Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year

RSE-09493; No of Pages 10 Remote Sensing of Environment xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse

Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year M. Melin a,⁎, J. Matala b, L. Mehtätalo c, J. Pusenius b, P. Packalen a a b c

University of Eastern Finland, School of Forest Sciences, P.O. Box 111, 80101 Joensuu, Finland Natural Resources Institute Finland, Yliopistokatu 6, 80100 Joensuu, Finland University of Eastern Finland, School of Computing, P.O. Box 111, 80101 Joensuu, Finland

a r t i c l e

i n f o

Article history: Received 6 January 2015 Received in revised form 29 June 2015 Accepted 18 July 2015 Available online xxxx Keywords: ALS Lidar Ecology Moose Habitat analysis GPS Forest structure

a b s t r a c t During the last decade, Airborne Laser Scanning (ALS) data has been used increasingly in numerous studies to assess e.g. bird habitats and biodiversity, but less often in studies focusing on ground dwelling mammals. Here, we utilized ALS data to study the role of forest structure in moose habitat use. The data consisted of 18 GPScollared moose in western Finland. ALS data was extracted around the moose locations. The aim was to examine the habitat selection patterns of moose within a year. Special attention was given: 1) to winter to detect when moose-related forest damages occur (when moose favor young forests) and 2) to the calving period to examine the possible role of forest structure in determining where females give birth and where they move with the newborn calves. During calving, females were occupying forests with minimal amounts of vegetation below the height of five meters. Shortly after calving, the females and their calves moved to forests with dense vegetation under the height of five meters, which is explained by the increasing demand of food both for the growing calf and the lactating mother. In summer and autumn, all moose were found more often in mature forests with higher and denser canopies, which is explained by the fact that their most preferred food during autumn (e.g. berries and twigs) grows in mature forests. This trend diminished as autumn turned to winter and moose started to favor vegetation below five meters. In general, we were able to see clear patterns and differences in habitat use between sexes, as well as with calves, and we gained new and more accurate information about the role of the forest structure to calving females. The results show that ALS data alone can yield valuable additional information about wildlife ecology. © 2015 Elsevier Inc. All rights reserved.

1. Introduction A large number of studies conducted in the 21st century have advanced the use of Airborne Laser Scanning (ALS) in forest inventory and planning. Its usefulness is based on the fact that the height distribution of ALS data is related to the vertical structure of the tree canopy. Variables calculated from the data can then be linked to attributes, such as tree height, basal area and ultimately, growing stock and volume. However, when thinking about forest wildlife, attributes such as availability of food, shelter and cover are also very often determined by the structure of the surrounding forest. In the 1960s, MacArthur and MacArthur (1961) acknowledged the importance of threedimensional (3D) vegetation structure in assessing habitat suitability of an area. As ALS produces detailed 3D data about vegetation structure, it can be highly useful in assessing habitats, organism–habitat relationships and biodiversity (Hill, Hinsley, & Broughton, 2014; Müller & Vierling, 2014). The range of species that can be studied with ALS ⁎ Corresponding author at: Suonrannantie 134, 82220 Niittylahti, Finland. E-mail address: markus.melin@uef.fi (M. Melin).

includes marine and terrestrial as well as avian (Vierling, Vierling, Gould, Martinuzzi, & Clawges, 2008). Here, we tested how ALS data can be used to describe the role that forest structure has in moose habitat selection (Alces alces). Moose is the most important game species in Fennoscandian area. It is also an important keystone species for the local forests, as it modifies the species composition through extensive browsing (McInnes, Naiman, Pastor, & Cohen, 1992). It also causes damage to forestry due to its tendency to browse in young seedling stands (Korhonen, Ihalainen, Miina, Saksa, & Viiri, 2010; Lavsund, 1987). In many cases (not always), moose are migratory and have different winter and summer habitats that they migrate between according to the seasonal cycle (Singh, Börger, Dettki, Bunnefeld, & Ericsson, 2012). During the summer, habitat selection of moose is diverse because summer provides moose with plenty of easy sources of food (both trees and green plants). Preferred foods include grasses, deciduous trees and shrubs, and certain water plants (Hjeljord, Hövik, & Pedersen, 1990; Bergstrom and Hjeljord, 1987). During autumn, moose start to eat more mushrooms, twigs and plants, such as blueberry (Vaccinium myrtillus) and common heather (Calluna vulgaris), which become a very common part of the

http://dx.doi.org/10.1016/j.rse.2015.07.025 0034-4257/© 2015 Elsevier Inc. All rights reserved.

