R08.04.01.01 · TASK CLASS

Yield maps & prescription maps

Clean yield data and build variable-rate prescriptions.

20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.

Typical applications: QGIS · John Deere Operations Center · FarmWorks
Class GUI-need label: HY · Hybrid

Task proposals · 20

Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.

R08.04.01.01-T01 Clean raw combine yield points and interpolate a 10 m yield surface in QGIS

Original instruction

In QGIS load input/yield_2025.shp and input/boundary.shp. Remove points with header up, speed <1.5 or >10 km/h, swath <80% of full width, yield outside mean ±3 SD, or within 15 m of the boundary. Save output/yield_clean.gpkg, then IDW-interpolate (power 2, radius 30 m) to a 10 m raster output/yield.tif.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click type · Text & forms
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 3 / 5

Proposed verifier

Kept point IDs must have Jaccard ≥0.98 with the gold cleaned set (SD computed after the other filters, per spec order). yield.tif: 10 m, field extent, RMSE <0.1 t/ha vs reference IDW.

Proposed partial credit

40% point filters (scaled by Jaccard); 20% headland removal; 40% interpolated raster.

R08.04.01.01-T02 Delineate three management zones from multi-year yield and NDVI in QGIS

Original instruction

In QGIS, cluster input/yield_norm_2023.tif, yield_norm_2024.tif, yield_norm_2025.tif and ndvi_2025.tif into 3 k-means classes, polygonize, dissolve, and eliminate polygons under 0.5 ha into their longest-border neighbour. Number zones 1–3 from low to high mean yield and save output/zones.gpkg.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click type · Text & forms vis · Visual judgment
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 3 / 5

Proposed verifier

Each zone IoU ≥0.8 with the gold zone after label matching; no polygon <0.5 ha; zone numbering ordered by mean yield. Zone count exactly 3.

Proposed partial credit

35% clustering; 20% sliver elimination; 30% zone IoU; 15% ordering.

R08.04.01.01-T03 Turn management zones into a variable-rate nitrogen prescription shapefile

Original instruction

In QGIS open input/zones.gpkg and add fields N_RATE (lb/ac) = 120, 150, 180 for zones 1, 2, 3 and N_KGHA (kg/ha, 1 decimal). Export to output/rx_n.shp in EPSG:4326 for the rate controller, with field names of 10 characters or fewer.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click type · Text & forms
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 2 / 5

Proposed verifier

rx_n.shp: CRS EPSG:4326, geometry equal to the zones, N_RATE per zone exact and N_KGHA = N_RATE × 1.1209 within 0.1.

Proposed partial credit

40% N_RATE per zone; 30% kg/ha conversion; 30% CRS and export format.

R08.04.01.01-T04 Create a zone-based seeding prescription in John Deere Operations Center (sandbox)

Original instruction

In the Operations Center sandbox org 'FrontierCUA Test Farms', upload input/zones.zip to field 'North 80', create a seeding prescription 'N80_Seed_2026' with 30,000/34,000/38,000 seeds/ac for zones 1–3 and product DKC62-08, and add it to a Work Plan. Do not send it to any machine.

Platforms: web · Any OS (web)
Applications: John Deere Operations Center
Interfaces: brow · Browser
GUI need · task setting: G2 · GUI-essential
GUI need · ordinary access: G2 · GUI-essential
Interaction primitives: click · Point & click type · Text & forms drag · Drag & drop
Work product: WP08 · App/system state
Conditions: risk · Irreversible/risky
Proposed difficulty: 3 / 5

Proposed verifier

Sandbox state queried by the grader: prescription exists on North 80 with name, three zone geometries matching the upload, the three rates and product; Work Plan references it; no transfer to a machine is logged. The agent has no API credentials.

Proposed partial credit

20% zones uploaded; 40% prescription rates and product; 20% work plan; 20% no machine send.

R08.04.01.01-T05 Post-calibrate a field's harvest data to scale tickets in FarmWorks

Original instruction

In FarmWorks Office import the 2025 monitor data in input/2025_harvest/ for field 'East 120' and apply a post-calibration so the field total equals the scale-ticket total of 412.6 t at the standard 15.5% moisture. Write the adjusted mean yield (t/ha) to output/answer.json and export the calibrated yield points to output/east120.shp.

