Adaptation vs mitigation evaluation, vulnerability measurement under deep uncertainty, longitudinal cohort design for slow-onset change, and attribution in a changing baseline. Walk out with a climate adaptation evaluation design brief.
4 modules~3 hoursInteractiveIndia-context
Your progress
0% complete
Your Capstone
Climate Adaptation Evaluation Design Brief
Walk in with a climate adaptation programme. Walk out with an evaluation design brief covering evaluability framing, vulnerability measurement, longitudinal design, and attribution strategy.
Module 1 . ~25 min
Adaptation vs mitigation -- what can be evaluated
Mitigation evaluation is relatively straightforward: did CO2 emissions decrease? Adaptation evaluation is fundamentally harder because adaptation success means "bad things did not happen" -- it is measuring a counterfactual absence. A farmer who diversifies crops because of a climate-resilience programme and then survives a drought is a success. But without the drought, the adaptation looks like wasted effort. And you cannot schedule a drought for your endline.
What makes adaptation evaluation different
Moving baseline -- the climate is changing during your evaluation period. "Improvement" relative to a worsening baseline is still improvement, but standard pre-post designs miss this.
Long time horizons -- adaptation benefits manifest over decades. Most evaluations are 2-3 years. You are measuring early signals of a long-term process.
Context dependence -- what works for drought adaptation in Bundelkhand will not work for flood adaptation in Assam or coastal erosion in Sundarbans. Generalisability is limited.
Maladaptation risk -- an intervention can reduce vulnerability on one dimension while increasing it on another (e.g., borewells for drought resilience that deplete groundwater for the next decade).
Evaluable adaptation outcomes in India
Outcome type
Example
Timeframe
Measurability
Adoption of climate-smart practices
Crop diversification, water harvesting, drought-resistant varieties under PMKSY
1-2 years
High (survey-based)
Asset protection
Livestock loss reduction, crop loss reduction during extreme events
Event-dependent
Medium (requires event coinciding with evaluation)
Livelihood diversification
Number of income sources, non-farm income share
2-3 years
High
Institutional adaptive capacity
Panchayat climate action plan, early warning system functionality
NABARD's Watershed Development Fund programmes in Maharashtra have been evaluated for climate adaptation value. The evaluation measured: (a) change in cropping pattern (shift to drought-resistant varieties), (b) groundwater recharge (well water levels pre-post monsoon), (c) income stability (coefficient of variation in income across years), and (d) coping strategy index during the 2015 drought. The comparison was between watershed-treated and untreated villages in the same taluka.
Your Evaluability Framing
Fill these for your climate adaptation programme. Answers flow into the capstone.
e.g., "Climate-resilient agriculture programme, 5 blocks in Bundelkhand, UP, NABARD-funded"
Saved
Self-check
A climate programme distributes solar pumps to replace diesel pumps in Maharashtra. Is this adaptation, mitigation, or both?
Adaptation -- reduces climate vulnerability
Mitigation -- reduces emissions from diesel
Both, but the evaluation must measure each pathway separately -- emission reduction (mitigation) and whether solar pump reliability during grid outages improves irrigation during dry spells (adaptation)
Neither -- it is an energy programme
Correct. Solar pumps have co-benefits. But an evaluation claiming "climate impact" must specify which pathway it is measuring. The mitigation pathway (emissions) and adaptation pathway (irrigation reliability) require different indicators and different counterfactuals.
Module 2 . ~30 min
Vulnerability measurement under deep uncertainty
Climate vulnerability is a composite concept: Vulnerability = f(Exposure, Sensitivity, Adaptive Capacity). Each component is itself multi-dimensional. Measuring it as a single index is tempting but methodologically fraught.
The IPCC framework in practice
Exposure -- the degree to which the community is physically exposed to climate hazards. Data: IMD historical climate data, IITM climate projections, district-level disaster frequency from NDMA.
Sensitivity -- the degree to which the community is affected by exposure. Depends on livelihood structure, health status, infrastructure quality. Data: Census, SECC, NREGA rolls, district handbooks.
Adaptive capacity -- the ability to adjust, moderate, or take advantage of change. Depends on assets, social capital, institutional support, information access. Data: primary survey, SHG records, panchayat records.
The index trap
Composite vulnerability indices (LVI, CVI, various SAPCC-based indices) are popular with funders because they produce a single number. But the weighting of sub-indicators is arbitrary, and small changes in weighting can reverse the conclusion. Better approach: report sub-indicators separately and show the direction of change in each, then construct the index as a summary only.
Deep uncertainty vs risk
Climate evaluation faces "deep uncertainty" -- we do not know the probability distributions of future climate events. A programme designed for 1.5 degrees C warming may fail at 2.5 degrees C. Your evaluation should explicitly state what climate scenario it is assessing against. Use IITM or CMIP6 downscaled projections for your district and specify the RCP/SSP pathway.
Your Vulnerability Measurement Plan
Design vulnerability measurement. These flow into your capstone.
e.g., "SSP2-4.5 downscaled for Bundelkhand, IITM projections to 2040"
Saved
Self-check
A vulnerability index shows your programme area's score improved from 0.62 to 0.54 (lower = less vulnerable). Can you conclude the programme reduced vulnerability?
Yes -- the index decreased, showing improvement
Not necessarily -- the index change could be driven by one sub-indicator, the weighting is arbitrary, and without a comparison area you cannot attribute the change to the programme
Only if the change is statistically significant
Yes, if you use the IPCC framework
Correct. Composite indices are summaries, not evidence. Report sub-indicators separately, compare to a control area, and test sensitivity to alternative weighting schemes before drawing conclusions.