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

2

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

diet. Thus, the selection of habitats favors areas with these kinds of food sources (i.e. more mature forests). During winter, the tree species and the age of the forest are the dominant factors affecting habitat selection. Winter habitats are typically characterized by Scots pine (Pinus sylvestris) dominated forests, peat lands, or shrub lands, and when compared to the surrounding landscape, they include more pine-dominated forests and forests of young successional stages (Nikula, Heikkinen, & Helle, 2004). Cassing, Greenber, and Grzegorz (2006) suggested that Scots pine stands that are 5–15 years old are most preferred. In Finland, pine is the most consumed source of food during winter due to its high availability (in the seedling stands) (Heikkilä & Härkönen, 1993). Differences have been found in terms of moose habitat use between males and females (Cederlund et al., 1987; Cederlund & Sand, 1994; Nikula et al., 2004), but so far, few studies have assessed the effect of calves. Calving typically takes place in May or early June, but the exact moment varies due to factors, such as the timing of last autumn's rut, progress of spring, availability of food and the condition of the mother (Bowyer, Kie, & Ballenberghe, 1998; Keech et al., 2000; Saether, Andersen, Hjeljord, & Heim, 1996). Moose typically give birth to one or two calves. In addition to timing, the characteristics of calving sites vary within the landscape. Factors, such as elevation, slope, distance to water and islands have been described as affecting calving site selection (Addison, Smith, McLaughlin, Fraser, & Loachim, 1990; Bowyer et al., 1998; Chekchak, Courtois, Ouellet, Breton, & St-Onge, 1998; Poole, Serrouya, & Stuart-Smith, 2007). Forest structure also plays a major role; past studies have described favoring both for more dense and closed forests and for more open landscapes for calving sites (Bailey & Bangs, 1980; Bowyer et al., 1998; Chekchak et al., 1998; Langley & Pletscher, 1994). Now, due to the ongoing changes that forest management causes in the forest structure (e.g. thinning, clear-cutting), it is important to understand the habitat requirements of moose in relation to forest structure (Markgren, 1974). As mentioned, ALS makes it easy to accurately measure this structure and so its potential has been widely realized and studied over the last decades (see, for instance, Graf, Mathys, and Bollmann (2009); Coops, Duffe, and Koot (2010); Goetz et al. (2010); Flaherty (2013); Palminteri, Powell, Asner, and Peres (2012)). Thorough reviews on recent studies have been provided by Davies and Asner (2014), Hill et al. (2014), and Müller and Vierling (2014). While ALS revolutionized the process of analyzing forests in 3D, Global Positioning System (GPS) based tracking collars revolutionized methods for tracking and locating wildlife. Moose has been a popular focus for these studies (Dettki, Löfstrand, & Edenius, 2003; Dussault et al., 2004; Lowe, Patterson, & Schaefer, 2010; van Beest, Rivrud, Loe, Milner, & Mysterud, 2011). Still, the method of integrating ALS data with exact animal locations from GPS-collars is relatively new. For moose, it has been tested by characterizing summer and winter habitats, in mapping the availability of forage, and in analyzing behavioral responses to thermal stress (Lone et al., 2014a,b; Melin, Packalen, Matala, Mehtätalo, & Pusenius, 2013; Melin, Matala, et al., 2014). Here, we hypothesize that there are structural differences in the areas occupied by moose at different times of the year and that these differences can be identified from ALS data. Our aims were to use ALS to describe the role that forest structure has in habitat use of moose during different seasons, and to see how sex, and especially the presence of calves, affect it. Special attention will be given to two times of year, winter and the calving period, because these are the most crucial times regarding survival, especially to females and their calves.

