Local catch history models reduced winter bank vole correlation in four Swedish areas in 2003–2025

We tested how much spatial correlation in seasonal bank vole catch change remained after local catch history adjustment. We used Swedish monitoring records from four fixed areas in 2003–2025. Each area and season had a separate linear model. The models used starting catch and an earlier autumn catch. Mean winter correlation across six area pairs was 0.458 before adjustment and 0.084 after adjustment. The reduction was 0.374, with a nominal 95% year block interval of 0.217 to 0.518. With four year test blocks excluded from fitting and a one year training buffer, mean winter prediction error correlation was 0.060. Two and six year test blocks gave 0.065 and 0.027. Each winter model predicted better than a local mean baseline. The three northern areas retained an error correlation of 0.461 in the four year test. Summer error correlation was 0.292 across all four areas. We also reproduced an earlier SCI seasonal estimate. The added model checks are exploratory. They show that local catch history predicts part of raw winter correlation. They do not identify a cause of synchrony or rule out shared environmental effects. The package preserves source workbooks, provenance, plans, code, predictions, resampling draws, figures, and independent checks.

bank vole · spatial synchrony · seasonal dynamics · local catch history · cross validation · open data