AI-Powered Irrigation Intelligence

Telemetry Insights devices don't use timers or fixed schedules. Every irrigation decision combines real soil moisture data, localized daily weather forecasts, crop science, and — after 90 days — a machine learning model trained on your device's own history. This guide explains how every layer works.

How Telemetry Insights AI Makes Every Irrigation Decision

Four intelligence layers — working together, every night, automatically

Data Sources

🌱

Soil Sensor

Real-time root-zone
moisture, temperature,
humidity & pressure

LiveRoot Zone

🌤️

Daily Weather

Localized forecast via
3rd Party Weather API
rain probability & ET₀

NightlyPer Location

📈

Your History

Every irrigation event
recorded from day one
at your location

PassiveAutomatic

AI Intelligence Engine

🧪

Soil Science

Computes your correct irrigation threshold from USDA soil data + your plant type — not a guess

FC − (PAW × MAD)

⚡

Weather

Adjusts daily based on rain forecast and heat demand — assigns SUPPRESS, DELAY, NORMAL or BOOST

ET₀ × Kc → RPS

🤖

Machine Learning

After 90 days, XGBoost trains on your local history and refines every decision automatically

Activates after 90 days

Today's Irrigation Decision

SUPPRESS

Rain is coming — significantly raise threshold to prevent watering

DELAY

Rain possible — raise threshold, water less

NORMAL

No change — base threshold in use

BOOST

Hot & dry — lower threshold, water more

💧

Solenoid opens exactly when needed — and not a moment sooner

Right amount of water · Right time · Right plant · No manual intervention

Your AI Gets Smarter Over Time

🟢

Day 1

Soil science + weather active immediately

📊

Day 30

Irrigation history building

🤖

Day 90

XGBoost model trained and validated

🚀

Ongoing

ML decisions continuously improve

The Three Watering Modes

Two toggles on the device detail screen control the active mode: AI and Manual. When AI is on, Manual is disabled. Turn AI off to enter Threshold mode — that enables the Manual toggle. To run a manual cycle, set the watering duration first, then toggle Manual on. Toggle Manual off to stop and return to Threshold mode.

Device vitals screen in AI mode showing NORMAL modifier badge and soil moisture bar — AI Mode — weather modifier badge, live RPS and ET₀ values.

Device vitals screen in Threshold mode showing manual moisture slider — Threshold Mode — fixed moisture % you set manually.

Device vitals screen in Manual mode showing on/off toggle and cycle duration — Manual Mode — direct solenoid control with duration timer.

🤖 AI Mode

AI toggle on. Thresholds computed from soil science, adjusted nightly by weather. Manual toggle disabled.

📊 Threshold Mode

AI toggle off, Manual toggle off. Watering triggers when moisture drops below your set threshold.

🔧 Manual Mode

AI off, Manual toggle on. Set duration first, then toggle on. Toggle off to stop and return to Threshold.

The Watering Mode indicator at the bottom of the vitals card shows the current mode — AI, Threshold, or Manual — and pulses while active. Use the AI and Manual toggles to change modes.

The moisture bar color shows soil status at a glance: red/coral = below threshold (watering due), orange/yellow = approaching threshold, blue/green = well above threshold. The droplet icon indicates whether the solenoid is currently open.

Long Watering Protection

All devices include built-in protection against runaway irrigation — a stuck solenoid, a downstream leak, or a threshold that keeps re-triggering. Three escalating alert thresholds are configured per customer account under Customer Details → Long Watering Alerts.

⚠️ Minor — 300 sec

Push notification sent. Watering continues — this is an early heads-up that the cycle is running long.

🔶 Major — 420 sec

Second notification sent. Watering still running — customer should check the device or manually stop from the app.

🚨 Critical — 540 sec

Watering stopped automatically. Device switches to Threshold mode with threshold set to 0% — no automatic watering until the customer reviews and re-enables.

Notifications are delivered via push notification and email — both configurable per customer account. The critical shutoff is automatic and does not require the customer to be watching the app.

After a critical shutoff the device remains in Threshold mode at 0% — it will not water again automatically until the customer increases the threshold or switches back to AI mode. This prevents repeated triggering while the root cause is investigated.

Customer Details page showing Long Watering Alerts with Minor 300 sec, Major 420 sec, Critical 540 sec thresholds and text and email notification options — Long Watering Alerts — configured per customer account with escalating thresholds.