variation is less than 10 m in most parts of the study area. Scots pine is the most common species, but Norway spruce (Picea abies) and downy birch (Betula pubescens) are also present (METLA, 2013). The landscape is characterized by fields and peat lands and the proportions of downy birch can be relatively high in peat lands. The inland waters of the area are small lakes, ponds and rivers. A total of 48% of the inland waters are less than one hectare in size and 94% are less than 10 ha. The rivers flow steadily without large rapids and the majority are less than 20 m wide. The density of the moose population (after the hunting period) in the study area is around 3.5 moose per 1000 ha (RKTL, 2011), which is a typical density. 2.2. Data 2.2.1. Moose data The data were obtained from 15 moose equipped with GPS-collars (Vectronic). Of these, 11 were female and four were male. Of the 11 females, five were calving during the study and the calf stayed with the mother during the whole study period. The status of the calving females was checked in early summer (were any calves born) and early winter (were the calves still alive after the hunting season). Each of the calving moose used in this study had one calf at heel during the study period. Moose locations were collected from January 2009 and until the winter of 2011. The collars measured the positions hourly, and every fourth hour, the information was sent to a database (WRAM 2011) via a GSM-network (Global System for Mobile Communications). This research focused on a 365-day period, so we selected a period that included the most moose individuals (i.e. a common period during which most of the moose were followed for the full time of 365 days). This period was between April 10 (2009) and April 11 (2010). An additional reason to delineate the period in this way was that now we were able to see how the calves born in the spring of 2009 affected the mother cows' behavior during the rest of the year. The average fix rate of GPS-positioning was 99.61% (ranging from 98.9 to 99.8%). The times of no fix seemed to happen with no reference to season or time of day. However, problems with the GSM network caused periods of blackout, from 4 h up to even a few days during the late summer and early autumn of 2009 and especially in June (as noted in Melin et al. (2014)). Because of these reasons, not all the positions were usable. 2.2.2. ALS data ALS data were provided by the National Land Survey of Finland (License no. TIPA/517/10-M) and were collected between April 24 and May 5, 2009 using a Leica ALS50 laser scanning system. The test site was measured from an altitude of 2000 m above ground level using a field of view of 40°. The nominal sampling density was about 0.8 measurements per square meter. The footprint of a single laser pulse was about 0.35 m at ground level. The National Land Survey of Finland had previously classified ALS points as ground and non-ground points. The Digital Terrain Model (DTM) was interpolated from the ground points using inverse distance weighted interpolation (Shepard, 1968). Next, the DTM was subtracted from the orthometric heights of the ALS echoes to scale the heights to above ground level (AGL). The laser scanner used in the study captured a maximum of four range measurements (echoes) for each emitted pulse. Here, only the echo categories “first of many” and “only” were used, because they represent surface hits. 2.3. Methods

2. Methods 2.1. Study area The studied moose moved within an area of around 2000 km2 on the west coast of Finland between the latitudes 62.254 N and 63.503 N. The topography of the study area is extremely flat. The overall topographical

2.3.1. Determining the calving period Past studies have shown that the calving period typically occurs in May or June. In the Bertram and Vivion's (2002) study, the calving periods of moose in Alaska ranged between May 14 and June 9, with median dates of May 24 and 25. Whereas, in the Kostroma experimental moose farm (natural or near-natural environment), Bogomolova and

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

Kurochkin (2002) found that 70% of the calves were born during the first half of May. In Haydn (2012), the calving period of moose in Sweden varied between May 12 and June 8. To study the issue, we analyzed the females' movement patterns hour-by-hour and searched for changes in movements that suggested that calving had occurred. According to Poole et al. (2007), mother cows move less during and after calving. Such a decrease in movement pattern should thus be visible from the movement data. Haydn (2012) also used the method of analyzing movement patterns to determine the calving period. The analysis was done by comparing the mean distances between two consecutive GPS-locations (displacement) on one day during different periods of spring and summer (24.4–10.6). 2.3.2. Analyzing forest structure around moose locations To gain information about forest structure around individual moose, we created a circular buffer with a 25-m radius around each moose

3

location and extracted the ALS data inside this buffer (Fig. 1). The 25m radius was concluded to be sufficient enough to describe the vegetation at the near vicinity of moose. Also, since the GPS-collars have a positioning error of 5–10 m, the 25-m radius guarantees that the moose locations were within buffered areas. Next, we analyzed the extracted points within these circular buffers and computed metrics that contained information about the height and density of the forest. The metrics described the proportion of echoes above five meters (p_canopy), echoes that hit the ground (p_ground) and echoes between the ground and the height of five meters (p_shrub). The final metrics were chosen after a thorough investigation of numerous variables and how they were linked to moose habitat use. p_ground was an obvious choice since its value directly (inversely) describes the amount of vegetation in the target area, which is linked to, for instance, the availability of food and shelter. The other two variables are related to the vertical profile of the vegetation. Here, the main point was to distinguish the canopy

Fig. 1. Integration of the datasets: ALS data extracted and visualized around two moose locations from GPS-collars (Melin et al., 2014). The circle has a 25-meter radius.