Platforms: win · Windows
Applications: FarmWorks
Interfaces: desk · Desktop GUI
GUI need · task setting: G3 · GUI-only
GUI need · ordinary access: G3 · GUI-only
Interaction primitives: click · Point & click type · Text & forms
Work product: WP02 · Records/table
Conditions: licensed · Licensed pro app
Proposed difficulty: 3 / 5

Proposed verifier

Exported points must sum (by area) to 412.6±0.1 t at 15.5% moisture; mean yield within 0.01 t/ha of reference; calibration factor stored in the FarmWorks job. Points must retain FarmWorks attributes (not recomputed in a GIS). GIS tools and Python are not installed on the VM, and a file-system audit log flags an east120.shp not written by FarmWorks.

Proposed partial credit

25% data imported; 45% calibrated total; 20% mean yield; 10% export.

R08.04.01.01-T06 Digitize a grassed waterway and cut it out of management zones in QGIS

Original instruction

input/zones.gpkg covers a field crossed by a grassed waterway visible in input/ortho.tif. Digitize the waterway as a polygon in output/waterway.gpkg, remove it from all zones so nothing is applied there, recompute AREA_HA for each zone, and save output/zones_cut.gpkg.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: G2 · GUI-essential
GUI need · ordinary access: G2 · GUI-essential
Interaction primitives: path · Continuous path click · Point & click vis · Visual judgment nav · Pan/zoom/scroll
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 3 / 5

Proposed verifier

Waterway polygon IoU ≥0.7 with gold (no vector of it exists in inputs); zones overlap the gold waterway by <2% of its area; AREA_HA within 0.01 ha of geometry area. No waterway vector exists in the inputs and segmentation tools are absent; the PyQGIS console and qgis_process are disabled.

Proposed partial credit

50% waterway digitized (scaled by IoU); 30% removed from zones; 20% areas recomputed.

R08.04.01.01-T07 Build a variable-rate lime prescription from grid soil samples in QGIS

Original instruction

From input/soil_samples.gpkg (1 ha grid, pH and buffer pH) interpolate pH and buffer pH by IDW to 20 m, compute lime rate with the equation in input/lime_rec.pdf, set rates <0.5 t/ha to 0 and cap at 4 t/ha, and export 20 m grid polygons with RATE_T_HA to output/rx_lime.shp. Report total product (t) in output/answer.json.

Platforms: mac · macOS win · Windows
Applications: QGIS · Adobe Acrobat Reader
Interfaces: desk · Desktop GUI
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click type · Text & forms
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 4 / 5

Proposed verifier

Grid cell rates within MAE 0.05 t/ha of reference with min/max rules applied; total product within 2%. Grid aligned to the field boundary bounding box at 20 m.

Proposed partial credit

25% interpolation; 30% rate equation; 20% min/max rules; 25% export and total.

R08.04.01.01-T08 Generate a P and K fertility prescription in Ag Leader SMS for an InCommand display

Original instruction

In Ag Leader SMS Advanced import soil test results input/soil_2025.csv for field 'South 40', create P2O5 and K2O recommendations with the rules in input/recs.xlsx, build a dual-product prescription, and export it for an InCommand display to output/rx_pk/. Save the SMS project.

Platforms: win · Windows
Applications: Ag Leader SMS · Microsoft Excel
Interfaces: desk · Desktop GUI
GUI need · task setting: G3 · GUI-only
GUI need · ordinary access: G3 · GUI-only
Interaction primitives: click · Point & click type · Text & forms
Work product: WP04 · 3D/spatial model
Conditions: licensed · Licensed pro app multiapp · Multi-app
Proposed difficulty: 4 / 5

Proposed verifier

Exported prescription per-cell P2O5 and K2O rates within 1% of reference; product names, units (lb/ac) and field boundary in the export match; recommendation rules stored in the SMS project. Excel macros (VBA) are disabled, no Python or GIS is installed, and a file-system audit flags rx_pk/ files not written by Ag Leader SMS.

