Mortality and recruitment in an urban forest (Forest Park in Portland, Oregon) between 1993 and 2003
- First Online:
- Cite this article as:
- Broshot, N.E. Urban Ecosyst (2011) 14: 553. doi:10.1007/s11252-011-0170-y
- 198 Views
I examined tree recruitment and mortality over a ten-year period at permanent plots in an urban forest, Forest Park, in Portland, Oregon. The density and diameter at breast height (dbh) for all trees living and dead were measured in 1993 and again in 2003. Data were analyzed using paired Student t-tests. I found significantly fewer live and significantly more dead trees in 2003 than in 1993. The increase in mortality was significant for all species of trees and for all sizes except large diameter trees. Mortality rates ranged from 0% to 67% at my sites. Recruitment was lower at all sites in 2003 with significantly fewer seedlings and saplings. The high mortality and low recruitment resulted in a net loss of trees at all sites. Loss of trees was not offset by increasing tree diameter, which suggests self-thinning is not the cause. No strong correlation with an urban to rural land use gradient was observed. The results may be related to global climate change or pollution. The high mortality of trees of all species in many diameter classes without a concomitant increase in recruitment could lead to dramatic changes in forest structure.
Urbanization has been shown to affect natural areas in a variety of manners. Many urban forest ecosystems exhibit reduced species diversity (Airola and Buchholz 1984, Sharpe et al. 1986, Pysèk and Pysèk 1990, Tait et al. 2005, Broshot 2007). Tree stand characteristics affected by urbanization include changes in density, basal area and/or importance value of trees in urban forests (Airola and Buchholz 1984, Rudnicky and McDonnell 1989, Broshot 2007). Tree growth rates have been shown both to increase (Iakovoglou et al. 2001) and decrease (Quigley 2002; Carreiro and Tripler 2005). Tree regeneration has been shown to decrease for some species, yet increase for others, especially earlier successional or non-native species (Lehvävirta and Hannu 2002, Zipperer 2002, Komuro and Koike 2005).
Several studies show urbanization may be interfering with normal succession processes in forests (Kowarik 1990, Matlack 1993, Leak and Smith 1996, Porter et al. 2001, Quigley 2002, Komuro and Koike 2005, Broshot 2007). Tree mortality and recruitment are important parameters in determining how succession will proceed, but limited studies have been done in urban areas. Most of the urban studies have assessed street trees (Gilbertson and Bradshaw 1985, Nowak et al. 2004), which do not necessarily respond similar to trees in natural forested urban areas. Several workers have reported reduced recruitment in urban forests (Greller 1975, Airola and Buchholz 1984, Fahey 1998, Lehvävirta and Hannu 2002, Zipperer 2002, Bertin et al. 2006, Broshot 2007), and some have reported increased tree mortality in urban areas (Davis 2004, Nowak et al. 2004, Carreiro and Tripler 2005, Broshot 2007). Nowak et al. (2004) reported that tree mortality in urban, forested areas was moderate as compared to that in more urban areas, but made no comparison with more rural forests.
Tree recruitment and mortality have been shown to be patchy in nature (Franklin et al. 1987; Peet and Christensen 1987; Busing 2005) making it important to determine whether changes seen in urban forests are different from those in more rural forest systems. In order for succession to proceed, mortality of canopy trees must be offset by recruitment of later successional species. A combination of high mortality and low recruitment could lead to dramatic changes in forest structure (Franklin et al. 1987, Lorimer et al. 2001, Bertin et al. 2006, Lutz and Halpern 2006).