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

4

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

and shrub layers from one another so that their structures could be analyzed separately. The height limit of five meters was proven to be valid for this purpose. In Dettki et al. (2003), vegetation below five meters was considered as a food source for moose. Also, if the limit had been lower (e.g. two meters), then the probability of catching the shrub layer with ALS would have been very low. With the limit of five meters, some pulses were certain to hit the shrub vegetation, if it existed in the amounts that would be meaningful for moose (as either food or shelter). With the canopy layer above five meters, we did not make further vertical divisions (e.g. 5–10 m) and so anything above five meters was considered as the tree canopy. The selected metrics contained information about the openness of the forest, about the amount of food or the cover it can provide, and about canopy density and height, which can be linked to the age of the forest. In Fig. 1, image a has a p_canopy value of 0.84, meaning that 84% of the echoes reflected above five meters (a dense and high canopy). The same value for image b is only 0.09. The ALS data was similarly extracted around every GPS location (n = 78,816) and separately around males (n = 20,552) and females with (n = 27,363) and without calves (n = 30,901). The grouping was necessary since there are known differences in habitat use between male and female moose (Miller & Litvaitis, 1992; Nikula et al., 2004). Through this, we were able to see how the structure of forests at moose locations changed as the year progressed. 2.3.3. The effects of sex, calf and season We wanted to examine how moose habitat selection, in relation to forest structure, varied seasonally and how much this was affected by sex and the presence of calves. Linear mixed effects modeling (LME) was chosen for studying this issue. LME is an extension of linear regression that is used particularly for grouped data. As with linear regression, the models still describe the relationship between the response- and independent variables, but with coefficients that can vary according to the grouping. Here, the grouping was applied to the moose data based on sex and the presence of calves. Models were created separately for the variables p_canopy, p_ground and p_shrub. The models quantified the effects of different times of the year, sex and the presence of a calf on the structure of the forest at the moose's location. The first created model focused on the full year and used months to distinguish between different times of the year. In addition, we created two models for the mentioned important seasons: winter and the calving period. These models used weeks to analyze the possible changes in the types of forests where the moose were found. The models were formulated as: ymi ¼ f ðtimemi ; sexm ; calf m Þ þ bm þ emi where ymi is the value of the ALS metric y (p_shrub, p_canopy, p_ground) around moose m on day i, timemi indicates the period that was used to distinguish between different times of the year (either month or week), sexm indicates sex (male or female), calfm indicates the presence of a calf (calf or no calf), bm is the random moose effect with variance δ2b and emi is a normally distributed residual with variance δ2 and where cor(emi, emi') = ρðtimemi − timemi0 Þ . The linear models included the main effects and interactions between the variables (time * sex, time * calf) because we wanted to see whether the effects of sex and/or the presence of a calf were stronger or weaker during different seasons. The contrasts of the models were defined so that the responses of consecutive months/weeks were compared against one another, i.e. April to March, May to April, and June to May. This was done to see whether the possible differences in the structure of forests in moose habitats are statistically significant between different months, and more importantly, to see when the significant changes occur. To reduce the spatio-temporal dependence of the GPS observations, the values of the calculated ALS variables around locations from one day were averaged together. The remaining dependence was modeled by individual-specific random effects, which took into account the correlation of observations caused by the possibility of different habitat