Setting Up AI Mode

Go to Device Settings → Watering & AI Setup and configure two fields before enabling AI mode:

1Soil Type — the texture at sensor depth. Tap Auto-detect from GPS to look up your soil type from the USDA soil survey automatically.

2Application Type — what you're irrigating. Determines how conservatively the AI manages moisture. Drip Defender is locked to Foundation.

If either field has not been configured when you tap to enable AI mode, a setup prompt will appear guiding you through soil type and application type selection before AI activates.

Also available: Soil Temperature and Moisture Calibration offsets (±18°F / ±10%) and Soil Moisture Sensitivity (1–5% hysteresis band).

Device Settings → Watering & AI Setup — soil type, application type, and calibration.

Soil Parameters

The AI uses USDA-NRCS hydraulic properties for each of 13 soil texture classes to compute the irrigation threshold. The three key values per soil type:

FC — Field Capacity

Moisture % the soil holds when full. The upper "stop watering" bound.

PWP — Wilting Point

Moisture % below which plants can't extract water. The danger floor.

PAW = FC − PWP

Plant-Available Water — the usable range the AI manages within.

Threshold Low = FC − (PAW × MAD) | Threshold High = FC × 0.95

Service Manager Soil Parameters table showing all 13 soil texture types — All 13 soil types — configurable in Service Manager.

Edit Soil Parameter form showing Sandy Clay Loam with HSG C, FC 30, PWP 18, PAW 12, Ksat 0.15 — Sandy Clay Loam detail — HSG C triggers cycle-soak mode.

Soils with HSG C or D (Sandy Clay Loam through Clay) have low infiltration rates. The AI automatically enables cycle-soak mode — delivering irrigation in three shorter bursts with a 20-minute pause between each to prevent surface runoff.

Application Types

The application type tells the AI what you're irrigating and how conservatively to manage soil moisture. It sets the MAD (Management Allowed Depletion) — the fraction of plant-available water that can be consumed before irrigation triggers. Lower MAD = more frequent watering. Higher MAD = longer intervals. Water Buddi supports 47 application types across 9 segments.

Foundation application type detail showing MAD 0.3, High Buffer 0.9, Segment Foundation — Foundation — MAD 0.30, tight buffer 0.90 to prevent over-saturation.

Vineyard application type detail showing MAD 0.45, High Buffer 0.95, Segment Agricultural — Vineyard — MAD 0.45, Agricultural segment, deficit irrigation strategy.

Crop Grains application type detail showing MAD 0.55, Label Field Crops — Grains (Corn, Wheat, Soy) — Field Crops — Grains — MAD 0.55, seasonal Kc multipliers apply.

Foundation (MAD 0.30)

Conservatively moist at all times. Prevents shrink-swell cycles in expansive soils that cause foundation movement and structural cracking.

Vineyard (MAD 0.45)

Mild deficit irrigation — intentional controlled water stress at the right growth stages improves grape concentration and wine quality.

Field Crops — Grains (MAD 0.55)

Tolerates moderate depletion. Seasonal Kc multipliers adjust water demand automatically from 0.20× in winter to 1.15× at summer peak.

Weather Modifiers

Every night at 02:00 UTC the weather batch fetches a localized forecast for each device's GPS location, computes a Rain Propensity Score (RPS) from 0–100, and assigns one of four modifiers for the coming day.

SUPPRESSRPS ≥ 75 — Rain is highly likely. Irrigation skipped entirely. Delta −5 applied but solenoid locked closed regardless of moisture level.

DELAYRPS 40–74 — Rain is possible. Threshold raised by −2.5% — soil must be drier than usual before watering triggers. Conserves water ahead of a wet event.

NORMALRPS 20–39 — No significant weather adjustment. Base threshold used unchanged. Most common daily state.

BOOSTRPS < 20 + high ET₀ — Hot and dry. Threshold lowered by +3% — device waters proactively to compensate for high atmospheric water demand.

Service Manager AI Modifiers table showing BOOST +3, NORMAL 0, DELAY -2.5, SUPPRESS -5 — AI Modifiers table — four scoring bands with configurable deltas.

Edit AI Modifier form for SUPPRESS showing RPS minimum 75, delta -5, pill color bg-blue-600 — SUPPRESS modifier detail — RPS minimum 75, delta −5, blue pill.

