Dedicated methodology page
How the Nyx cotton PoC turns public environmental data into interpretable signals
This page makes the PoC logic explicit. It separates field-cluster evidence from regional context, lists the indicator formulas and weights, and shows where Nyx is measuring directly versus estimating a composite summary. The goal is transparency, not mystique.
Claim boundary
The representative field-cluster output below is a PoC worked example built from the established Nyx formulas and scoring rules. It is designed to show how the pipeline will present results once the automated data refresh is running.
Core composite equations
LCPI = 0.30×RP + 0.30×SMP + 0.25×HP + 0.15×WCP
FDR = 0.30×RainfallPersistence + 0.30×HeatPersistence + 0.25×ChronicWaterStress + 0.15×FloodFactor
Pilot scope
Three framing rules keep the PoC credible
Pilot entity
Representative cotton field cluster in Bahawalpur district
Deliberate PoC assumption rather than a verified contracted farm.
Spatial model
Field-cluster evidence + district and basin context
Fine-resolution and context-resolution signals are kept distinct.
Worked season
2023 cotton season within a 2021-2024 monitoring window
May-October is the reference crop window used throughout the PoC.
Indicator formulas
Every visible signal should be traceable back to one auditable transformation
The PoC uses a small number of indicator families and keeps the transformations explicit. Fine-resolution NDVI is calculated over the representative field cluster, while rainfall, soil-moisture, heat, flood, and structural water-stress layers provide wider context around the same origin story.
NDVI crop condition
Formula
NDVI = (B8 - B4) / (B8 + B4)
Aggregation: Scene-level polygon means aggregated across May-October.
Visible output: Seasonal mean NDVI and seasonal peak NDVI.
Sentinel-2 Level-2A surface reflectance with cloud filtering.
Rainfall anomaly
Formula
Rainfall anomaly ratio = (Rcurrent - Rbaseline) / Rbaseline
Aggregation: CHIRPS daily or monthly totals summed over May-October.
Visible output: Below-normal to above-normal seasonal rainfall signal.
Longer historical baselines are feasible because CHIRPS extends back to 1981.
Soil-moisture anomaly
Formula
SM anomaly ratio = (SMcurrent - SMbaseline) / SMbaseline
Aggregation: District-context mean for seasonal soil-moisture context.
Visible output: Dryness or wetness relative to usual conditions.
SMAP is contextual and should not be presented as field-precise.
Heat anomaly
Formula
Heat anomaly = LSTcurrent - LSTbaseline
Aggregation: Seasonal mean MODIS LST or ERA5-Land cross-check context.
Visible output: Near typical, moderate, significant, or extreme heat pressure.
MOD11A2 8-day LST is the preferred PoC heat layer.
Flood exposure index
Formula
FEI = 100 × (0.25×A1 + 0.50×A2 + 0.75×A3 + 1.00×A4)
Aggregation: Area share by JRC flood-depth class using the 1-in-50-year layer.
Visible output: Low, moderate, or high river-flood exposure context.
Static hazard context rather than evidence of a realized flood event.
Water-scarcity context
Formula
Aqueduct category mapped to numeric context score
Aggregation: Basin or regional classification rather than farm-performance logic.
Visible output: Low to extremely high structural water stress.
WRI Aqueduct is the main simple screening source for chronic water pressure.
Composite score weights
LCPI and FDR are simple on purpose
Both composite indices are designed to be explainable. LCPI summarizes recent pressure. FDR summarizes directional structural risk. Neither score should be framed as a regulatory metric or a complete product footprint.
| Component | Normalization | Weight |
|---|---|---|
| Rainfall pressure (RP) | min(max(((-1 × rainfall anomaly ratio) / 0.40), 0), 1) × 100 | 30% |
| Soil-moisture pressure (SMP) | min(max(((-1 × soil-moisture anomaly ratio) / 0.30), 0), 1) × 100 | 30% |
| Heat pressure (HP) | min(max((heat anomaly °C / 4.0), 0), 1) × 100 | 25% |
| Water-scarcity context pressure (WCP) | Mapped Aqueduct context score | 15% |
| Component | Method | Weight |
|---|---|---|
| Repeated rainfall stress | Share of recent seasons with rainfall anomaly ≤ -0.10 | 30% |
| Repeated heat stress | Share of recent seasons with heat anomaly ≥ 1.0°C | 30% |
| Chronic water stress | Mapped Aqueduct structural risk score | 25% |
| Flood factor | min(FEI, 100) | 15% |
Source notes
The PoC works only if scale and source limitations remain visible
Nyx should surface dataset roles and caveats rather than burying them. That is especially important when fine-resolution imagery is presented alongside coarser environmental context layers.
| Dataset | Role | Scale | Source note |
|---|---|---|---|
| Sentinel-2 L2A | Field-cluster vegetation and canopy condition | Fine resolution | Best suited for NDVI and seasonal crop-condition logic. |
| CHIRPS | Rainfall totals and rainfall anomalies | Medium resolution | Strong baseline depth for same-season anomaly comparisons. |
| SMAP | District-scale soil-moisture context | Coarse resolution | Useful context layer, but too coarse for farm-specific claims. |
| MODIS MOD11A2 / ERA5-Land | Seasonal heat-stress context | Medium to coarse | Used for anomaly logic and cross-checking of warmer-than-usual seasons. |
| JRC Global River Flood Hazard | Static river-flood exposure screening | Hazard-layer context | Represents screening exposure, not a realized seasonal event. |
| WRI Aqueduct | Regional structural water-stress context | Basin / regional | Context classification only; do not collapse it into a field metric without explanation. |
Supporting documents
Download the full Nyx PoC document set or return to the overview
The downloadable documents mirror the website narrative and give stakeholders a portable reference for executive framing, methodology, scope, and claim boundaries.