Both vegetation abundances and community compositions play important roles for the functions of green roofs (e.g. stormwater retention, habitat provision, aesthetic appearance). However, green roof vegetation can change significantly over time, which may consequently affect the functions related to them. This study investigated vascular plant covers and species compositions on 41 roof sections located in Sweden’s subarctic and continental climate zones. For the roof sections with a known originally intended vascular plant composition (n=32), on average 24±9% of the intended species were present in surveys while unintended species made up 69±3% of the the species found. The Intended species dominated plant cover (93±3%) and Sedum acre (58±36% cover) was the most commonly found species. As revealed in previous studies, substrate depth had a positive relationship with plant cover and species richness. The vascular plant cover of the roofs in this study was not related to species richness as hypothesized but instead had a significant negative correlation with moss cover. The results in this study emphasize the importance of substrate depth for both plant abundance and species diversity, and that even in a cold climate, colonising unintended species can have a great contribution to the species richness of green roofs. However, since most colonising species formed sparse cover on the roofs, their potential benefit to green roof functions that benefit from a dense vegetation cover (e.g. stormwater management and thermal performance) could be limited while the intended vegetation performs these functions more effectively.

During the vegetation survey, the following variables were recorded. When referring to their use as modelling parameters (and for convenience here) these variables are italicized:

• Species richness (number of vascular plant species detected on a roof section)
• Cover of individual vascular plant species in quadrats (%)
• Total vascular plant cover (%)
• Total moss cover (%)
• Substrate depth (mm)
• Age (years)
• Roof area (m²)
• Roof width (m)
• Roof length (m)
• Aspect (in cardinal direction: North, East, South or West)

Stratified random sampling was used to estimate the vegetation cover on the selected roof sections. For this, sampling quadrats (each consisting of a 1×1 m wooden frame) were evenly spread along transects on each roof section, avoiding edge zones such as ventilation shafts. The number of quadrats used on each section (3-54) was limited by the size of roof sections. The quadrats covered a minimum of 6% of the roof section surfaces and vegetation cover was estimated visually with grids in each quadrat. The buildings’ owners or suppliers of the green roofs provided the age of the roof and the substrate depth was measured in each sampling quadrat as an average of three insertions using a 3 mm diameter metal rod. Total vascular plant and total moss cover were recorded separately as growing in separate layers, so their combined cover in a sampling quadrat could potentially exceed 100%. Vascular plant species were determined to species level when possible, using floras (Mossberg and Stenberg, 2010; Krok et al., 2013), but in exceptional cases only to genus (e.g. Taraxacum). The percent cover of each detected vascular plant species was recorded in each sampling quadrat and mean values were calculated for each roof section. Species represented by small single specimens were assigned a minimum of 0.1% cover. All vascular plant species found outside the sampling quadrats were also recorded, during a further limited time spent searching for rare species on each roof section after scrutinizing the quadrats. The species found outside of quadrats were not assigned any percent cover values, but were included in the list of species, and thus contributed to the recorded species richness, in each section.