preferences of individuals and by the differences in the habitat potentials within an individual's home range. In general, with presence-only data, we cannot be sure about the full habitat use of the animal since we only know where it was during the time of positioning. Therefore, an average from the daily values might distort some of the forests that the moose were in. We compared the daily means of the variables with their daily medians and saw that they had a nearly linear relationship with one another among all three moose groups. However, when the daily modes were contrasted similarly we saw differences between the daily means and modes especially during summer and early autumn (July–September). The differences were most clear with the variable p5, which most likely is a direct cause of the known circadian trends in moose habitat use during summer and especially during the times of high temperatures (Melin et al., 2014). The trend was that the daily means of p5 were underestimates of the corresponding modal values, which means that the areas moose were mostly using were in higher forests than what is suggested by the daily means. During the time between late autumn and spring there were no major differences (or systematic differences) between the modes and the means, which suggests that there were no significant circadian trends or circadian differences in moose habitat use during these seasons. After this analysis, we felt that the daily means gave proper information about the structure of forests in moose habitats except for the summer months (or the months with the highest temperatures and the biggest differences between day and night temperatures). All the analysis was done in R (R Development Core Team 2015). 3. Results 3.1. Timing of calving period The movements of pregnant female moose were at their minimum between April 28 and May 17. The week of least movements was between May 3 and 10 when females, on average, moved 57 m per hour. At the same time, males moved 209 m/h and the other females, 101 m/h. A mean three-day period of least movements occurred between May 3 and 5, during which the females only moved, on average, 18–35 m per hour (m/h). We therefore concluded that calving took place within the first half of May, and most likely between May 3 and 5. Supporting this, Poole et al. (2007) found that during the calving period, the females moved only around 15–17 m/h, with movements increasing slowly after calving. The movements in our data suggested that the calving week, at their minimum, ranged between 15 and 18 m/h, and after this, movements began to rise slowly and were never as low again. 3.2. Habitat use and forest structure within a year — the effects of sex, calf and season 3.2.1. Month-by-month analysis Fig. 2 illustrates how time of year, sex and the presence of calf affected what type of forest the moose were found in. The Figures show the predicted values given by the models (Section 2.3.3) for each month. Variation between the levels of the solid (males) and the dotted (females without calves) lines show the difference relating to sex in the given month (the sex effect). The variations between the dotted (females without calves) and the dashed (females with calves) line then show the differences between females with and without calves (the calf effect). So, without the female factor, the dotted line would meet the solid line. Similarly, without the calf factor, the dashed line would meet the dotted line. Lines set in bold indicate statistical significance. For instance, in the p_ground graph of Fig. 2, the dotted vertical lines in May are set in bold, meaning that in May, all of the groups (males and females with and without calves) were found in areas that were significantly different from one another in terms of landscape openness (p_ground). When a horizontal (solid, dashed or dotted) line is set in bold, it means that the forests

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

5

Fig. 2. Monthly variations in the structure of forests at moose locations within a year, as predicted by the linear mixed effect models. The x-axis denotes the month and the y-axis denotes the predicted value of the ALS variable for that month. Solid line = males, dotted = females without calves, dashed = females with calves.

that moose were found during the corresponding month were significantly different from the forests occupied in the previous month. For instance, in the p_ground graph of Fig. 2, all the lines are set in bold between January and the previous month of December, which means that in terms of landscape openness, the areas used in January were significantly different from those used in December among all three moose groups (males and females with and without calves). The strongest differences were seen from females with calves (dashed line). The differences began during the calving period (May and June) when females with calves were found in forests with lower canopies and more open ground compared to other females. As the summer progressed to autumn, the same trend continued and the females with calves kept moving to more open areas (dashed line in p_ground, Fig. 2). This pattern was seen from the other moose as well, but they were not found in open areas as the females with calves were. In general, as autumn progressed, the moose used forests characterized by lower canopies and more open ground. The clearest change in the full year analysis happened after December when moose began to select open areas with relatively

more shrubs/young forests (i.e. vegetation below five meters – p_shrub). The use of areas with more shrubs peaked in March among all groups, which is linked to known moose habitat selection in the winter. However, our results also show that from December to April, the monthly differences between the occupied areas were always significant with at least one of the variables depicted in Fig. 2. This means that there was a significant difference between the types of areas that moose occupied during the different winter months. The model approach showed that the presence of a calf and the moose's sex affected habitat selection, in addition to showing that this selection had significant monthly variations. Since the clearest variations were found in winter and spring/early summer, this provided increased justification for studying these periods in more detail on a weekly basis. Table 1 shows the random effect parameters of the monthly model. 3.2.2. Week-by-week analysis — calving period The graphs from the weekly models are to be interpreted similarly to those in Fig. 2. Setting in bold indicates statistical significance, either between the groups (vertical lines) or between months (horizontal lines).

