Satellite Blueberry Crop Advisory: 124-Acre Proof

The reported outcome in one frame. Both 30–50% figures are the grower’s own estimates, given in our in-app feedback form. The rest of this article opens up the advisory he was reading and traces where each saving actually came from.

Most precision-agriculture case studies work backwards from a happy quote. This one works backwards from the data. In June 2026, a commercial blueberry grower in Citrus County, Florida, completed the feedback form attached to his satellite blueberry crop advisory and reported that the report had helped him optimise irrigation by 30–50%, optimise fertiliser application by 30–50%, and take action on pest, disease and weed pressure. Those are large claims. So we opened the same advisory he had been reading — all 25 blocks of it — and asked a harder question: which specific lines in this report could plausibly produce savings of that size?

The answer turned out to be unusually clean, and it is not the answer most people expect. The savings did not come from a clever irrigation formula telling him to apply less water on the days he irrigated. They came from the advisory telling him, on nearly half of all farm-days, not to irrigate at all — and from telling him, on eight of his twenty-five blocks, not to apply any fertiliser at all. The value was in the zeroes.

Illustrative. Southern Highbush blueberry on raised pine-bark beds — the production system this case study covers, and the pre-dawn hour the advisory kept recommending as the irrigation window. Representative imagery, not a photograph of the grower’s own blocks.

The operation: 25 blocks, four growth stages, one week

John Vansickle farms Southern Highbush blueberry (Vaccinium corymbosum) across 25 discrete blocks totalling 124.3 acres (50.3 hectares) in Citrus County, on Florida’s Nature Coast. Block sizes run from 1.77 acres to 11.29 acres. He onboarded onto our platform on 21 December 2025; his first advisory was generated the same day, drawing on a Sentinel pass captured three days earlier.

By the time he submitted the feedback that prompted this case study, the account had received 1,048 individual block advisories across 40 distinct satellite acquisition dates, at a median revisit interval of five days. This is the detail that makes the operation a useful test case: it is not a demo plot. It is a real commercial portfolio, monitored continuously for close to six months before the grower was asked whether any of it had been worth reading.

The complicating factor — and the reason a whole-farm average would have been useless to him — is that his 25 blocks were not in the same place in the season. On the week in question, the advisory placed them across four different phenological stages simultaneously.

Nearly half of all block-days in the week carried the same instruction: do not irrigate.

What the satellite blueberry crop advisory actually said

Our satellite blueberry crop advisory — delivered through JEEVN AI, our satellite and AI-based farm advisory product — issues a separate report per block per acquisition. Each report carries five working sections: irrigation, fertiliser, pest and disease, weed, and soil management, plus a growth and yield estimate. Below are the three sections the grower said had changed his decisions, shown as he saw them in the app.

Irrigation: the instruction was often “no water”

The irrigation section does not give a single weekly figure. It gives a day-by-day schedule for the seven days following the satellite pass, and for each day it publishes four things: the quantity in millimetres, the forecast rainfall, the probability of precipitation, and an evapotranspiration class. When forecast rain is sufficient, the recommended quantity drops to zero and the time window reads N/A.

The irrigation card as the grower sees it in JEEVN AI. Note the summary line: total water for the period, the number of irrigation days, and the best time window. The per-day rows expose the rainfall probability that drives whether a day is an irrigation day at all.

On one representative post-harvest block, the seven-day schedule for 12–18 June read: 3.5 mm, zero, 3.8 mm, zero, 4.0 mm, zero, 3.8 mm. The three zeroes were not omissions. They were issued against forecast rainfall of 2.0 mm, 1.5 mm and 0.5 mm at precipitation probabilities of 40%, 30% and 25% respectively. Total recommended irrigation for the week: 15.1 mm across four days, not seven.

Illustrative. This is what a skip day looks like from the ground. The advisory issues the zero before the rain arrives, off the forecast — which is the only way the water is actually saved rather than merely wasted more visibly. Representative imagery.

Two further details matter for the saving. First, the recommended window across the portfolio sat between 05:00 and 07:30 — pre-dawn, when evaporative loss is lowest. Second, where the advisory listed alternative delivery methods, drip was flagged as recommended at 3.8 mm while overhead sprinkler was listed at 4.5 mm and micro-sprinkler at 4.0 mm for the same day and the same crop demand. Choosing drip over overhead sprinkler is, on those numbers, roughly 16% less water for the same delivered requirement.

The obvious objection is that a satellite cannot know what a probe in the ground knows. It is a fair challenge, and the honest answer is that the two measure different things at different scales — a probe reads one point precisely, while the satellite reads the whole block approximately. For an irrigate-or-skip decision across 25 blocks, block-wide coverage is the more useful of the two.

Fertiliser: eight blocks were told to apply nothing

The fertiliser section reports five nutrients — nitrogen, phosphorus, potassium, sulphur and zinc — and for each one it publishes the current modelled soil composition, the ideal range for the crop, an application rate, and both an organic and a chemical source. Critically, when a nutrient already sits inside its ideal band, the application rate returns zero.

