# FieldLens 3D: scientific basis and model notes **Adaptive AgroTech · Remote sensing for precision farming** This package demonstrates a realistic *decision workflow*: map variation, inspect multiple sensing layers, check acquisition quality, collect independent field evidence, and approve a targeted investigation. All imagery, reflectances, measurements, field geometry, time series, and scenario outcomes are synthetic. No real farm data, actual geographical coordinates, calibrated crop model, predicted yield, or validated economic savings are used. The export is a scouting record, not a machine-ready fertilizer or irrigation prescription. ## What the demo represents The 240 m × 160 m synthetic maize field has an area of 3.84 ha. The 24 × 16 analysis grid contains 384 cells, each 10 m × 10 m, or 0.01 ha. These are **decision cells**, not raw camera pixels. A survey's centimetre-scale imagery can be aggregated into a much coarser management grid in real work. Crop details in the scene are representational geometry and are not individual plant measurements. The scenario contains a water-limited patch, a nutrient-limited patch, poor establishment, a wetter patch, and comparatively vigorous reference areas. The hidden simulated field state is revealed only through the ground-check action. The initial zone names describe observed patterns rather than diagnosing their cause. A warm area or a low red-edge index can have several causes; some edge-of-patch alerts deliberately do not resolve to a confirmed constraint after a ground check. ## Spectral layers The visible RGB-like scene conveys canopy appearance; it is rendered geometry, not a photograph. The model stores plausible synthetic reflectances for red, near infrared and red edge. It calculates each cell's indices from the exact reflectances: `NDVI = (NIR − Red) / (NIR + Red)` `NDRE = (NIR − RedEdge) / (NIR + RedEdge)` NDVI responds to vegetation greenness and cover because chlorophyll absorbs red light while leaves reflect much of the near infrared. Dense canopies can reduce NDVI's sensitivity to additional canopy variation. Red-edge measurements may retain useful sensitivity to chlorophyll variation in denser canopies. Neither index independently identifies disease, nitrogen shortage, or an appropriate input rate. Soil background, crop stage, canopy structure, other stresses and sensor calibration matter. [1–3] The model's default reference is approximately NDVI 0.83 and NDRE 0.45. These values and the classification cutoffs are **illustrative choices for this synthetic maize scenario**, not universal crop-health thresholds. Synthetic reflectance residuals vary with the illumination challenge; the demo does not simulate a complete radiometric processing pipeline. ## Thermal interpretation The thermal layer displays **surface temperature**. In well-covered cells it is predominantly a canopy signal; in sparse cells exposed soil contributes. The model's notional air temperature is 24 °C. Water limitation warms the synthetic crop, while wet patches are cooler. A warm measurement alone is not enough to diagnose drought. Real thermal interpretation requires consideration of canopy cover, soil background, wind, radiation, humidity, emissivity, sensor calibration and acquisition time. Research combining vegetation and thermal measurements has demonstrated the importance of separating surface composition from crop water status. This package does **not** implement or label its output as CWSI or WDI: these indices require additional observations and assumptions that are outside the model. [4] ## Acquisition quality and spatial resolution The displayed quality/confidence score is a **synthetic relative quality score, not a statistically calibrated probability or an accuracy claim**. The changing-illumination control lowers the score and slightly perturbs the optical signal; high values represent a deliberately difficult survey with variable light, not the statement that every uniform overcast flight is unusable. Real acquisition needs a suitable reflectance calibration workflow and appropriate image overlap. [5] The optical ground sampling distance uses nadir-view pinhole geometry: `GSD [cm/pixel] = height [m] × pixel pitch [m] / focal length [m] × 100` The hypothetical optical camera has 3.45 μm pixel pitch and an 8.8 mm actual focal length. Thus 60 m gives 2.35 cm/pixel; 80 m gives 3.14 cm/pixel. These assumptions do not describe a named camera. GSD is pixel spacing on the ground, **not positioning accuracy**. The package does not compute a thermal camera GSD or simulate the loss of every subpixel plant detail as height changes. [6] The serpentine mission, flight time and coverage visualization are conceptual. They are not operational aviation guidance, a legal flight authorization, or a complete