Spectral Responses of Urban and Peri-Urban Woody Canopy to Recurrent Drought: A Matched Three-City Pilot

Urban conditions may increase woody-vegetation drought stress, although management and canopy composition can modify urban–peri-urban differences. We compared spectral anomalies of persistent woody canopy in Berlin, Madrid, and Vienna using Harmonized Landsat–Sentinel-2 observations from 2013 to 2024. Seventy urban–peri-urban pairs were matched on elevation, long-term temperature, precipitation, and retained woody area. Of 26 predefined candidate recurrent-drought events, 12 met the 80% observation-support criterion. Outcomes were event medians of within-pair urban-minus-peri-urban anomalies in the two-band enhanced vegetation index (EVI2) and normalized difference moisture index (NDMI). Each index produced seven positive and five negative contrasts; medians were 0.00047 for EVI2 and 0.00065 for NDMI. Event-level models adjusted for city and current drought severity examined antecedent severity and inter-event interval. The standardized predictor coefficients ranged from −0.00225 to 0.00252 in terms of index units. All four heteroskedasticity-consistent 95% confidence intervals included zero. The same pattern emerged in pair-within-event resampling, spatial buffering, and models based on growing-season severity, with every interval crossing zero. As the primary regressions contained 12 events and five fitted terms, there were only seven residual degrees of freedom, resulting in imprecise estimates and low statistical power. Climate antecedents were identifiable for all 26 candidates, and analytical inclusion was determined by vegetation observation support. Across the retained cases, we found neither a consistent urban spectral disadvantage nor a clear relationship between the contrast and drought history. The matched design is reproducible, but the evidence it provides comes from a small sample selected based on observation availability and from dense canopies that persisted throughout the study period. Accordingly, these estimates should not be generalized to whole urban forests or interpreted as measures of physiological response, mortality, or canopy loss.