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

6

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

Table 1 Random effect parameters of the monthly model. Variable

δ2b

δ2

ρðtimemi − timemi0 Þ

p_ground p_shrub p_canopy

b0.001 b0.001 0.002

0.02 0.005 0.021

0.43 0.44 0.44

Females with calves were generally mimicking the non-pregnant females until April changed to May (when calving began), when they were found in areas that were significantly different from the other groups. The calving period was deduced to have taken place between May 3 and May 10 (Section 3.1), and more likely at the start of May (May 3). First (May 3), calving females were found in areas that did not have much vegetation below five meters (dashed line, p_shrub, Fig. 3). As May progressed, the females and their newborn calves moved to areas offering more vegetation below five meters. The most significant changes

happened in May. In June, when the calf was around one month old, the differences diminished and the areas the mother and the calf were in had significantly more shrub vegetation compared to the areas where the female gave birth. The forest structure in areas used by males had no major variation during the period. The solid graphs with all of the variables remained at a rather steady level, although the p_ground graph confirms the results of Fig. 2. Here, we see that the areas where males were in June were different from those used in April or May. The p_ground graph of males in Fig. 3 shows a steady rise from April to June, i.e. the males were moving to more open areas. Interestingly, after April, the females without calves were found in forests with the highest and densest canopies, with this trend continuing in June. In May, the females without calves were found in areas with a dense and high canopy, which was also significantly different from where the other moose groups were (the vertical lines set in bold, Fig. 3). Table 2 shows the random effect parameters of the calving period model.

Fig. 3. Weekly variations in the structure of forests at moose locations during spring/early summer, as predicted by the linear mixed effect models. The x-axis shows different dates on a weekly basis and the y-axis denotes the predicted value of the ALS variable for that particular time. Solid line = males, dotted = females without calves, dashed = females with calves.

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx Table 2 Random effect parameters of the calving period model. Variable

δ2b

δ2

ρðtimemi − timemi0 Þ

p_ground p_shrub p_canopy

0.003 b0.001 0.003

0.016 0.004 0.02

0.40 0.35 0.43

3.2.3. Week-by-week analysis — winter With the winter weekly models, we wanted to gain more insights into the specific period when the use of young forests (and seedling stands) peaked, i.e. when the forests might be damaged from heavy browsing. Differences between the groups were seen, but they were not the main focus here. Moose strongly favored forests with more vegetation below five meters from February 14 onwards, meaning that the favored areas were now in young forests. For example, the variable that indicated the use of seedling stands (p_shrub) peaked in the second half of March. This indicates that March was the time when seedling stands were most

7

favored, suggesting that this potentially is the time of most heavy browsing. This behavior was the strongest with males and it began to diminish after March 21. From February 21, females made a significant change and went for areas that, in addition to offering more shrubs, also had higher and denser canopies (dotted line, p_canopy, Fig. 4). Females with calves made a similar move from March 7 onwards (dashed line, p_canopy, Fig. 4). Males also showed this behavior; their p_canopy values started to increase from February 14 onwards. At the same time, the p_ground values began to drop with all the groups, meaning that the areas they were using had increasing amounts of vegetation at both the shrub level and in the canopy layer, which indicates that the favored areas were offering both, food and cover. The moose groups did occupy different areas during, for instance, March, but still, the biggest differences were caused by time, not by sex or the presence of a calf. The groups used different types of forests as winter progressed from January to April, which was also noted in Fig. 2. Table 3 shows the random effect parameters of the winter model.

Fig. 4. Weekly variations in the structure of forests at moose locations during winter, as predicted by the linear mixed effect models. The x-axis shows different dates on a weekly basis and the y-axis denotes the predicted value of the ALS variable for that particular time. Solid line = males, dotted = females without calves, dashed = females with calves.