photogrammetric survey design. Real flight planning also needs wind, terrain, clearance, forward/side overlap, image exposure and mission reserves. ## Ground checks and management decisions The ground-check action produces a **simulated** 0–20 cm soil volumetric water content, a synthetic tissue-N result when adequate canopy exists, and a canopy-cover estimate. Local illustrative comparator values are 28% soil VWC, 3.35% tissue N on a dry-matter basis, and 94% canopy cover. These are scenario reference values, not universal adequacy thresholds. Real measurements require representative sampling, crop-stage-specific interpretation, root-zone information and laboratory or sensor quality control. For sparse stands, a tissue result may be absent; the interface must show it as unavailable rather than zero. A stand gap can cause low NDVI and a warmer surface without implying a fertilizer requirement. Nutrient assessment requires independent agronomic evidence. Irrigation depth requires the soil water-holding capacity, rooting depth, current deficit, crop demand and system constraints; a zone map alone does not supply it. [7] Approval records a human decision to prioritize the suggested investigation. It does not activate a sprayer, irrigator or autonomous equipment. The code never assigns a kg N/ha fertilizer rate or a guaranteed irrigation saving. Changing the water or nutrient scenario invalidates previous simulated field checks; changing acquisition conditions retains those checks because the simulated soil and crop state has not changed. ## Temporal view The six survey points run from 25 days before the current observation to today. They are deterministic synthetic comparison records illustrating divergence between zones as the canopy develops. They are not a forecast, a measured season, or the outcome of an applied treatment. Today matches the current cell's displayed indices and temperature. ## Transparent model assumptions In `js/data.js`, smooth deterministic spatial functions define scenario patches. Red, NIR and red-edge values are generated to give physically plausible illustrative index contrasts. The exported calculation is reproducible and does not use a neural network or claim trained AI inference. The default field has 101 cells (1.01 ha) flagged for investigation; this count varies with scenario and quality controls. No percentage of yield increase, water saved, fertilizer saved or carbon avoided is claimed. The value demonstrated is locating variation, directing ground investigation, maintaining an evidence trail and leaving operational decisions with people. ## Sources Sources were checked on 13 October 2026. They support the remote sensing principles and workflow. They do **not** validate the synthetic numerical model. 1. **NASA Earth Observatory.** *Measuring Vegetation (NDVI & EVI).* Explanation of red/NIR reflectance and NDVI. https://science.nasa.gov/earth/earth-observatory/measuring-vegetation-ndvi-evi/ 2. **U.S. Geological Survey.** *Landsat Normalized Difference Vegetation Index.* Standard red/NIR index and surface-reflectance products. https://www.usgs.gov/landsat-missions/landsat-normalized-difference-vegetation-index 3. **MicaSense.** *Overview of Agricultural Indices.* Sensor manufacturer's explanation of NDVI sensitivity, red-edge measurements, and interpretation limits. https://support.micasense.com/hc/en-us/articles/227837307-Overview-of-Agricultural-Indices 4. **Hoffmann, H., Jensen, R., Thomsen, A., Nieto, H., Rasmussen, J., & Friborg, T. (2016).** *Crop water stress maps for an entire growing season from visible and thermal UAV imagery.* Biogeosciences, 13, 6545–6563. Primary research; barley, not maize. Used for the principle of combined thermal/vegetation interpretation, not as calibration of this demo. https://doi.org/10.5194/bg-13-6545-2016 Author institution record: https://pure.au.dk/portal/en/publications/crop-water-stress-maps-for-an-entire-growing-season-from-visible-/ 5. **MicaSense.** *Best practices: Collecting Data with MicaSense Sensors.* Manufacturer's acquisition and calibration guidance. https://support.micasense.com/hc/en-us/articles/224893167-Best-practices-Collecting-Data-with-MicaSense-Sensors 6. **PIX4D.** *Computing the Flight Height for a given GSD.* Camera geometry, actual focal length, altitude and pixel spacing. https://support.pix4d.com/hc/en-us/articles/202557469 7. **University of Minnesota Extension.** *Variable Rate Irrigation.* Operational explanation of management zones, soil information and irrigation scheduling. https://extension.umn.edu/natural-resources/conservation/agricultural-soil-and-water/irrigation/variable-rate-irrigation No third-party source text or imagery is embedded in the scene. The Adaptive AgroTech logo remains a separate, unmodified branding asset.