The fertiliser card triages the five nutrients into Need Attention / Monitor Closely / Optimal Level, then gives current-versus-target for each and names both an organic and a chemical source. Nutrients already inside the target band generate no application.

On the 11 June advisory, eight of the twenty-five blocks returned zero on all five nutrients at once — nitrogen, phosphorus, potassium, sulphur and zinc were all already inside their ideal ranges. For those eight blocks, covering roughly 30 acres, the correct action for the week was to spend nothing and apply nothing.

The remaining seventeen blocks did get rates, and the spread between them is the second half of the story. Prescribed nitrogen ranged from 6.5 to 46.9 kg/acre — a 7.2-fold difference across blocks of the same crop, on the same farm, in the same week.

Pest, disease and weed: stage-aware, not calendar-driven

The third section the grower credited is pest, disease and weed management. Across the 25 blocks, the 11 June advisory raised 36 separate high-confidence alerts — defined here as a modelled probability of 80% or above. Every alert carried both an organic and a chemical control option.

The pest, disease and weed card. Risks are counted and banded (high / moderate), tagged by type, given a percentage, and paired with both an organic and a chemical option — so the grower chooses the control, not just the target.

What makes this section credible rather than generic is that the alerts tracked growth stage. On the blocks at harvest, the dominant alert was Spotted Wing Drosophila (Drosophila suzukii) — a ripe-fruit pest — typically at 85%, peaking at 96% on one post-harvest block. On the three blocks still at bloom, the top alert was instead Flower Thrips (Frankliniella bispinosa) at 85% — a bloom-stage pest that would be meaningless on a harvested block. The advisory was not broadcasting one pest list across the farm.

Illustrative. Ripe fruit with intact bloom is exactly the target of Spotted Wing Drosophila, which is why the alert dominated the 20 blocks at or past harvest and was absent as the top risk on the three blocks still at bloom. Representative imagery.

On the weed side, Yellow Nutsedge (Cyperus esculentus) was flagged on 24 of 25 blocks and Crabgrass on 18, with probabilities reaching 92%. Bermudagrass and Goosegrass appeared on the remainder. Each weed alert, again, carried an organic option (heavy mulching, hand weeding) alongside a chemical one.

Reverse-engineering the reported 30–50%

So: can the advisory above actually produce a 30–50% reduction in water and fertiliser? We tested both claims against the portfolio data rather than accepting them.

Where the water saving comes from

Across the 25 blocks, the week’s advisory covered 175 block-days. Of those, 97 carried a positive irrigation instruction and 78 carried zero. The saving therefore depends entirely on what the grower would otherwise have done on those 78 days.

We modelled two honest baselines. Against a grower irrigating every day at each block’s own average recommended rate, following the advisory delivers a 44.4% reduction in applied water. Against a grower irrigating every day at each block’s own peak rate, the reduction is 48.7%. Both land inside the 30–50% band the grower reported, from the upper half of it.

There is an independent sanity check available. University of Florida IFAS guidance holds that a mature blueberry planting may need more than one inch of water per week during fruit development and rapid shoot growth. The advisory’s portfolio mean for this week was 16.5 mm — about 0.65 inches. That is well under the peak-season benchmark, and it should be: twenty of the twenty-five blocks were at or past harvest, where the crop coefficient the advisory cites (Kc 0.7–0.8 for post-harvest Vaccinium corymbosum) is materially lower than at fruit development. The saving comes from stage-aware demand modelling, not from under-watering the crop.

Where the fertiliser saving comes from

The fertiliser claim resolves the same way. Summed across the portfolio, the advisory prescribed about 1,764 kg of nitrogen. A uniform blanket programme applying the portfolio’s mean rate to all 124.3 acres would have applied roughly 2,528 kg — 30.2% more. A blanket programme pitched at the highest block’s rate, which is how growers often protect against the worst case, would have applied about 5,829 kg — 69.7% more.

So the reported 30–50% fertiliser saving is bracketed by our two baselines: 30.2% at the conservative end, 69.7% at the wasteful end. The grower’s own estimate sits precisely in that range.

Grower-reported outcomes, as submitted

The four cards below reproduce John Vansickle’s answers to our in-app feedback questionnaire for the advisory dated 12 June 2026, exactly as recorded. They are structured survey responses, not paraphrases.

Illustrative. The grower’s own public review of the app, left six months before this advisory, was four words long: “great app helps me in the field.” The advisory is built to be read standing in the row, not at a desk. Representative imagery, not a photograph of the grower.

Eight of twenty-five blocks were told to apply no nitrogen, phosphorus, potassium, sulphur or zinc at all.

We should be straight about the limits of this evidence. These are the grower’s own estimates, given in a three-question form, not metered water volumes or invoiced fertiliser tonnages. What our portfolio analysis establishes is consistency: the advisory he was reading contained instructions capable of producing savings of the size he reported, through a mechanism we can point at line by line. That is a stronger claim than a testimonial and a weaker one than an audit.