Module 3 . ~30 min
Longitudinal cohort design for slow-onset change
Climate adaptation effects are slow-onset. Groundwater recharge from watershed treatment takes 3-5 monsoons. Agroforestry returns take 5-7 years. Soil organic carbon improvement takes a decade. Evaluating these with a 2-year pre-post design is like judging a tree by its first-year height.
Cohort design principles
Panel data -- follow the same households over time, not different cross-sections. This lets you measure within-household change. Panel attrition in rural India averages 8-15% per year; budget for tracking.
Climate event windows -- your evaluation period must include at least one significant climate event (drought, flood, cyclone) to test whether adaptation works under stress. This is not schedulable. Plan for opportunistic data collection when events occur.
Slow-onset indicators -- track continuously, not just at baseline and endline. Soil moisture sensors, well monitoring, satellite-derived vegetation indices (NDVI) can provide continuous data between survey rounds.
Comparison group selection -- in watershed programmes, downstream villages are contaminated (they receive water from treated watersheds). Select comparison villages from adjacent, untreated micro-watersheds with similar agro-climatic conditions.
Remote sensing as complementary data
For climate adaptation evaluation in India, freely available satellite data can supplement primary data collection:
Data source
Indicator
Resolution
Access
Sentinel-2 / Landsat
NDVI (vegetation health), crop area, land use change
10-30m, every 5-16 days
Free (Copernicus, USGS)
GRACE-FO
Groundwater storage anomaly
~300km (district level)
Free (NASA)
IMD gridded data
Rainfall, temperature at district level
0.25 degree grid
Free (IMD)
ISRO Bhuvan
Soil moisture, drought indices for India
Variable
Free
Worked example
Foundation for Ecological Security's commons-restoration programme in Rajasthan was evaluated using a 5-year panel of 600 households across treated and comparison villages. NDVI analysis from Sentinel-2 confirmed that vegetation cover in treated commons increased 23% over 4 monsoons versus 4% in comparison villages. Household surveys showed that fodder availability (from commons) during the 2023 drought reduced distress livestock sales by 35% in treated villages. The satellite data and survey data told the same story from different angles.
Your Longitudinal Design
Design the longitudinal component. These flow into your capstone.
Opportunistic data collection? Rapid-response survey module?
Saved
Self-check
Your 2-year evaluation of a watershed programme finds no significant difference in crop yields between treated and comparison villages. Does this mean the programme failed?
Yes -- no yield difference means no impact
Not necessarily -- watershed effects on yields may take 3-5 monsoons to manifest; check intermediate indicators (groundwater levels, soil moisture, practice adoption) which may already show change
Yes, if the sample size was adequate
Only if neither year had a drought
Correct. Watershed programmes are slow-onset interventions. Two years may be too early for yield effects. The evaluation should report on the causal chain: Did practices change? Did groundwater improve? Did soil moisture increase? If yes, yield effects are plausible in later years. Early absence of yield effects is not programme failure -- it is timeline mismatch.
Module 4 . ~25 min
Attribution in a changing baseline
The fundamental attribution challenge in climate adaptation is: the baseline is changing. Rainfall patterns shift. Temperature trends move. Groundwater levels drop even without extraction because of regional changes. Your programme is attempting to improve outcomes against a worsening backdrop.
Three attribution strategies
Difference-in-differences with climate controls -- compare treated and comparison areas over time, but control for climatic variation using IMD gridded data. If treated areas experienced the same rainfall shock as comparison areas but had better outcomes, the difference is attributable to the programme.
Contribution analysis -- trace the causal chain from intervention to outcome, documenting the contribution of the programme relative to other factors (government schemes, market changes, autonomous adaptation). Does not require a counterfactual, but is less rigorous.
Process tracing + climate event analysis -- document how treated communities responded to a specific climate event versus how comparison communities responded. This is qualitative attribution through mechanism documentation.
The autonomous adaptation problem
Indian farmers are not passive. They adapt autonomously -- shifting crops, migrating seasonally, changing irrigation practices. NSSO data shows that 35% of Indian farmers changed cropping patterns between 2013-2023 without any programme intervention. Your evaluation must distinguish programme-driven adaptation from autonomous adaptation. Ask: "Would they have done this anyway?"
The NAPCC alignment
India's National Action Plan on Climate Change (NAPCC) and its State Action Plans (SAPCC) provide the policy framework. Your evaluation can add policy value by explicitly mapping findings to SAPCC priorities. If your programme demonstrates that community-based watershed management reduces drought vulnerability, and this aligns with the SAPCC's water security mission, the finding becomes policy-actionable.
Your Attribution Strategy
Design the attribution approach. These flow into your capstone.
"This evaluation will tell us ___ and will NOT tell us ___."
Saved
Self-check
Your climate adaptation evaluation compares treated and comparison villages. Both experienced similar rainfall. Treated villages had 20% less crop loss. But you discover that 40% of treated-village farmers had also enrolled in PM-KISAN and used the cash transfer for drought-resistant seeds. Can you attribute the reduced crop loss to your programme?
Yes -- the comparison group did not have your programme
Not cleanly -- the PM-KISAN co-intervention confounds attribution; you must control for PM-KISAN enrolment or use contribution analysis to estimate the relative contribution of each programme
No -- the evaluation is invalid
Only if PM-KISAN was equally available in comparison villages
Correct. Co-interventions are endemic in Indian rural programmes. MGNREGA, PM-KISAN, PMKSY, state schemes -- all operate simultaneously. Clean attribution requires either controlling for co-interventions statistically or using contribution analysis to document the relative role of each factor.
Capstone
Your Climate Adaptation Evaluation Design Brief
Click Build my brief to compile everything.
Climate Adaptation Evaluation Design Brief
Click "Build my brief" -- your module answers will be pulled into the artefact.
Your brief will appear here when you click "Build my brief".