Proposed partial credit

20% soil data import; 35% recommendation rules; 30% prescription rates; 15% controller export.

R08.04.01.01-T09 Summarize an on-farm nitrogen strip trial from yield and as-applied data

Original instruction

input/as_applied_trial.gpkg holds applied-rate points for four N-rate treatments in three replicated strips, but no strip polygons. In QGIS digitize the 12 strip cores (excluding 5 m strip borders and 20 m transitions) as polygons with treatment and rep attributes in output/strips.gpkg, clip the cleaned yield points to them, and write output/trial_summary.csv plus the best rate in output/answer.json.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: G2 · GUI-essential
GUI need · ordinary access: HY · Hybrid
Interaction primitives: path · Continuous path click · Point & click vis · Visual judgment
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 3 / 5

Proposed verifier

Each strip-core polygon IoU ≥0.8 with the gold core and correct treatment/rep attributes; per treatment×replicate means within 0.05 t/ha of reference; best rate equal to gold. No strip geometry exists in the inputs and the PyQGIS console is disabled.

Proposed partial credit

40% strip cores digitized (scaled by IoU); 15% attributes; 30% means; 15% best rate.

R08.04.01.01-T10 Lay out a printable yield map with legend, scale bar and north arrow in QGIS

Original instruction

In QGIS style input/yield.tif with 7 quantile classes on a red–yellow–green ramp, then build an A4 landscape print layout with the map, legend, metric scale bar, north arrow and title 'North 80 — 2025 Corn Yield (t/ha)'. Save output/yield.qgz and export output/yield_map.pdf.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: G2 · GUI-essential
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click drag · Drag & drop type · Text & forms
Work product: WP03 · Visual layout
Conditions: None declared
Proposed difficulty: 2 / 5

Proposed verifier

Project parsed: renderer quantile with 7 classes and RdYlGn ramp; layout A4 landscape with map, legend, scale bar (meters), north arrow and exact title text, items not overlapping. PDF page size A4 and title text extractable. The PyQGIS console and qgis_process are disabled, and a file-system audit flags a .qgz or PDF not written by QGIS.

Proposed partial credit

30% styling; 40% layout items; 20% title; 10% PDF.

R08.04.01.01-T11 Estimate and correct combine grain-flow delay from yield transitions

Original instruction

input/yield_raw.gpkg shows yield shifted at pass ends because the monitor's flow delay is wrong. Inspect transitions at headland entries/exits in QGIS, estimate the true delay in seconds, shift yield values along each pass accordingly (scripting allowed), and save output/yield_shifted.gpkg and the delay in output/answer.json.

Platforms: mac · macOS win · Windows
Applications: QGIS · Python
Interfaces: desk · Desktop GUI term · Terminal
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: HY · Hybrid
Interaction primitives: vis · Visual judgment nav · Pan/zoom/scroll code · Code & commands
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 5 / 5

Proposed verifier

Delay within ±1 s of the seeded value; shifted values correlate r ≥0.98 with the reference correction and pass-end artifacts are removed (edge-yield ratio within 5% of field mean).

Proposed partial credit

40% delay estimate; 40% shifted data; 20% artifacts removed.

R08.04.01.01-T12 Map persistently unprofitable areas from four years of yield and set them aside

Original instruction

input/ has 10 m yield rasters for 2022–2025 (corn/soy rotation), prices.csv and costs.csv. Flag cells with negative net return in at least 3 of 4 years, polygonize, then edit the polygons in QGIS into farmable set-aside areas: straight edges parallel to the pass direction in input/ab_line.gpkg, at least 24 m wide, fragments <0.2 ha merged or dropped. Save output/set_aside.gpkg and the total area in output/answer.json.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: G2 · GUI-essential
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click path · Continuous path vis · Visual judgment
Work product: WP04 · 3D/spatial model
Conditions: long · Long-horizon
Proposed difficulty: 5 / 5

Proposed verifier

Edited polygons must cover ≥85% of the gold persistent-loss area with ≤25% extra; every polygon ≥24 m wide (negative-buffer test) with edges within 5° of the pass direction or its normal; none <0.2 ha; area within 10% of the reported value. The PyQGIS console is disabled.

