Remote SensingScientist Academy
Menu
Apply for cohort

Field Lab 07 · operational tutorial · western Estonia

Plan, fly and process an eBee mission

Follow one chronological workflow from the ecological question to an analysis-ready raster stack. Every stage names its inputs, output, software location, quality check, failure response and next action.

Begin at Step 01 ↓Open the detailed Drone Lab →
30 Jun–2 Jul 2024Acquisition window
Saardu · KeemuWestern Estonia sites 1–2
Koera · KudaniWestern Estonia sites 3–4
eBee X · 120 plotsFixed-wing mapping + 1 m² references

How to use this lab

Do the step. Check the evidence. Then continue.

  1. Field LabUse Steps 01–22 to understand and execute the complete mission.
  2. Drone LabOpen the workstation-level eMotion, PPK, GCP and Pix4D procedure.
  3. Reference deskOpen the science explanation linked from a step only when you need it.
Download the complete 22-step mission checklist ↓

The sensor records radiation and image geometry. Field observations supply ecological meaning and reference responses. The UAV did not directly measure species identity, CCI, leaf area, vegetation height or AGB.

Phase 0 · before the flight

Preparation makes the mission traceable

This eBee X checklist reconciles the 2024 field notes with the documented PPK/GCP workflow. DJI-specific calibration or base-station actions are intentionally not transferred into this SOP.

A

At home · before leaving

B

On site · before launch

C

Instructional template · do not invent missing values

2024 UAV field note

Site
________________
Date
________________
Mission name
________________
Flight / log number (including the recorded suffix)
________________
Aircraft
________________
Camera / payload
________________
Multispectral / RGB / thermal
________________
Start time
________________
End time
________________
Time zone
________________
Planned altitude
________________
Planned GSD
________________
Forward overlap
________________
Side overlap
________________
Wind
________________
Cloud / light conditions
________________
GCP IDs
________________
Number of images
________________
Notes / warnings
________________
Landing condition
________________
Operator comments
________________

Clickable scientific workflow

See what each stage means before you operate it

Open a card for the meaning, the 2024 example, its scientific purpose, the check and a direct link to the relevant tutorial step.

  1. WHAT IT MEANSDefine the response, observational unit and claim before choosing a sensor.

    OUR 2024 EXAMPLERelate plot-level field responses to plot summaries of UAV predictors.

    WHYThe question determines the products, support and validation—not the other way around.

    WHAT TO CHECKCan you state the response, predictor family and non-claim?

    Go to the relevant practical step →

Begin here · complete operational sequence

Plan, fly, process and hand off one defensible mission

Open one step at a time. Every step names the action, workspace, evidence entering and leaving, stop condition and next decision.

  1. WHAT

    Turn the research question into a flight specification before opening the mission planner.

    ACTION
    1. Name the ecological response and the field observation that represents it.
    2. Draw the site boundary and identify the 1 m² plot support.
    3. Choose the required products, sensor, GSD, altitude and overlap as one connected decision.
    WHERE

    Project brief + GIS map; then the mission-planning worksheet.

    WHY

    Altitude and overlap are not defaults to copy. They determine coverage, image geometry, processing load and whether the raster support can answer the question.

    INPUT
    • Research question
    • Site boundary
    • Field-plot design
    • Required raster products
    OUTPUT
    • Mission specification
    • Sensor choice
    • Target GSD and overlap
    • Explicit non-claims
    CHECK
    • The requested product can support the ecological response.
    • The boundary includes safe operational margin.
    • GSD is reported separately from effective spatial resolution.
    IF THIS FAILS

    Stop and revise the question or product list. Do not compensate for an unclear objective by collecting more imagery.

Real 2024 project evidence · Saardu

Same site, different measurement product

Every panel comes from a verified processed Saardu project raster. Select a product to compare what changes when the measurement domain, spectral band or derivation changes over the same survey footprint.

RGB processed map for the Saardu 2024 UAV study site
RGBProcessed / derived 2024 project example · Saardu

FORMULA / SOURCES.O.D.A. RGB orthomosaic

BANDS / DOMAINVisible Red + Green + Blue channels

WHAT YOU ARE SEEINGThe processed visible-light orthomosaic for the same Saardu survey footprint.

WHY IT LOOKS DIFFERENTThe source RGB colour rendering is retained and resampled to the common teaching grid; it is not a single reflectance band.

WHAT THE PIXELS REPRESENTDisplayed red, green and blue channel values in the processed orthomosaic.

WHAT WE CAN INFERVisible spatial context, boundaries and texture after geometric QA.

WHAT WE CANNOT INFERQuantitative surface reflectance, a plant trait or species identity from colour alone.

Same footprint: image, bands and derived products

Supporting science · open when a step links here

Field Lab reference desk

The operational timeline is the spine of this lab. These concise chapters explain the science behind its decisions without interrupting the chronological workflow.

