Environmental monitoring data and frog abundance

We calculated M30DAT for each river as the running mean of consecutive water temperature measurements from the previous 30 days for each day between May and October, averaging 2009, a cool year, and 2010, a warm year. We determined the temperature over the warmest 30 days of the summer, a period that coincides with rapid growth and development of tadpoles. We assessed water temperature from spring to autumn of 2009 and 2010 in 12 northern California river systems, pairing drainages affected by dams with drainages lacking dams (See in Article: Figure 1, and Appendices S1 and S2 in Supporting Information). Each pair was in the same catchment, having similar geomorphology and habitats. We positioned data loggers (Thermocron iButtons DS1921G, 0.5°C accuracy, recording every 2 hr), in typical egg and tadpole habitats (Figure S2 in article). We obtained additional water temperature data collected by dam operators in our six catchments (Placer County Water Agency, Pacific Gas & Electric, and San Francisco Public Utilities Commission) for 63 monitoring stations, of which 51 were used by frogs for reproduction and comprise the sample size of our analyses. We characterised physical habitat of occupied sites by performing Principal Component Analysis (PCA) of the following variables at each monitoring station: mean of M30DAT for 2009 and 2010 (see above), mean annual discharge over the full period of record for the nearest stream gauge (United States Geological Survey, http://relicensing.pcwa.net/html/science/hydrology.php), elevation, and catchment area upstream of the station. Variables were scaled prior to PCA on a correlation matrix. Following PCA, we extracted the first two principal components, which accounted for 73% of the total variance.Discharge and drainage area loaded strongly on PC1, whereas M30DAT and elevation loaded strongly on PC2 (Table S2). We used k-means clustering of PC1 and PC2 to delineate the riverine environments occupied by breeding frogs. We examined scatter plots of PC1 versus PC2 using k = 4 and delineated distinct clusters of sites.Data on frog abundance refer to number of clutches per km of river reach. We conducted surveys of frog reproduction by searching for clutches of eggs which females attach to rocks in shallow water. Rana boylii has a lek mating system in which frogs congregate and oviposit at areas of coarse sediment deposition in a river, such as lateral cobble bars (Kupferberg, 1996). By searching several riffle pool sequences, approximately 500 m upstream and downstream of the water temperature monitoring stations, we covered multiple depositional environments where leks might occur. We repeated visits (≥29) throughout the breeding season to ensure high detectability of clutches and followed an established protocol (Kupferberg et al., 2012) searching both banks and marking clutches to prevent double counting. We standardised clutch abundance by the linear distance searched. Survey periods varied among rivers because frogs breed earlier in the spring (March–May) in coastal rain-driven systems, and later (May–June) in the inland rivers at the foothills of the Sierra Nevada mountains (henceforth, inland rivers), which receive rain and snowmelt