Recent studies have reported increased tree mortality over broad geographical areas (van Mantgem and Stephenson 2007, Adams et al. 2009, van Mantgem et al. 2009). High worldwide tree mortality has been positively correlated with increased water deficits caused by higher temperatures related to global climate change (Guarín and Taylor 2005, Mueller et al. 2005, van Mantgem and Stephenson 2007, Adams et al. 2009, van Mantgem et al. 2009). Because urban forests often experience warmer temperatures due to an urban heat sink effect (Bornstein 1968, Kim 1992), tree mortality could provide early indication of the impact a changing climate may have in the future (Carreiro and Tripler 2005). The issue is also confounded by the fact that urban forests are often exposed to higher levels of air pollutants, particularly ozone that has been shown to increase mortality of younger trees (Lehvävirta and Hannu 2002, Gregg et al. 2003). In addition, the high recreational use of urban forests may increase mortality by trampling and/or soil compaction (Bagnell 1979, James et al. 1979, Bhuju and Ohsawa 1998, Littlemore and Barker 2001, Lehvävirta and Hannu 2002, Zipperer 2002, Moffatt and McLachlen 2004).
In 1993, I found reductions in tree species richness and diversity, as well as in the density and diameter of shade-tolerant (later successional) species of trees such as western red cedar and western hemlock with urbanization. I also found a dearth of young trees, especially shade tolerant species, in the more urban areas of the park (Broshot, 2007). Since 1993, development of the land surrounding Forest Park has continued and recreational use has intensified (Broshot, unpublished data). My initial study showed changes along an urban to rural land use gradient, however, long-term studies on urban forests are needed (Nowak et al. 2004). In this study, I returned to permanent plots in 2003 and repeated earlier measurements in Forest Park.
This paper presents the results of a 10-year study of tree recruitment and mortality in an urban forest. I had several objectives: (1) to determine whether levels of tree recruitment and mortality have changed over the 10 years of the study and if so, whether those rates vary by tree size or species; (2) to determine whether tree recruitment or mortality correlated with level of urbanization; and (3) to propose potential reasons for the findings, which could lead to further research.
The areas surrounding my sites have various land uses that create an urbanization gradient running from crowded urban Portland residential neighborhoods and industrial sites on the southeast to rural lands on the northwest and surrounding the old-growth (Broshot 2007). Over the ten-year period between data collections, land surrounding the park has continued to be developed, mostly for single-family residential homes.
Forest Park extends in a northwesterly direction along the Willamette River floodplain on the east side of the Tualatin Mountains. Elevations within the park range from 20 m near the base of the Tualatin Mountains to 335 m above sea level at the crest. Soils at my sites are fairly uniform, being composed of Cascade and Goble types, both of which have an underlying fragipan between 51 and 114 cm deep that prevents water penetration. Much of the area includes slopes that exceed 30% grade, making erosion, slumping, and landslide hazards high. The climate for the area is maritime with mild temperatures, prolonged cloudy periods, warm dry summers and wet cool winters (Green 1983).
The study area lies in the Douglas fir (Pseudotsuga menziesii Mirbel Franco)—western hemlock (Tsuga heterophylla Raf. Sarg.) vegetation zone. Western hemlock and western red cedar (Thuja plicata Donn.) are considered climax species for the area. Although Douglas-fir is an early seral species, many forests are dominated by it because it routinely lives for more than 500 years (Franklin and Dyrness 1988, Agee 1991, Franklin and Spies 1991). In addition, Douglas-fir is a fire-adapted species able to survive fires that kill other tree species, including western hemlock and western red cedar. In western Oregon, the return time for fire is estimated to have been 230 years before Europeans began a policy of fire suppression in the mid 1800s (Agee 1991). Frequent disturbances in more urban areas favor invasion by bigleaf maple (Acer macrophyllum Pursh) and red alder (Alnus rubra Bong) (Franklin and Dyrness 1988). As a result of continual disturbance, many sites in the Portland area are dominated by those two species of shade-intolerant deciduous trees (Houle 1996).