Why zone-level satellite data beats a whole-farm average

Underneath every section of a satellite blueberry crop advisory sits the same asset: a per-pixel read of the block from orbit. The advisory’s own references list the inputs — Sentinel-2 NDVI and NDWI for canopy vigour and water status, Sentinel-1 radar soil-moisture and vegetation indices, FAO-56 Penman–Monteith for reference evapotranspiration, and UF/IFAS blueberry crop coefficients to convert that into a blueberry-specific demand.

Illustrative. Within-block variation of the kind the satellite quantifies: most rows dense and dark, two irregular patches visibly thinner and paler. A whole-block average reports this field as healthy. The zone read tells you which twelfth of it to walk. Representative imagery.

Zone-level crop health for a single block. On this block the advisory scored 91.86% of the area healthy and 8.14% stressed — the purple patches. The neighbouring block, mid-renovation, reads as bare rows. An average would have hidden both facts.

That 8.14% figure is the whole argument for zone-level monitoring in miniature. A block reported as “92% healthy” is a block where a targeted intervention on one-twelfth of the area is worth more than a blanket pass over all of it — and where a farm-level average, blended across 25 blocks, would have reported nothing actionable at all.

The soil section rounds out the picture and shows the model is not simply agreeing with the grower. Across all 25 blocks, pH ran 4.8 to 5.2 — comfortably inside the 4.5–5.5 band blueberry requires, and a genuine strength of the operation. But soil organic carbon came back 0.10–0.20%, flagged low on every block. Salinity was low throughout. So the advisory’s standing recommendation across the portfolio was organic-matter building — composted pine bark, the organic source it names for sulphur — rather than more mineral fertiliser.

Yield estimation as the cross-check

Each block advisory also carries a growth and yield estimate. Across the portfolio these ran 1,950 to 3,510 kg/acre expected, against a modelled local potential band of 1,814–3,628 kg/acre. The lowest expected figure belonged to the same post-harvest block carrying the 96% Spotted Wing Drosophila alert — which is the internal consistency you want to see. The block flagged hardest on pest pressure was also the block projected to yield least.

Twelve of the twenty-five blocks are shown, ordered by area, with the portfolio total. The pattern that matters is in the last three columns: no two blocks got the same combination of water, nutrient and alert. That is the difference between a farm advisory and a weather forecast.

How to get this running on your own farm

The workflow that produced everything above is the standard one, and it does not require any hardware in the field.

1. Mark your blocks separately. This grower’s results depend on 25 boundaries, not one. Blocks at different stages need different advisories — draw them as they are actually managed.
2. Set crop and sowing date per block. The crop coefficient and the phenological model both key off this, and it is what makes stage-aware pest alerts possible.
3. Read the irrigation-days count, not just the volume. The skip days are where the water saving is.
4. Act on the zeroes in the fertiliser card. A nutrient inside its ideal band is a cost you can decline this cycle.
5. Correct the model when it is wrong. Stage labels and field data are grower-editable, and the feedback form is read — as the July correction in this case study shows.

Explore the platform at sat.farmonaut.com, read more about our satellite and AI-based farm advisory in the JEEVN AI overview, or contact our team for portfolio-scale onboarding. JEEVN AI is a paid product; plan detail is on the platform.

Frequently asked questions

Glossary

NDVI / NDWI

Satellite indices for canopy vigour and water content respectively, derived from Sentinel-2 reflectance. They drive the zone-level health map.

Evapotranspiration (ET) and Kc

ET is atmospheric water demand; the crop coefficient Kc scales it to a specific crop and growth stage. The advisory cites Kc 0.7–0.8 for post-harvest Southern Highbush blueberry.

FAO-56 Penman–Monteith

The standard method for computing reference evapotranspiration, named in the advisory’s own irrigation references.

Phenological stage

Where a block sits in its season — bloom, fruit development, harvest, post-harvest. It determines nutrient demand, water demand and which pests are relevant.

Soil organic carbon (SOC)

A measure of soil organic matter. Flagged low (0.10–0.20%) on every block in this portfolio, driving an organic-matter recommendation rather than a mineral one.

Case-study basis: JEEVN AI block advisories for a single Farmonaut account, satellite acquisition date 11 June 2026 (25 blocks), plus the account’s in-app feedback submission for the advisory dated 12 June 2026 and a public app-store review dated 22 December 2025. Grower name, crop, county, block count and area are reported with the account holder’s details as held on the platform. The 30–50% irrigation and fertiliser figures are the grower’s own estimates from a structured feedback form, not metered or invoiced measurements; the 44.4–48.7% water and 30.2–69.7% nitrogen figures are our modelled comparisons against stated baselines and are presented as ranges for that reason. Pest, disease and weed percentages are modelled probabilities, not confirmed infestations. Yield figures are estimates, not guarantees. Imagery: the four platform screenshots and the crop-health zone map are genuine captures of this account’s own advisory and satellite data; the five wider field photographs are AI-generated illustrative imagery, each labelled as such in its caption, and do not depict the grower, his blocks or his equipment. External agronomic benchmarks are attributed to University of Florida IFAS and the Florida Blueberry Growers Association as cited within the advisory. JEEVN AI is a paid product.