Panel data analysis modeling of marine plastic debris

Sammanfattning

Plastic debris accumulation on Nordic beaches is a growing environmental concern, yet quantitative evidence linking oceanographic and meteorological conditions to beach-level deposition remains scarce. This thesis uses an unbalanced panel of seven beaches along the Swedish western and southern coast surveyed three times annually between 2012 and 2025 by Håll Sverige Rent. Combined with hourly oceanographic and meteorological data from three SMHI buoys and weather stations, the analysis estimates six panel regression models, for both fixed effects transformation and first-difference estimation with cluster-specific fixed effects. The first-differenced models are shown to be superior in handling this type of data based on residual diagnostics as well as coefficient and standard error estimates. Maximum wind speed emerges as the most consistent predictor: a one meter per second increase is associated with a 10.52 to 13.89 percent increase in plastic items per 100 metres, all else constant. The study also addresses the issue of having few clusters by utilising leave-one-cluster-out jackknife. The jackknife indicates that maximum windspeed proves robust and that mean sea surface temperature, included as a control variable, exhibits a consistently negative association with plastic counts. By contrast, current-related variables and the proportion of onshore winds prove sensitive to cluster omission, limiting the conclusions that can be drawn about these mechanisms. The findings provide tentative empirical support for wind intensity as a driver of plastic accumulation on the Swedish western and southern coast and serve as groundwork for future panel studies of marine debris.