I collected data from the 25 randomly located field sites during the summer of 1993; I returned to the same sites and recollected data in the same manner in the summer of 2003. Three, 30 m long, parallel transects were established at each study site. Transects were laid so that they ran up slope at each location. The 30 m transect length limited the variability in slope, aspect and elevation, each of which was measured at midpoint on each transect and found to not be correlated with any measure of urbanization. All members of tree species, living or dead, within a 250 square meter quadrat (4.17 m on either side of the 30 m transect) were identified to species and the diameter at breast height (dbh) of each was measured. Each quadrat was 20 m away from the adjacent quadrat. In this study, I defined trees as members of arboreal species with a dbh greater than 10 cm. Saplings were defined as members of tree species that were greater than 1.37 m tall with a dbh less than 10 cm. All tree seedlings (members of tree species less than 1.37 m tall) that were present on each transect were identified to species and the basal diameter and height were measured. I noted whether dead trees were standing or blown down.
Data were compared between 1993 and 2003 using two-tailed, paired Student t-tests. Data were also analyzed using analysis of variance (ANOVA), where each section of Forest Park (e.g., city, middle, far), as well as the old-growth stand was designated as a treatment. Tukey-Kramer HSC values were used to determine significance among the categorical variables in each ANOVA. This particular test was selected because it is a conservative test and because I did not have equal numbers of samples in each group (Johnson and Wicher 1982, Sokal and Rohlf 1995; JMP 2005). The statistical program JMP 6.0 (SAS Institute 2005) was used for all statistical analyses. To meet the assumptions of normality required for t-test and ANOVA, some variables were transformed (Sokal and Rohlf 1995). Tree variables were reported only for those species that were present in at least three sections and whose density was large enough to examine statistically.
The total number of live trees measured declined from 1,956 stems in 1993 to 1,555 stems in 2003. Although 17 tree species were measured, only six had high enough densities to be analyzed; five of the tree species were non-native. Over the same ten-year period, the total number of dead trees increased from 554 to 1124. The total number of seedlings (trees <1.37 m tall) decreased from 76 to only 37 stems.
Mean living and dead tree density per hectare and stand diameter (standard error) in 1993 and 2003. P-value from two-tailed, matched student t-test. n = 75 for all categories
# live trees/ha
# live shade tolerant/ha
live tree total dbh/250 m2
# dead trees/ha
# dead shade tolerant/ha
% dead trees
Mean density per 250m2 (standard error) of live seedlings (<1.37 m tall), saplings(<10 cm dbh) and trees in the different sections in 1993 and 2003. p-value in right column is for the ANOVA comparing the sections; Values with different letters are significantly different (Tukey-Kramer HSD) from one another (p < 0.05). t-test p-value below each column relates to the significance of the variable in the 2 years. N = 24 for the city, middle and far sections; N = 3 for the old-growth
30.6 (2.0) b
34.3 (3.7) b
10–20 cm dbh
20–40 cm dbh
40–60 cm dbh
>60 cm dbh
rate of recruitmenta
Mean density per 250m2 (standard error) of dead seedlings (<1.37 m tall), saplings(<10 cm dbh) and trees in the different sections in 1993 and 2003. p-value in right column is for the ANOVA comparing the sections; Values with different letters are significantly different (Tukey-Kramer HSD) from one another (p < 0.05). t-test p-value below each column relates to the significance of the variable in the 2 years. N = 24 for the city, middle and far sections; N = 3 for the old-growth
% dead trees
<10 cm dbh
10–20 cm dbh
20–40 cm dbh
>40 cm dbh
On average, mortality doubled at sites from 1993 to 2003. Forest Park appears to be experiencing high levels of tree mortality, with significantly higher mortality in all tree species and in all but the largest trees. A study of old growth forests in Yosemite Valley, California showed similar findings (Guarín and Taylor 2005). Studies in other locations have shown high mortality in very small and very large trees (Busing 2005). Frequently mortality rate varies by tree species (DeBell and Franklin 1987, Lorimer et al. 2001, Smith et al. 2005), but I found high mortality in all tree species. I also found reduced rates of recruitment as indicated by significantly fewer seedlings and saplings. This finding is similar to other studies (Lorimer et al. 2001, Bertin et al. 2006), but van Mantgem and Stephenson (2007) found no significant change in recruitment rates in the Sierra Nevada.
