**Fertilizer and Irrigation**

Since 2011, on building 3, subplots were established, as shown in Table 1, to investigate the effects of additional irrigation and subplots of the same size were established to investigate the effects of fertilizer use. These subplots were compared with ongoing monitoring of the plant lists in the normal roof positions.

**Table 1.** The sizes of the 10-cm depth research subplots on building 3 from 2001; on each part of the roofs, the following sized areas were selected for treatments with additional irrigation and fertilizer, and the "normal" comparison plots for the four test plots (in m2).


The different sizes are due to the roof construction.

Additional irrigation was applied once a year in May using 10 L/m2 tap water. The fertilization subplots were fertilized with 10 g/m<sup>2</sup> Kristalon 16-11-16 NPK plus MG on the same day. Figure 3a shows the effect of the fertilizer on the eastern part of the roof and Figure 3b gives an impression of the western part with the sedum mats. *Sedum sexangulare* was stimulated to flower intensively while *Allium schoenoprasum* and *Petrorhagia saxifraga* showed significantly better coverage and performance.

#### **Statistics**

For the statistics, SPSS\_Vers.27 procedures, such as descriptive statistics, dependent *t*-tests with two variables each, tests of significance, cluster analysis, and analysis of variances (ANOVA), were completed using the datasets.

**Figure 3.** (**a**) View to the eastern roof section, left: growing media Op, right: Ul-media. The line demarcating fertilizer application in June 2020 can be clearly seen, about 4 weeks after the application. (**b**) Western roof area, the fertilized area on the Op media is in the foreground while the Ulo section is in the background. The more intense blooming can be seen on 20 June 2020, four weeks after the fertilizer application.

#### **3. Results**

The effects of using different growing media, substrate depths, and fertilizers were significant and are demonstrated in the following section by the statistical improvements. The additional application of 10 L/m<sup>2</sup> water had no impact on the vegetation and will not be described in detail.

#### *3.1. Cluster Analysis 2001–2020: Test Based on Plant Species Observation*

The annual application of fertilizer affected the visual outcome as demonstrated in Figure 2.

**Test 1:** Analysis of the first year on building 3 showed the development of the plant species richness in the turf mats in Table 2. The results, supported by the cluster/correlation analyses, are shown in Table 2, with the species mix on all turf mats with no additional treatments being highly significantly similar with 0.896\*\*\*, as demonstrated by the examples of the turf mats on the Ulo-West and Op-West test plots. Both mats in the year 2020 were also highly significant, with 0.963\*\*\*.

**Test 2:** How did the growing media influence the number of species and the coverage value? In general, this was demonstrated here by the north test plots in 2001. The Op media in all cases showed higher species richness, as shown by way of the example in Table 3, with 25 species on the Ulo medium and 29 species on the Op medium.

At all times in 2001 and 2020, the Op media showed higher coverage values. In general, the number of observed species decreased over the past 20 years, but in contrast, the fertilizer test plots showed higher coverage values in this time.

**Table 2.** Cluster analysis: correlation based on the number of species in 2001 and 2020 on the west roof, 10-cm media. High significance between species in 2001 on both media (Ulo and Opti). Also high significance in 2020 on both media. Time is more important than the different media.



**Table 3.** Changes in the number of species from 2001 to 2020, 10-cm media.

In Table 4, the similarity is demonstrated by the Pearson index. Again, the similarity of the mats on the north side in 2001 was highly significant. The species similarity remained significant until 2020 in the shade on both test plots of "normal" and "fertilizer". This is demonstrated by the high grass coverage. In contrast to this, the south section differed, with higher numbers of sedum species. Shade is a significant factor determining the plant species mix.

**Test 3:** Species development of the cuttings, eastern exposure test plots.

**Table 4.** Correlation and significance (two-tailed) of the north test plots, 10-cm media, based on the number of similar plant species, in a comparison between 2001 and 2020. High significance \*\* on the 5‰ level between both media Ulo and Op 2001, it remain similar until 2020 on the shade plots. The sunny parts are significant different from these.


The dark material surface of Ulo has significantly higher temperatures during summer than other comparable professional roof growing media. This limited the plant growth of the seedlings and cuttings. Additionally, spontaneous plants found it difficult to establish themselves in this first year. Table 5 show the low number of species in the first year, with 9 (Ulo) and 14 (Op) species and low coverage values of 9% and 14%. These low rates were ultimately determined by the growth of *Petrorhagia saxifraga* on both east test plots, but these plants only covered a small area. In addition, the sedum cuttings must first establish themselves on the test plots. The vegetation on both areas needed a few years to achieve the estimated coverage value of 60%.

**Table 5.** Changes in the number of species on the east area with sedum cuttings from 2001 to 2020, 10-cm media.


Like the other exposure, there were differences between the fertilized and normal areas in terms of the plant species diversity. The effect of the fertilizer on all test plots is to reduce the species richness while increasing the coverage values.

Table 6 shows the correlation on the seeded eastern plots. The differences in early 2001 between the plant mix on Ulo and Op indicate widely different development. This was apparent in 2001, with very low vegetation coverage and fewer species compared to the turf mat on the other sites. For many years, the Ulo test plots in particular had low performance values. Finally, all indicators in 2020 showed that in the long run, similar development was achieved for both media.

**Table 6.** Correlation and significance (2-tailed) of east test plots, 10-cm media, based on the number of similar plant species, in a comparison between 2001 and 2020. Significance here means: the fertilizer equalize differences between the different vegetation stands also on the east plots.


<sup>1</sup> There are not enough similar species for the correlation pairs, and significance expression cannot be calculated.

The hierarchical cluster in Figure 4 demonstrates the similarity of the eastern test plots. From the perspective of plant species development, these seeded plots are valuable observation areas. Considering the aim of achieving full coverage on the roof as soon as possible, it is important to note here that this process took nearly 10 years. These aspects must be evaluated in terms of the aims of green roofs and the risk of wind or water erosion on green roof areas that are not fully covered.

**Figure 4.** Dendrogram of the similarities between the eastern test plots from 2001 to 2020.