01
Flight design and the fixed-wing decisionPlatform · altitude · overlap · GSD

A fixed-wing eBee X was appropriate for systematic coverage of spatially extensive meadow sites. A multirotor remains preferable for hovering, tight launch areas and fine local inspection; neither platform is universally superior.

FIXED-WINGSystematic area mapping
  • Efficient forward coverage
  • Longer endurance potential
  • Cannot hover
  • Needs suitable launch/recovery space
MULTIROTORFlexible local inspection
  • Hover and oblique views
  • Tighter launch/landing
  • Fine local positioning
  • Usually less area per battery
SystemPayloadAltitudeOverlapOutput GSDPurpose
MultispectralParrot Sequoia106–109 m AGL80% forward · 75% side≈10 cm/pixelGreen, Red, Red Edge and NIR reflectance products
RGBsenseFly Duet T · S.O.D.A. RGB119–125 m AGL85–88% forward · 80–86% side≈2.7 cm/pixelVisible context, texture, boundaries and photogrammetric geometry
ThermalsenseFly Duet T · 640 × 512 thermalProject mission recordPaired Duet T acquisition≈15.6 cm/pixelThermal infrared product; not reflectance
  1. 01
    Define survey boundary

    Include the target meadow, required plots and operational margin; identify obstacles and an approved fixed-wing recovery area.

  2. 02
    Choose sensor

    Choose RGB, Sequoia or Duet T from the required measurement domain—not from convenience alone.

  3. 03
    Choose required GSD

    Relate research detail to nominal GSD, altitude, camera footprint, coverage and practical endurance.

  4. 04
    Set overlap

    Set forward and side overlap so consecutive images and adjacent flight lines share enough stable content for reconstruction.

  5. 05
    Choose landing approach

    Evaluate the landing direction and area under the current approved eBee procedure; this tutorial does not replace manufacturer training.

  6. 06
    Plan GCP distribution

    Where control is used, distribute visible surveyed targets around and within the block; avoid one corner or a single line.

  7. 07
    Save and document

    Record mission name, site, flight date, payload, planned GSD, altitude and both overlaps; capture the settings screen.

Pixel size is not always effective spatial resolution. Blur, motion, reconstruction, geolocation and resampling can reduce the detail that is scientifically distinguishable.

02
What RGB, Sequoia and Duet T measureSpectrum · reflectance · thermal emission · indices

Band explorer

550 nm centre · 40 nm bandwidth

Green

Visible green reflectance; useful context for vegetation colour and green–NIR contrasts. Pigments, illumination, canopy and background all contribute.

Project index

NDVI

(NIR − Red) / (NIR + Red)

Why these bands: Red absorption contrasted with NIR scattering.

Project role: General vegetation-condition contrast.

Boundary: Can saturate; soil, canopy, atmosphere and shadows matter.

RGB · S.O.D.A.

Visible colour and geometry

High-resolution visible imagery supports context, boundaries, texture and reconstruction. It is scientific evidence, not merely a normal picture.

SEQUOIA

Four reflectance bands

Green 550/40 nm, Red 660/40 nm, Red Edge 735/10 nm and NIR 790/40 nm are calibrated separately.

DUET T

Thermal emission

S.O.D.A. RGB is paired with a 640 × 512 thermal camera. Thermal values represent emitted infrared/apparent surface temperature when calibrated—not reflectance.

THERMAL ≠ REFLECTANCETHERMAL ≠ DIRECT SOIL MOISTURETHERMAL ≠ DIRECT PLANT STRESS
03
Positioning and controlTime · RINEX · PPK · GCP · CRS

PPK combines rover and reference GNSS observations after flight. RINEX transfers receiver-independent observation/navigation data. In the documented 2024 project example, reported geotag uncertainty improved from approximately 0.806 m to 0.049 m; that is not a universal guarantee of map accuracy.

TIME CHECK

For the July 2024 mission, Estonia local time was UTC+3. Verify the source time basis and request reference data for the complete flight plus margin.

GEOMETRY

Where is this pixel?

  • geotagging and PPK
  • GCPs
  • EPSG:3301
  • bundle adjustment
  • DSM and orthorectification
RADIOMETRY

What does its value represent?

  • sensor response
  • illumination and calibration
  • reflectance
  • thermal signal
04
What Pix4D reconstructsTie points · bundle adjustment · dense surface · orthorectification

  1. WHAT IT MEANSPhotographs view the same ground from several camera positions.

    OUR 2024 EXAMPLEForward and side overlap create repeated coverage along and between flight lines.

    WHYReconstruction requires common image content.

    WHAT TO CHECKDoes the analysis area have sufficient achieved overlap?