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

8

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

Table 3 Random effect parameters of the winter model. Variable

δ2b

δ2

ρðtimemi − timemi0 Þ

p_ground p_shrub p_canopy

0.002 b0.001 0.003

0.016 0.007 0.018

0.48 0.48 0.50

4. Discussion In this study, we focused on the effect of forest vegetation structure on the full year habitat selection of moose, with special attention to periods of winter and calving. The important finding was that forest structure in moose locations had a significant seasonal variation that was affected by sex and the presence of a calf, but also that it was detectable from ALS data. 4.1. Monthly variations The clearest finding was the favoring of forests with high and dense canopies from spring to autumn, a preference that diminished as winter approached (Fig. 2). Hjeljord et al. (1990) noted a shift towards more mature forests from spring to autumn. The researchers also noted that during autumn, 70–80% of the moose feeding occurred in older forests. Between May and August, the areas with closed forest canopies were utilized most, but the trend continued into autumn as well. Here, we must acknowledge the fact that the daily means of the variable p5 seemed to be an underestimate of the height of the forest moose were mostly using. Thus, the trend in practice is even stronger; moose utilize the older forests even more than what is portrayed in Fig. 2. This would make the results more similar to the 70–80% use of old forests discovered by Hjeljord et al. (1990). It seemed that the circadian differences were related to temperature, since the differences between the daily modes and means of the variable p5 were so clear during summer and early autumn with all the moose groups. This is in line with the findings of Melin et al. (2014) who also saw a significant circadian trend in the use of high canopy forests during summer, and this pattern was present with both males and females (14 out of the 18 moose in their study showed this pattern). After autumn, we saw the beginning of winter migration, when moose favored more open forests with lower canopies (the winter habitats) over the mature forests used in late summer and autumn (Fig. 2). During winter, our results showed the increasing use of young forests from January onwards. The selection peaked in March, suggesting that this was the time when most browsing pressure on seedling stands takes place. The trend was the strongest with males and the weakest with calving females, which could be explained by the previous findings that male moose, due to their larger body size, should select habitats with even more pine-dominated young forests, since they need significantly larger amounts of food (Cederlund, Ljungqvist, Markgren, & Stalfelt, 1980; Nikula et al., 2004). Snow depth can also play a role during winter, with all of the groups showing a tendency towards selecting areas with higher numbers of larger trees (rising p_canopy values from February to April, Fig. 2), which would have less snow than open areas (p_ground values dropped simultaneously). There were not very large (or statistically significant) monthly differences between the groups, except for the calving period and during late autumn and winter. Interestingly, after calving, females with calves were found in areas with more vegetation below five meters than in areas with vegetation above it (Fig. 2). This might indicate that they were searching for areas with more food for the calf and for the mother to meet lactation needs. van Beest et al. (2011) saw the effect of calves on movements disappearing as autumn progressed. We also saw this change from early summer to late autumn. Nikula et al. (2004) found no major differences between male and female habitat use except for during winter. We reached similar results, but also note that there

were statistically significant differences between habitat occupancy during different winter months (from December to April, Fig. 2). This suggests that different winter conditions have a great effect on where moose are found during different winter months. 4.2. Calving period Spring was the period when we focused on the sex- and calf-related differences. As mentioned, studies about the structure of calving sites have been contradictory, with support gained for both the use of dense forests and for open landscapes (Addison et al., 1990; Bailey & Bangs, 1980; Bowyer et al., 1998; Chekchak et al., 1998; Langley & Pletscher, 1994). According to our results, the areas chosen by females during the time of birth offered some canopy cover above five meters, but practically no vegetation below it (dashed line, low p_shrub values at May 3, Fig. 3). After giving birth, the preferences turned to areas with increasing amounts of vegetation under five meters. Singh et al. (2012) suggested that females, instead of maximizing their own survival, tended to select safer habitats to maximize the survival of their offspring (Clutton-Brock et al., 1996, Gaillard, Festa-Bianchet, Yoccoz, Loison, & Tiogo, 2000). It seems here that the females did not favor the areas with any shrubs during calving because the number of shrubby areas where the moose were found increased significantly after the calf's birth. Poole et al. (2007) suggested that the selection of calving sites by ungulates is related to trade-offs between minimizing risk of predation and meeting nutritional needs for lactation, while Trembley, Solberg, Sæther, and Heim (2007) suggested a need to determine which factors influence the selection of a calving area based on the risk perceived by females. Our results emphasized a need for both good cover and good visibility. During calving, the females did not search for completely open areas; rather, they mostly favored the ones offering the best visibility, while also offering some type of canopy. After this, selection preferences clearly turned to forests offering more food for the growing calf and the lactating female (dashed p_shrub line rising from May 3 onwards, Fig. 3). 4.3. Winter During winter, the differences between groups were not the main focus, but we found supporting evidence for the previous findings that male moose, due to their larger body size, need significantly more food than do females and should thus select habitats with more young forests (Cederlund et al., 1980; Nikula et al., 2004). In Fig. 2, it can be seen that males favored areas with vegetation below five meters (Fig. 4). Cassing et al. (2006) suggested that during winter, moose typically utilize seedling stands that are from 5 to 15 years old. These age classes are well linked to our height range of 0.5 to 5 m, and the clear peak (Figs. 2 and 4) in the variable accounting for this height range (p_shrub) proved this. Also related to these results, Melin et al. (2013) noted that moose winter habitats were characterized with significantly more vegetation between the heights of two and eight meters, which during winter is directly linked to the amount of edible vegetation. Another aim was to determine the time period in winter when the young forests (and seedling stands) were most heavily utilized and thus damaged. This was achieved through the week-by-week tracking and modeling. In Fig. 4, we see that moose used young forests from February 14 onwards, and mostly on March 21. The most significant shift towards these areas happened between February 28 and March 7 (p_shrub – Fig. 4). Our results show that moose utilized young forests throughout the winter, but the highest browsing pressure occurred in the second half of March (Figs. 2 and 4). Interesting trends were also seen from February 14 onwards. After this day, all of the groups began to move to the areas with more shrubs, but at the same time, the p_canopy values increased as well, while the p_ground values dropped (Fig. 4). This indicates that as the winter progressed, the favoring turned towards areas that were able to offer both food (p_shrub) and