Proposed partial credit

25% per-year net return and persistence rule; 30% coverage of loss areas; 30% farmable geometry constraints; 15% area.

R08.04.01.01-T13 Fix a misassigned crop on harvest records in Operations Center (sandbox)

Original instruction

In the Operations Center sandbox, field 'Creek 60' has 2025 harvest data recorded as corn, but the planting record and yields of 3–4 t/ha indicate soybeans. Correct the harvest operation's crop type and variety to match the planting record, and leave other fields unchanged.

Platforms: web · Any OS (web)
Applications: John Deere Operations Center
Interfaces: brow · Browser
GUI need · task setting: G2 · GUI-essential
GUI need · ordinary access: G2 · GUI-essential
Interaction primitives: click · Point & click nav · Pan/zoom/scroll vis · Visual judgment
Work product: WP08 · App/system state
Conditions: None declared
Proposed difficulty: 3 / 5

Proposed verifier

Sandbox state: Creek 60 harvest operation crop = soybeans and variety equal to planting record; harvest data for other fields unmodified (change log). No API credentials are available to the agent.

Proposed partial credit

40% crop type corrected; 30% variety corrected; 30% no collateral changes.

R08.04.01.01-T14 Measure application accuracy of an as-applied map against its prescription

Original instruction

In QGIS compare input/as_applied.gpkg (applied N rate points with swath width) to input/rx_n.shp. Compute the percentage of applied area within ±10% of the target rate and list zones where mean applied rate was below 90% of target. Write output/accuracy.csv and output/answer.json.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click type · Text & forms
Work product: WP02 · Records/table
Conditions: None declared
Proposed difficulty: 3 / 5

Proposed verifier

Area-weighted percentage within 0.5 points of reference; under-applied zone list exact.

Proposed partial credit

30% join of points to zones; 40% accuracy percentage; 30% under-applied list.

R08.04.01.01-T15 Build a prescription grid aligned to an AB guidance line with headland exclusion

Original instruction

In QGIS create a grid of 12 m × 30 m cells aligned to the AB line in input/ab_line.gpkg (heading 37°), with one swath edge on the AB line, clipped to input/boundary.shp minus a 24 m headland. Assign each cell the majority zone rate from input/rx_n.shp and export output/rx_grid.shp.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: geom · Precise geometry click · Point & click vis · Visual judgment
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 4 / 5

Proposed verifier

Cell long axis within 1° of the AB heading; cell width 12±0.1 m with a swath edge within 0.2 m of the AB line; no cell inside the 24 m headland; majority rate correct for ≥95% of cells. PyQGIS is available (Hybrid route accepted); grading is purely geometric.

Proposed partial credit

35% grid orientation and alignment; 20% cell size; 20% headland clip; 25% rates.

R08.04.01.01-T16 Export a nitrogen prescription as ISOXML TASKDATA from FarmWorks

Original instruction

In FarmWorks import input/rx_n.shp for field 'North 80', convert the lb N/ac rates to L/ha of UAN 32% (density 1.32 kg/L), set product 'UAN 32%', and export an ISOBUS TASKDATA prescription (grid type 2) to output/TASKDATA/.

Platforms: win · Windows
Applications: FarmWorks
Interfaces: desk · Desktop GUI
GUI need · task setting: G3 · GUI-only
GUI need · ordinary access: G3 · GUI-only
Interaction primitives: click · Point & click type · Text & forms
Work product: WP04 · 3D/spatial model
Conditions: licensed · Licensed pro app
Proposed difficulty: 4 / 5

Proposed verifier

TASKDATA.XML parses as ISO 11783-10: one TSK with PDT 'UAN 32%', GRD type 2 with a binary grid file; setpoint values decode to the converted L/ha within 1% per cell; field boundary present. No Python or ISOXML library is installed, and a file-system audit flags TASKDATA files not written by the FarmWorks exporter.

