Your farm workspace
AI-generated conceptual crop illustrationAI-generated illustration · not field evidence
Your season, in focus

Farm overview

From the first seed to a better-informed harvest.

Live workspace QA agent

An automated local agent checks data integrity, monitoring gaps and page controls. Watching runs while this tab is open; it is not a human chat service.

RAG solution centre

Local keyword retrieval + extractive answer generation from cited passages. No cloud LLM connected. Uploaded documents are untrusted reference material, never agent instructions.

Knowledge library & document upload

Upload plain-text or Markdown extension guidance, max 500 KB. Documents stay in this tab; no automatic upload to any provider. For PDF, export its text first.

ML harvest predictor · multivariable k-nearest neighbours

Train on your own historical data for one crop and growth stage. Features: NDVI and NDMI; target: yield kg/ha. Holdout validation leaves each district out in turn. Model and data stay in this tab. This predicts yield, not disease or pesticide residues.

crop,stage,district,season,ndvi,ndmi,yield_kg_ha,source

No model trained.
Compare model error with farm decisions. Predictions require comparable timing, crop and measurement protocol. There is no pretrained Pakistan-wide model.

Concept illustrations generated with AI. Actual satellite results are labelled separately.

Fields requiring a visit

Monitoring coverage

Next actions

Start with your own farm

Register a field, add a dated scouting visit, then record follow-up visits using the same sampling method.

Example records are labelled and kept separate from an empty workspace. Use Live satellite check for opt-in scene lookup and polygon-based index calculation. No automated disease model is connected.

Field register

Field location map

Map requires the online Leaflet library. Field coordinates remain available in the table.
Markers are farmer-entered locations. Imported boundaries are WGS84 GeoJSON, not cadastral verification.

Season & field profile

Stage estimates use elapsed days and a configurable harvest date. They are planning aids; inspect the actual crop before taking action.

Record a scouting visit

Symptom assessment

Select a symptom to see possible causes and practical checks. This is a rule-based field guide, not a diagnosis.

Confirm pest identity and locally applicable economic thresholds with extension staff before choosing a pesticide. A stressed plant does not automatically need a spray.

Scouting history

Measured-data workspace: import your field-level Sentinel-2 or other processed index values. Live satellite check can populate this ledger after an opt-in, quality-screened field analysis. Vegetation indices indicate canopy conditions; they cannot identify a specific disease or pesticide residue.

Vegetation & moisture history

Add a geospatial measurement

Measurement ledger

Reproducible GIS workflow

  1. Use a reviewed field boundary in WGS84; reproject for area calculations.
  2. Use atmospherically corrected Sentinel-2 L2A imagery. Mask cloud, cirrus, shadow, snow and invalid pixels using SCL.
  3. NDVI = (B08 − B04) / (B08 + B04).
  4. NDMI = (B08 − B11) / (B08 + B11); align 20 m and 10 m grids explicitly.
  5. Exclude boundary mixed pixels; report the median, cloud and valid coverage for each date.
  6. Compare the same field and crop stage, then verify anomalies on the ground.

Small fields and narrow plots may have too few uncontaminated pixels for dependable monitoring.

Quality rules in this portal

Cloud cover above 20% or valid coverage below 70% flags a measurement for review. These configurable-workflow defaults are screening rules, not universal scientific thresholds.

Charts retain flagged values with a different colour, but comparisons exclude them. No crop health score is inferred from a single NDVI threshold.

Sentinel-2 product documentation ↗

Pesticide application record

Review effects, not assumptions

Before/after changes do not prove a treatment caused recovery. Weather, irrigation, crop stage, sampling differences and untreated controls matter. Residue safety requires laboratory testing.

Integrated pest management

Scout first. Identify the pest, compare with a local action threshold, protect beneficial organisms and consider cultural or biological controls. Rotate modes of action only according to current label and extension advice.

Application & harvest interval ledger

Dates calculate elapsed label intervals only. They do not certify legal compliance, residue safety or readiness to harvest. Product-specific label rules take precedence.

