate invalidates the plateau. These are adjustable software heuristics, not species-specific moisture limits. The report uses the same detector as the logger. See the current detector rules.

A supported observed plateau can remain useful without an old dry reference or timely wet anchor. Those missing anchors still prevent a calibrated future forecast. A flat pot with no demonstrated earlier decline remains uncertain.

A plateau can also reflect measurement noise, changed conditions, or poor uptake. The model therefore advances an inspection opportunity rather than issuing an unconditional watering order. It does not rewrite the dry reference to match the plateau.

A Practical Decision Sequence

1. Confirm that the evidence belongs together. Check the plant/container identity, current setup, actual observation times, last watering, and any corrections or unusual changes. Do not combine old-pot readings with a new setup.
2. Locate the current observation on its own curve. Consider the difference from the old dry reference, recent measured losses, any plateau, and whether the current curve agrees with learned history.
3. Apply the plant-specific rule. A dry-looking surface, a reached forecast date, or another plant being watered is not enough. The exceptions below take precedence over a general cactus routine.
4. Choose the supported action. For the main daily report, require a confirmed plateau and valid plant-specific guards for a conditional Water candidate. Keep reference-only hits in the hold/monitor category, or request a useful weight when due. Keep plant-specific Check-only exceptions separate. Actual watering still requires the usual readiness check; do not invent a water date to complete a table.
5. When watering actually occurs, log it accurately. Record the application style, nutrients if used, and measured volume only if known. Record a separate post-drain weight or include the actual post-drain weight with Water.

There is no automatic extra four- or six-day drought after reaching the reference. Equally, a pot that remains damp should not be watered to satisfy a forecast. The existing small-pot dryness guidance explains the limitations of a skewer observation without treating it as a calibrated test.

Collection Exceptions

P35/P36 were repotted September 23 into their chosen Blush Mauve and speckled-brown pots, both on setup 2 with 80% Molly's Succulent Mix plus 20% added perlite. Routine weights were 1247 g and 713.5 g respectively; the repot checks recorded Dry medium and firm plants. These readings are not verified dry baselines. No Water event appears in either repot batch, and prior watering remains unknown. Keep both mesemb pots outside automatic feed/plain-water alternation; repotting creates no feeding event.

Do not copy the source publications' regional seasons or illustrative watering intervals into a fixed calendar for this indoor collection. Exact identities and additional exceptions remain in the plant profiles.

Watering Application and Amount

Record the application that actually occurred. A Partial or Spot event still changes the current watering cycle. Historical blank application values remain legacy/unspecified and are accepted for compatibility.

Water amount (mL) is optional measured liquid volume. Leave it blank if unknown. It is different from the pot's gram increase and from Nutrient amount, which describes the product quantity or concentration with units. The model does not convert “grams above dry” into a required number of milliliters.

Missing or Conflicting Evidence

The September 13 installation of the AW200 and AeroLight 240 W changed growing conditions, not the physical weighing assembly. It does not automatically advance Pot setup. The model does not read lamp settings or recognize a lighting note as a numerical correction; actual subsequent weights must show the response. See the installed setup record.

Sources and Implementation References

• University of Minnesota Extension: cacti and succulents — general watering and drainage guidance.
• SANBI: Pleiospilos nelii — split-rock leaf replacement and care context.
• Wisconsin Extension: Aeonium — observed growth/rest distinctions.
• Practical care notes and the linked plant profiles — collection-specific evidence and exceptions.
• Logger source — dryDownCycles_, fullWateringForForecast_, dryDownModelForPlant_, cycleInspection_, and watering guidance.
• Operator runbook — model behavior and the distinction between a conditional date and a watering instruction.