Tree mortality data from other studies
% dead trees
Annual mortality rate (%)
Stage of stand development
Location of study
Acker et al 2006
Mount Ranier, Wa
Franklin et al. 1987
Andrew's Forest, Or
Franklin et al. 1987
Wind River, Wa
Franklin et al. 1987
Mt Ranier, Wa
Guarin and Taylor 2006
Lorimer et al 2001
Great Lakes region
Lorimer et al 2001
0–3.8 (sugar maple)
Great Lakes region
Lorimer et al 2001
Great Lakes region
Lutz et al. 2009
Nowak et al. 2004
Smith et al.
Sierra Nevada, Ca
van Mantgem and Stephenson 2007
Sierra Nevada, Ca
Broshot 2003 study
young to mature
The average rate of recruitment as examined by the number of new trees (not just seedlings and saplings) present on sites in 2003 ranged from 0 per ha in the old-growth to 84 per ha in the middle section. Given that some of these trees may have been up to a decade old, this number is very low compared to data from other studies (Lehvävirta and Hannu 2002, van Mantgem et al. 2009). The combination of low recruitment and high mortality resulted in a net loss of trees ranging from 428 trees per ha in the old growth to 156 trees per ha in the middle section. The number of trees lost exceeded the number of new trees in all sections and at every site. I also found that the total diameter of trees in the park declined significantly. This suggests that the high tree mortality was not the result of self-thinning. During the thinning stage of stand development, a loss in tree density should be balanced by an increase in tree diameter (Peet and Christensen 1987, van Mantgem et al. 2009). DeBell and Franklin (1987) observed slight increase in tree biomass that balanced the loss of biomass to mortality in old-growth forests. I also found significantly fewer shade tolerant trees in every section of the park compared to what was found in 1993. The loss of canopy trees should allow the release of shade tolerant trees in the understory under normal circumstances (DeBell and Franklin 1987, Franklin and Dyrness 1988), but the dearth of shade tolerant trees in the understory could lead to a major disruption in the successional processes in Forest Park. DeBell and Franklin (1987) found mortality in Douglas-fir in old growth forests in Washington was compensated for by growth of western hemlock. This was not observed in this study where I found high mortality and reduced diameter of all species relative to 1993.
If the process of self-thinning is not responsible for the observed increase in mortality, is there an effect due to urbanization, which could show up along an urban to rural land use gradient. In both 1993 and 2003, I found significantly more dead trees in the far than either in the city or middle sections of Forest Park, however the annual mortality rate was not significantly different among the different sections. In both years there were significantly more live seedlings in the far section than the city or middle sections and although there was a tendency for fewer seedlings in all sections in 2003, the findings were not significant. There were significantly fewer saplings (trees less than 10 cm dbh) in 2003. In 1993 there were significantly more saplings and shade tolerant trees in the far section than in the city and middle sections of the park, however there were no significant differences among the sections in 2003. A large proportion of the trees in Forest Park are dying without shade tolerant saplings in the understory to grow into the consequential gaps. In 1993, it appeared that succession was being impeded by urbanization in the section of the park closest to downtown Portland (Broshot 2007). The 2003 findings suggest that the park is becoming more homogeneous. It seems unlikely that disturbances associated with surrounding land usage have expanded to the extent that all sections face the same pressures. This leads to the conclusion that something else is responsible for the observed changes.
Tree mortality is a complex process. Mortality of young trees is often caused by suppression or self-thinning, whereas deaths of older trees are usually brought about by a combination of factors. Biotic factors include competition, herbivory, insects and disease. Abiotic factors include environmental stresses such as drought, wind and pollution (Franklin et al. 1987, Tkacz et al. 2008). Competition and self-thinning can be ruled out as the major cause for my findings because the diameter of trees and the density of shade tolerant trees also declined. Tree mortality can also be episodic due to major storms or diseases (Franklin et al. 1987), but there have been no major insect or disease outbreaks. Herbivory by ungulates does not appear to be causing widespread death of young trees in Forest Park (Broshot, unpublished data). Only 5.8% of tree mortality resulted from blow down, whereas 39% of the dead trees were still standing. Thus some other reason must account for the findings.