    Go to the relevant practical step →

Pix4D is not simply stitching pictures. Repeated features link overlapping views; bundle adjustment estimates camera geometry; dense reconstruction estimates the visible surface; the DSM then supports orthorectification into map geometry.

05
Products, indices and failure boundariesRGB · DSM · reflectance · thermal · vegetation indices
ProductStatusPixel / sample meaningScaleUseLimitation
RGB orthomosaicprocessedOrthorectified visible-colour mosaic≈2.7 cm/pixelContext, boundaries, textureColour and texture are not species or trait measurements
Green reflectanceprocessedCalibrated Green-band reflectance≈10 cm/pixelVisible vegetation/background informationIllumination and calibration quality remain consequential
Red reflectanceprocessedCalibrated Red-band reflectance≈10 cm/pixelPigment-sensitive contrast; NDVI inputNot a direct chlorophyll reading
Red Edge reflectanceprocessedCalibrated Red Edge reflectance≈10 cm/pixelRed-edge indices and modellingSensitive to registration and calibration
NIR reflectanceprocessedCalibrated NIR reflectance≈10 cm/pixelCanopy/leaf scattering and indicesWater, shadows and background can dominate
DSMderivedElevation of the visible upper surfaceoutput-grid dependentSurface structure and height-related predictorsDSM is not a bare-earth DEM
Point cloudderived3-D reconstructed surface samplesirregular 3-D supportGeometry QA and DSM productionDensity is not accuracy
Thermal productprocessedThermal infrared signal / apparent surface temperature where calibrated≈15.6 cm/pixelSurface-temperature pattern inspectionNot reflectance, soil moisture or plant stress directly
Vegetation indicesderivedFormula applied to compatible reflectance bandsaligned output gridCompact spectral predictorsBand ≠ index; index ≠ trait
BAND ≠ INDEXINDEX ≠ TRAITTRAIT PREDICTION ≠ DIRECT MEASUREMENT
NDVI(NIR − Red) / (NIR + Red)

Why: Red absorption contrasted with NIR scattering.

Use: General vegetation-condition contrast.

Failure boundary: Can saturate; soil, canopy, atmosphere and shadows matter.

GNDVI(NIR − Green) / (NIR + Green)

Why: Green-to-NIR contrast.

Use: Complementary canopy and pigment-sensitive predictor.

Failure boundary: Does not directly measure chlorophyll or photosynthesis.

SAVI((NIR − Red) / (NIR + Red + 0.5)) × 1.5

Why: Adds a documented soil-adjustment term.

Use: Sparse vegetation and exposed-background conditions.

Failure boundary: L = 0.5 is a parameter choice, not universal correction.

MSAVI(2NIR + 1 − √((2NIR + 1)² − 8(NIR − Red))) / 2

Why: Adaptive red–NIR soil adjustment.

Use: Complementary predictor where soil background is visible.

Failure boundary: Requires valid reflectance, masks and a non-negative radicand.

RNDVI(NIR − Red) / √(NIR + Red)

Why: Renormalizes the red–NIR contrast.

Use: Vegetation-density sensitivity in the project predictor set.

Failure boundary: Not NDVIRe; denominator and reflectance scale must be valid.

RTVIcore100(NIR − Red Edge) − 10(NIR − Green)

Why: Combines NIR, Red Edge and Green contrasts.

Use: Red-edge-sensitive canopy/pigment predictor.

Failure boundary: Empirical relationship must be validated for this campaign.

SReNIR / Red Edge

Why: NIR-to-Red Edge ratio.

Use: Canopy-condition and red-edge predictor.

Failure boundary: Unstable when the denominator is invalid or near zero.

CIre(NIR / Red Edge) − 1

Why: Shifted red-edge simple ratio.

Use: Chlorophyll-associated project predictor.

Failure boundary: An index, not a direct CCI or chlorophyll measurement.

HIGH NDVI ≠ HIGH BIODIVERSITYHIGH NDVI ≠ AUTOMATICALLY HIGH BIOMASSHIGH NDVI ≠ DIRECT CHLOROPHYLL

Final failure gallery

A completed process can still be unusable

STOP 01Wrong mission or unexplained image count
STOP 02Wrong UTC/local-time conversion or incomplete RINEX interval
STOP 03Weak or undocumented PPK solution
STOP 04Image/log association failure
STOP 05Wrong camera model or image group
STOP 06Wrong or ambiguous CRS/vertical reference
STOP 07Unverified radiometric calibration
STOP 08Insufficient achieved overlap or motion blur
STOP 09Seamlines, ghosting or water matching failure
STOP 10Band misregistration or mismatched raster grids
STOP 11NoData or scaling ambiguity
STOP 12Field plots outside accepted coverage

Portfolio challenge

Produce an auditable mission handoff

Evidence and software references

Verified project records and current technical documentation