Pakistan seasonal planning calendar

Sowing / transplantingGrowing seasonHarvest / picking

Illustrative lowland planning ranges, not planting instructions. Variety, elevation, irrigation and local advisories change dates. Spring and autumn maize are listed separately. Field-specific dates in your register override this calendar.

Your field milestones

کسان کی اردو رہنمائی

کسان کی رہنمائی

اپنے کھیت، فصل اور بوائی کی تاریخ درج کریں۔ فصل کا باقاعدہ معائنہ کریں اور متاثرہ پودوں کی تعداد محفوظ کریں۔ اسپرے سے پہلے کیڑوں کی شناخت اور مقامی زرعی ماہر سے تصدیق ضروری ہے۔ دوا کے لیبل کے مطابق حفاظتی لباس، دوبارہ داخلے کا وقفہ اور کٹائی سے پہلے کا وقفہ اختیار کریں۔

سیٹلائٹ اشاریے فصل کے دباؤ کا اشارہ دیتے ہیں، بیماری یا دوا کے اثر کی حتمی تصدیق نہیں کرتے۔ ریکارڈ اسی براؤزر میں محفوظ ہوتا ہے؛ بیک اپ ضرور ڈاؤن لوڈ کریں۔

Official knowledge & extension links

Punjab crop production guidance ↗ · Pest scouting and IPM brochures ↗ · Pest forecast bulletins ↗

Punjab crop life calendar (historical reference) ↗ · Copernicus Sentinel-2 documentation ↗

District directory is a selection aid, not an authoritative current boundary register. PBS administrative geography reference. Custom district names are supported.
Reviewed 21 October 2026. Punjab references do not establish recommendations for every Pakistani agroecological zone. Use your district agriculture extension office for local variety, threshold, pesticide and planting advice.

Pakistan agricultural intelligence

GIS · STATISTICS · POWER BI

Provincial proportional-symbol map

Symbols sit at approximate province centres, not field locations or administrative boundaries. Size represents the selected measure. Click a symbol for values.

Provincial comparison

Production share

Seasonal yield comparison

Area–yield relationship

Province-level association is not evidence of farm-level causality.

Provincial evidence table

District / regional farm statistics

Computed only from your registered fields. Not representative district or national survey estimates. Symptom counts pool the latest visit per field; sampling designs may differ.

Power BI handoff

Export a field-level fact table with province/district, crop, dates, area, observed symptoms, measured indices and actual/expected yields. Import the CSV in Power BI Desktop, set numeric types, then use the supplied DAX measures.

This is an export integration. No live Power BI report or .pbix file is embedded.
Observed harvest sampling

Crop-cut yield estimator

Enter representative quadrat samples, one per line: fresh harvested mass (kg), sampled area (m²). Use the same crop, harvest date and grain-moisture method.

Yield = total mass ÷ total sample area × 10,000 × (100 − measured moisture)/(100 − target moisture). This extrapolates samples; it is not a guaranteed field harvest.
Interpretable machine-learning baseline

Train & evaluate a yield model

Upload labelled historical data for one crop and one growth-stage protocol. CSV: crop,district,season,stage,ndvi,yield_kg_ha,source. At least 15 rows, 3 districts, and NDVI variation within each training fold.

No model trained. Uploaded training rows remain in memory for this tab session and are not included in farm backups.
Ordinary least-squares baseline with leave-one-district-out validation, not a production GeoAI forecast. No extrapolation outside training NDVI range. No calibrated uncertainty interval. A single vegetation index misses weather, soils, management, variety and phenology.

GeoAI model card & deployment gates

Before operational deployment: validate on independent seasons and spatial blocks, compare with climatology and agronomist baselines, check sensor/stage consistency, quantify uncertainty, and monitor drift. Disease classification requires expert-labelled images and a separately validated model; this portal does not claim one.

Crop stress & issue atlas

ILLUSTRATIONS + FIELD EVIDENCE

Illustrations explain symptom patterns; they are not photographs or AI diagnoses. Select a card for ground checks.

Observed symptom distribution

Latest observation per field. Missing visits are shown separately; zero observations do not mean healthy crops.