Proposed partial credit

25% import; 30% unit conversion; 30% valid ISOXML grid; 15% product metadata.

R08.04.01.01-T17 Control: clean yield points and interpolate with GeoPandas and GDAL

Original instruction

Control pair of T01. Using Python (GeoPandas) and GDAL from the terminal, apply the same filters to input/yield_2025.shp, save output/yield_clean.gpkg, and IDW-interpolate (power 2, radius 30 m) to a 10 m output/yield.tif.

Platforms: mac · macOS lin · Linux
Applications: GeoPandas · GDAL
Interfaces: term · Terminal
GUI need · task setting: CL · CLI-sufficient
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: code · Code & commands
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 2 / 5

Proposed verifier

Same checker as T01 (Jaccard ≥0.98 on kept IDs, raster RMSE <0.1 t/ha). Scored as the CLI control of T01.

Proposed partial credit

40% filters; 20% headland; 40% raster.

R08.04.01.01-T18 Remove double-harvested overlapping passes from a yield dataset

Original instruction

In input/yield_2025.gpkg some passes re-enter already harvested ground, producing low yields. In QGIS identify points whose swath overlaps a previously harvested swath by more than 50% (by time order), remove them, and save output/yield_nooverlap.gpkg and the removed IDs in output/removed.csv.

Platforms: mac · macOS win · Windows
Applications: QGIS
Interfaces: desk · Desktop GUI
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click vis · Visual judgment nav · Pan/zoom/scroll
Work product: WP04 · 3D/spatial model
Conditions: None declared
Proposed difficulty: 4 / 5

Proposed verifier

Removed ID set Jaccard ≥0.9 vs gold computed from swath polygons in time order; remaining dataset has no >50% overlaps.

Proposed partial credit

60% removed set (scaled by Jaccard); 40% clean output.

R08.04.01.01-T19 Choose the zone count with fuzzy k-means on soil and yield layers and build a seeding Rx

Original instruction

input/ has ec_shallow.tif, ec_deep.tif, elevation.tif and yield_norm_mean.tif. Standardize and run PCA, then fuzzy k-means for 2–6 zones; choose the zone count minimizing FPI and NCE, build zones, assign seeding rates from input/seed_rules.csv, and export output/zones.gpkg and output/rx_seed.shp with a short justification.

Platforms: mac · macOS win · Windows
Applications: QGIS · Python
Interfaces: desk · Desktop GUI term · Terminal
GUI need · task setting: HY · Hybrid
GUI need · ordinary access: CL · CLI-sufficient
Interaction primitives: click · Point & click code · Code & commands vis · Visual judgment
Work product: WP04 · 3D/spatial model
Conditions: long · Long-horizon
Proposed difficulty: 5 / 5

Proposed verifier

Chosen zone count equals gold; zone map adjusted Rand index ≥0.7 vs reference; seeding rates per zone exact from the rules; prescription export valid.

Proposed partial credit

20% PCA inputs; 25% zone count; 30% zone map agreement; 25% seeding Rx.

R08.04.01.01-T20 Compare hybrid performance across fields with Operations Center Analyze (sandbox)

Original instruction

In the Operations Center sandbox use Analyze to compare 2025 corn yield by variety across fields North 80, East 120 and Creek 60, export the variety report as CSV to output/variety.csv, and write the variety with the highest area-weighted mean yield to output/answer.json.

Platforms: web · Any OS (web)
Applications: John Deere Operations Center
Interfaces: brow · Browser
GUI need · task setting: G2 · GUI-essential
GUI need · ordinary access: G2 · GUI-essential
Interaction primitives: click · Point & click nav · Pan/zoom/scroll
Work product: WP12 · Answer/decision
Conditions: None declared
Proposed difficulty: 3 / 5

Proposed verifier

variety.csv must be the Operations Center export for the three fields and 2025 corn; best variety equals gold computed from sandbox data. Exports from other seasons or crops fail. The agent has no API credentials for Operations Center; the CSV must be the platform's browser export.

Proposed partial credit

30% correct filters; 40% export; 30% best variety.

Mapped research references

No benchmark reference is assigned to this class in the current mapping.