Moisture availability shifts heatwave carbon release from respiration stimulation to photosynthetic limitation

We obtained half-hourly (HH) flux and meteorological data from three networks: FLUXNET2015, AmeriFlux, and ICOS. Only records processed with the Nighttime Partitioning (NT) method were used to ensure consistency in deriving ecosystem fluxes. The following primary variables were extracted from the “FULLSET_HH” files: GPP (GPP_NT_VUT_REF), RECO (RECO_NT_VUT_REF), and NEE (NEE_VUT_REF). We then retained all FLUXNET2015 records and appended daily records from ICOS and AmeriFlux only for site-years not covered by FLUXNET2015. Half-hourly records were aggregated to daily resolution. For each site and day, we extracted NEE, RECO, GPP, SW, air temperature, VPD, precipitation and SWC. Daily NEE, RECO and GPP was computed as the sum of accepted half-hourly values. Meteorological summaries included TMAX, VPD, daily cumulative precipitation, daily mean SW, and daily mean SWC. Data quality was quantified as the fraction of valid daytime half-hours (qc_day); we retained days with qc_day > 0.5. Site metadata (latitude, longitude, IGBP class, MAT, and MAP) were merged by site ID. Analyses were restricted to vegetated IGBP classes and to sites with ≥ 5 years.