The increase in mortality in this study was observed in all species and sizes except larger trees (>60 cm dbh). Potential causes for the mortality could be the effects of air pollution or global climate change. Increased dust deposition of particulate pollution has been reported in urban areas (Pouyat et al. 1995, Lovett et al. 2000). Such pollutants could alter the soil ecosystem in such a manner as to reduce growth rate and vigor of trees (Ashby and Fritts 1972, McLaughlin et al. 1983, Pouyat et al. 1995, Zipperer et al. 2000, Tkacz et al. 2008). Gregg et al. (2003) reported decreased growth in rural trees due high ozone levels. Urban air pollutants interact with biogenic volatile organic compounds produced by forests and agricultural lands to raise ozone levels in areas downwind from cities (Campbell et al. 2000). The resultant ozone can accumulate to levels above a threshold (40 ppb—the AOT 40) that has been documented to cause foliar damage (Kinnee et al. 1997, Vollenweider et al. 2003, Goldstein et al. 2004). High levels of ozone cause visible damage to foliage, but reductions in growth without visible damage have also been documented (Gregg et al. 2003, Gregg et al. 2006). Plants exposed to ozone in controlled growth chambers lost control of stoma, resulting in reduced photosynthetic rates, which led to a significant reduction in biomass accumulation (Gregg et al. 2006). The Oregon Department of Environmental Quality reported that in 2008 the average level of ozone measured in areas surrounding Portland ranged between 75 and 85 ppb (Oregon DEQ 2010), levels sufficient to cause damage to vegetation (Vollenweider et al. 2003). This points out that measuring levels of ozone in Forest Park will be an important step.
Another possible explanation for the observed higher mortality could be related to global warming with concomitant increases in water stress. Several studies have recently reported higher tree mortality in many areas of the world (Mueller et al. 2005, van Mantgem and Stephenson 2007, Adams et al. 2009, van Mantgem et al. 2009). The increase in tree mortality has been observed in both tropical and temperate (both Europe and North America) forests (van Mantgem and Stephenson 2007, Adams et al. 2009, van Mantgem et al. 2009). A correlation has been observed between the observed increase in mortality and water stress resulting from the higher temperatures of global climate change (Guarín and Taylor 2005, Mueller et al. 2005, van Mantgem and Stephenson 2007, Adams et al. 2009, Lutz et al. 2009, van Mantgem et al. 2009). The high mortality in observed in my study may be the result of climate change, perhaps intensified by the urban heat sink effect (Bornstein 1968, Von Stulphagel et al. 1990, Carreiro and Tripler, 2005).
In this 10-year study, I found increased mortality in all species of trees in all sizes except very large trees. I also found a reduction in recruitment as exhibited by a reduction in seedlings and saplings in 2003. My data suggest that Forest Park may undergo a dramatic change in forest structure in the future. Although a forest in normal successional development could be expected to have increased mortality during the thinning stage, the loss of shade tolerant trees, low levels of recruitment and reduction in tree diameter all suggest that this is not the cause of my findings. The changes in tree mortality and recruitment did not seem to correlate with measures of urbanization using the urban to rural land use gradient. It is possible that the findings may be the consequences of increased ozone pollution or global climate change or a combination of the two. Higher temperatures result in increased ozone development (Goldstein et al. 2004). Warming temperatures could lead to higher ozone levels. The combination of water stress and higher ozone levels could be impacting the trees in Forest Park. Further study into the levels of ozone and the temperatures to which trees at the sites are exposed are necessary to help understand the observations.
This project was funded in part by Portland Parks and Recreation, Portland State University, and Linfield College. I wish to thank Jack Keyes and Todd Rosenthiel for helpful comments and reviews that improved the document. I also wish to thank Peggy Sedlak and Michelle Heater for assistance in the field.