Geostatistical field exploration

Latest visit in the preceding 7 days per field, same crop. At least 5 distinct field coordinates are required for descriptive distance analysis.

Empirical semivariogram: γ(h) = ½ mean[(zᵢ − zⱼ)²], with z = symptomatic plants (%). Distance uses great-circle kilometres. Bins are descriptive, with no fitted variogram, significance test, kriging surface or causal claim.

Field evidence map

Markers and imported boundaries are coloured by observed symptom incidence: green <10%, amber 10–25%, red >25%; grey = no visit in the selected window. These are display bands, not economic action thresholds.

Schedule a farm activity

Activity calendar

Import the file in your calendar app. This portal does not create cloud reminders or send notifications. Due dates are all-day events.

Activity register

Field water-budget planner

Enter locally measured or sourced values for an upland-crop planning period. This is a simplified water-balance calculation, not sensor-based automatic irrigation.

Water requirement breakdown

Validate local ET₀, crop stage, root-zone storage and field efficiency. This model omits paddy land preparation, ponding, seepage, salinity/leaching and irrigation-event scheduling. It must not be used as a complete rice irrigation prescription.

ETc = ET₀ × Kc × days. Net deficit = max(0, ETc − effective rain − usable root-zone water). Gross depth = net deficit ÷ efficiency fraction. Volume = gross mm × field hectares × 10.

FAO crop coefficient reference ↗

Saved water plans

Mark a plan applied only after the field operation. Calculated volumes are estimates, not metered water use.

Record costs & sales

Recorded farm balance

Recorded income minus expenses is a cash-ledger balance. It does not establish full farm profit; missing transactions, inventory, depreciation, credit, taxes and unpaid family labour can change results.

Yield × price scenario

Enter your own price. No live market quotation or guaranteed yield is supplied.

Scenario & sensitivity

Sensitivity cells show assumed revenue minus the same entered costs. Yield and price vary by ±20%; this is a scenario grid, not a probability forecast.

Transaction ledger

Field action centre

Rule-based reminders calculated from your records and Pakistan dates. Checks run when the page renders or you refresh. No background surveillance, SMS or push notifications.

Connect a field to Sentinel-2

PUBLIC SATELLITE DATA · NO ACCOUNT PASSWORD

Use a reviewed field polygon for field-level analysis. A point and an area value can find nearby scenes but cannot define a verified field boundary.

No satellite lookup performed. Choose a field and approve the location-sharing checkbox.

Satellite scenes

Scene cloud metadata covers a whole tile. Actual clear coverage inside your field is checked separately. Searches show up to 12 newest matches; no continuous monitoring service is running.

Field verification & satellite findings

These checks assess data suitability and canopy measurements. They do not verify title, cadastral boundaries, crop identity or a disease. Confirm possible causes with scouting, soil/water tests and local extension.

Field NDVI raster

Low NDVI: ochreHigher NDVI: greenGrey: masked
20 m analysis grid. Red and NIR are averaged from aligned 10 m pixels; SWIR/SCL use the native 20 m grid. Pixels near the boundary are excluded. This is not a true-colour photograph.

Analysis provenance

Successful quality-screened measurements are added to Satellite & index records with scene ID, acquisition date, source and screening metadata. Failed or insufficient analyses are not saved as valid observations.

Farm monitoring report

Import vegetation measurements

CSV columns: field,date,ndvi,ndmi,cloud,coverage,source. Use an existing field ID, ISO dates and values within the documented ranges. Quoted CSV values are supported.

Import field boundary

Select a field, then import a GeoJSON Polygon or MultiPolygon feature in WGS84 longitude/latitude. Coordinates are checked; topology and cadastral accuracy require GIS review.

Backup & restore

Everything is stored in this browser. There is no account sync, shared database or unattended monitoring service. Back up regularly.

Restore replaces this workspace after confirmation.

Workspace

No photos are sent to AI and no automated diagnostic claim is made. A production national monitoring service would need authenticated shared storage, a reviewed agronomy knowledge base, satellite processing and validated disease models.

AgriSentinel Pakistan · Muhammad Safdar · Evidence-led agriculture

Register a field