Decision-Making Context and Needs

Cotton production is a labour-intensive, climate-sensitive livelihood activity. In Bangladesh, cotton communities are increasingly exposed to higher temperatures and humidity, more frequent heatwaves, and disruptive extremes such as droughts and flash floods. Stakeholder consultations highlighted that extreme heat and humidity are already constraining safe time in the field, reducing daily productive hours and increasing the incidence of dehydration, cramps and heat-related illness.

Adaptation decisions the project was designed to support include:

• Work planning: when to work, when to rest, and what health precautions to take (hydration, shade, heat illness signs).
• Timing of key agronomic activities (planting, spraying, picking and harvesting) under heat and precipitation constraints.
• Targeting of extension support and health outreach by CottonConnect field teams, NGOs and local health services.
• Government and sector planning: understanding where cotton is grown, how conditions are changing, and what support measures are required.

The stakeholder consultation process highlighted the trend of lost labour hours due to adverse environmental conditions as an issue of local concern. The project analysed trends for the study region in Wet Bulb Globe Temperature (WBGT), a heat metric that considers air temperature, relative humidity, wind speed and direct solar radiation - factors that determine how much heat a person experiences and their ability to cool down through sweating. It is a metric that has been found to be a stronger predictor of negative heat-related health outcomes than air temperature or the heat index (a combination of air temperature and relative humidity). It is also used in national legislation and workplace health and safety. The analysis showed a clear upward trend over the past 3 decades in the amount of time workers would be recommended not to work to stay within safe physiological boundaries.

Figure 2: Total number of working days in which it would have been recommended not to work to remain safe in the heat conditions in Chuadanga district in Bangladesh.

Project end users and stakeholders span the cotton supply chain: farmers and pickers (women and men), ginners and spinners, seed companies, NGOs working with farmer groups, and government agencies including the Cotton Development Board and the Ministry of Agriculture. CottonConnect plays a central intermediary role, translating analytics into locally appropriate advice and disseminating it through established farmer engagement channels.

Earth Observation Data Products and Information

C5 produces and uses EO-derived products that directly address the decision needs described above. Key products include:

• Cotton distribution map (2023): National 10 metre land cover classification (using Copernicus Sentinel-1 and -2 data) identifying cotton, mixed cotton and other classes; used for targeting support and understanding production geography.
• Vegetation health time series and anomalies: Leaf Area Index climatology and standardised anomalies (using MODIS sensor data) and associated indicators used as proxies for crop condition and stress.
• Heat stress indices (historic archive): Hourly Wet Bulb Globe Temperature (WBGT) and Heat Index from reanalysis; translated to work-rest categories and a Lost Labour Index to quantify productivity constraints.
• District forecast heat risk summaries: Daily updated 15-day forecast time series (air temperature and heat index) aggregated over pilot districts; used to generate bulletin plots and risk tables.
• High-resolution temperature feasibility product: Prototype 2 metre height air temperature downscaling using geostationary satellite and Copernicus Land Surface Temperature data combined with reanalysis data in a machine learning model.

All project datasets and operational bulletin inputs were published to an open repository to support transparency and reuse by partners.

Underlying Observations and Data Sources

C5 integrates multiple EO and climate data sources. The main observations and datasets used include:

• Sentinel-2 optical surface reflectance (10–20 metres) for vegetation and crop mapping features.
• Sentinel-1 SAR backscatter (VV/VH; 10 metres) to capture crop phenology and improve cotton extent mapping in cloudy periods.
• MODIS vegetation biophysical parameters, in particular Leaf Area Index (LAI) at 500 metres every 5 days to construct a multi-year baseline and anomalies.
• ERA5 and ERA5-Land reanalysis (ECMWF/C3S) providing temperature, humidity, wind and related variables used to compute WBGT, Heat Index and derived work-rest guidance from 1980 onward.
• NOAA Global Forecast System open forecasts (2 metre height temperature and humidity) used to generate daily district-level forecast plots and heat-risk tables (up to 15 days lead).
• Bangladesh Meteorological Department (BMD) forecast information (used to define and check national heatwave thresholds where available).
• Himawari-9 AHI geostationary observations (albedo, brightness temperatures and geometry) and Copernicus Land Monitoring Service hourly Land Surface Temperature (LST) for high-resolution temperature downscaling feasibility.
• In-situ reference and training data: CottonConnect field samples for crop mapping; meteorological station data (used in model training/validation for downscaling feasibility).
• ESA CCI Land Cover used in reference dataset development for cotton mapping validation.

Figure 5: Cotton extent map

Figure 6: Farmer engagement & feedback session in Chuadanga

Impacts and Benefits

C5 was designed to produce measurable value to end users by converting EO and weather data into decisions that protect health and maintain productivity. Benefits observed and evidenced during the feasibility study include:

• Improved situational awareness and anticipatory guidance: farmers and intermediaries received forecast-based information and tailored precautions that were not previously available in a consolidated form.
• Health protection and productivity planning: the Health Stress Index operationalises work-rest guidance and a Lost Labour Index, enabling conversations about safe labour planning under rising heat exposure.
• Better targeting of support: the national cotton map and crop context analytics help CottonConnect and government stakeholders understand where cotton is grown and where interventions should be prioritised.
• Supply-chain resilience and ESG reporting potential: data products provide an evidence base for brands and supply-chain actors seeking to understand climate risk, labour impacts and adaptation measures in sourcing regions.

Evidence and performance results reported during the project include:

• Cotton distribution mapping achieved an overall classification accuracy of 0.9658 against reference data.
• The EO-based temperature model performed strongly in validation, closely matching ground-station measurements (RMSE 1.18, MAE 0.9; correlation 0.99).
• The feasibility study pilot covered 300 farmers across three districts (Kushtia, Meherpur and Chuadanga), with positive feedback reported from direct engagement and a final stakeholder workshop in Dhaka.
• An operational pipeline was demonstrated: forecast data downloaded daily, processed into plots and a risk table, and published to the repository for CottonConnect staff to support bulletin preparation.

Key delivery lessons included the need for hybrid dissemination (digital plus in-person) due to uneven smartphone/WhatsApp usage and literacy barriers, and a recurring user requirement for mid-term and seasonal forecasts (temperature and precipitation) to support planning across the full cotton production cycle.

Further information and guidance

Open data repository (C5 outputs): bit.ly/c5data

Core EO / climate datasets: Sentinel-1 and Sentinel-2 (Copernicus), MODIS LAI, ERA5/ERA5-Land (C3S), NOAA GFS forecasts, Himawari-9 AHI, Copernicus LST.

Suggested points of contact:

• CottonConnect – programme delivery, farmer engagement and bulletin dissemination (Bangladesh pilot). sunaina.chaturvedi@cottonconnect.org
• Assimila – EO and climate data processing, index development and operational pipeline design. gerardo.lopezsaldana@assimila.eu

References (project documentation):

• C5 Final Report (May 2024), ESA C5 project 2023-115 (Assimila and CottonConnect).
• C5 Marketing Overview (draft, 2024) – positioning and farmer quote used in Section 1.