The India Health and Climate Resilience Fund asked us to build a district learning system across four sites: Chamarajanagar in Karnataka, Dhubri in Assam, and Khunti and West Singhbhum in Jharkhand. The deliverables included district base papers, structured profiles that would anchor the Fund's fellowship programme and act as reference documents for grantees working in those districts.
Writing them took longer than we expected. The data existed, but in formats that could not be read against each other. Each district has a health profile and each has climate data. The two are held by different arms of government, organised around different administrative geographies and updated on different schedules. At district level in India, the basic problem in climate-health research is integrating data that already exists in separate systems.
Dhubri, in lower Assam, is a clear case. Flood risk data from the Brahmaputra Board covers embankment failure zones and inundation histories, but only by river segment, with no breakdown by block or panchayat. NFHS-5 gives us block-level data on anaemia, child nutrition and healthcare utilisation for Assam, but those geographies do not map cleanly onto the flood zones. District hospital data in the Health Management Information System records admissions by diagnosis and month. The HMIS categories predate any climate-health analysis, so "diarrhoeal disease" in July is not flagged as flood-related, even when the correlation is obvious to anyone who has looked at both datasets together.
Chamarajanagar had a different version of the same problem. It is one of Karnataka's most climate-stressed districts, with high heat exposure and a tribal population that has limited access to the health system, and the Biligiriranga Hills create microclimatic variation that district averages erase. The Karnataka Health Department's HMIS data showed the district's health system performing adequately on process indicators such as facility availability, staffing ratios and immunisation coverage. NFHS-5 community health data showed more on outcomes. Stunting, institutional delivery and treatment-seeking for fever all varied markedly within the district, and the variation correlated with forest dependency and distance to town.
Khunti and West Singhbhum are in Jharkhand's mining belt, where several climate-health problems overlap. There is heat stress from deforestation and water contamination downstream of mining operations. Tuberculosis rates are among the highest in the country, in a state with a disproportionate share of India's adivasi population. Mining-related environmental exposure is tracked by the Ministry of Mines and the Jharkhand State Pollution Control Board. TB incidence and treatment outcomes are tracked by the National TB Elimination Programme. Nothing in the administrative system cross-references the two.
The district base papers did not solve this integration problem. They forced us to attempt the integration, record where it failed, and state plainly what we were inferring and what the data supported. Writing down what we could not say, and why, turned out to be one of the most useful outputs for Fund grantees. They were designing programmes in these districts and needed to know the evidentiary floor before claiming outcomes.
We also found that the districts where the link between climate and health was most obvious on the ground were often those where the administrative data hid it most. Under heavy stress, people adapt their health-seeking, food strategies and mobility, and standard indicators are not designed to detect those adaptations. The base paper method combined secondary data synthesis with field visits and key informant interviews, and it picked up those adaptations better than any single data source would have.
The Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) and ERA5 reanalysis data were used for climate layers. Health system data came from HMIS and NFHS-5 district factsheets. Nutrition data was cross-referenced with Poshan Tracker where available.