Machine Learning Engine

After 90 Days

From day one, every irrigation event is recorded — soil moisture, temperature, humidity, pressure, ET₀, and RPS at that exact moment. After 90 days this dataset trains an XGBoost gradient-boosted model tailored to your location.

Use Trained Model — when toggled on, nightly scoring uses the ML model instead of fixed RPS bands. Toggle off to revert instantly — no downtime, no code deploy.

Per-Cluster Models — trains a separate model per geographic cluster (≥20 events/cluster) to capture micro-climate variations not visible in regional forecasts.

Train Now — triggers immediate training outside the weekly schedule. Takes 1–3 minutes. LED turns green when a validated model is ready to activate.

AI Engine — tracks data collection progress and controls model activation.

Crop Coefficients (Kc)

ET₀ measures atmospheric water demand — but a vegetable garden and a drought-tolerant shrub bed respond very differently to the same hot day. The crop coefficient (Kc) scales ET₀ to the actual water demand of each plant type (FAO-56 standard):

ET_actual = ET₀ × Kc | Kc range: ~0.40 (drought-tolerant) to 1.15 (high-demand crops)

Root depth (cm) is also stored per application type — it determines the soil water storage volume available to the plant, which scales BOOST adjustments so deep-rooted established trees aren't over-watered on hot days.

Service Manager Kc Parameters table showing 47 application types with Kc values — berries 1.05, container 1.1, crop_grains 1.15 — Kc Parameters — 47 application types with FAO-56 crop coefficients.

Edit Kc Parameter for crop_hay_alfalfa showing Kc 1.05, Root Depth 90cm, Notes Alfalfa Kc_mid FAO-56 — Hay/Alfalfa — Kc 1.05, root depth 90cm. High water demand crop.

Edit Kc Parameter for golf_fairways showing Kc 0.85, Root Depth 25cm, Notes Fairway turf managed lower than greens — Golf Fairways — Kc 0.85, root depth 25cm. Managed turf, moderate demand.

Seasonal Kc Multipliers

Agricultural crops have significant seasonal variation in water demand driven by growth stage. Monthly multipliers adjust the base Kc automatically:

Effective Kc = Kc_base × Kc_multiplier[month]

Example — Grains (base Kc 1.15): January multiplier 0.20 → effective Kc 0.23. June multiplier 1.15 → effective Kc 1.32 at summer peak. 120 monthly records cover all crop-type application types.

Seasonal Kc — crop_grains monthly multipliers, page 1 of 12.

Page 2 — crop_hay_alfalfa and golf_greens seasonal adjustments.

Soil Drift Alerts

The system monitors every device for divergence between its configured soil type and observed irrigation behavior. If the actual moisture trigger point consistently differs from the AI threshold by more than 15% across at least 10 irrigation events, a drift alert is raised.

Common causes:

Soil type doesn't match the actual texture at sensor depth
Compost, mulch, or peat has changed water retention
Sensor has shifted to a different depth
Compaction or organic matter changes over time

To resolve: update the Soil Type in Device Settings. Compute Now runs an immediate drift check outside the nightly schedule.

Soil Drift Alerts — all devices in normal range. Compute Now for immediate check.

Frequently Asked Questions

It rained — why did my device still water?

The weather batch runs at 02:00 UTC using tomorrow's forecast. SUPPRESS prevents irrigation when rain is predicted — it doesn't respond to rain that already fell. Tonight's batch will reflect what happened.

What does "No weather adjustment" mean?

Your device received the NORMAL modifier — no significant rain or extreme heat forecast. Base threshold is used unchanged. Most common daily state.

The threshold feels wrong for my plants. What should I try?

First try a different Application Type — that drives MAD and how conservative the threshold is. If you need full control, switch to Threshold mode and set a moisture % manually.

When does the machine learning model activate?

After a minimum of 90 days of irrigation event data. Training is automatic. Support staff review and activate it — expect 3–4 months from installation.

What is ET₀ and why does it matter?

ET₀ (reference evapotranspiration) measures atmospheric water demand in mm/day. A hot, sunny, windy day pulls far more moisture from soil than a cool cloudy one. The AI uses ET₀ × Kc to scale BOOST adjustments to your specific plant type's actual water demand.

Can I use AI mode on both Drip Defender and Water Buddi?

Yes. Drip Defender is locked to the Foundation application type. Water Buddi supports all 47 application types across Lawn, Garden, Trees, Golf, Sports, Commercial, and Agricultural categories.