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

shelter (p_canopy). The shelter then means that areas with high and dense canopies have also less snow in them, which makes the areas easier to move in. Females with calves used these types of forest area most heavily (p_canopy – Fig. 4), which is logical since the deep snow would be even harder for the calves to move in. Dussault et al. (2004) found that moose increased the use of stands that provided shelter against snow, but these were located along edges of stands with abundant food. Our results support this – the choice of habitat during midwinter seemed to be a mix landscape of both food and shelter.

5. Conclusions We showed here that ALS data could be used to analyze moose behavior and habitat preferences (in terms of forest structure) during different seasons and during important events, such as calving. However, certain weaknesses should be addressed before considering the practical implications for forestry. We did not have knowledge about tree species, which could be considered by additional data sources (e.g. stand register data or aerial images). Also, the results would probably be different if the study had been done in an area with slopes and hills, and in an area with fewer agricultural fields and peat lands. This would also mean that we should increase the amount of study animals, which was only 15 here. Thus, the next step would be to expand the study to include different landscapes, as well as to consider tree species so that it is possible to assess the habitats more thoroughly, not just look at its forest structure. Lindenmayer, Margules, and Botkin (2000) suggested the use of structure-based indicators, such as stand structural complexity, as potential indicators of biodiversity. Current research has shown that ALS is an excellent tool to study forest structure. Work et al. (2011) concluded that “the utility of lidar will be borne out when the force of this data is realized through mechanistic hypotheses related to habitat requirements of plants and animals”. Our approach gave information about the habitat requirements of moose in terms of forest structure. Today, as ALS coverage increases rapidly, there are great possibilities for mapping the abundance of ecologically important attributes. Based on the experiences from other studies and from the ones presented here, we can say that there is no doubt about the usefulness of this data in wildlife research. The data that valuable information about the target areas' characteristics and about its suitability to wildlife, for instance about its suitability for moose calving site. These kinds of new applications provide great added value to ALS datasets and can also give new knowledge about the ecology and habitat use of different species.

Acknowledgments We would like to thank the Alfred Kordelin Foundation (http:// www.kordelin.fi) for supporting the work carried out in this study. We would also like to thank the three anonymous reviewers. Your comments surely made this article better.

References Addison, E.M., Smith, J., McLaughlin, R., Fraser, D.J.H., & Loachim, D.G. (1990). Calving sites of moose in central Ontario. Alces, 26, 142–153. Bailey, T.N., & Bangs, E.E. (1980). Moose calving areas and use of the Kenai National Wildlife Refuge, Alaska. Proceedings of the North American Moose Conference Workshop, 16, 289–313. Bergstrom, R., & Hjeljord, O. (1987). Moose and vegetation interactions in northwestern Europe and Poland. Swedish Wildlife Research Supplement, 1, 213–228. Bertram, M.R., & Vivion, M.T. (2002). Moose mortality in interior Alaska. Journal of Wildlife Management, 66(3), 747–756. Bogomolova, E.M., & Kurochkin, Y.A. (2002). Parturition activity of moose. Alces Supplement, 2, 27–31. Bowyer, T.B., Kie, J.G., & Ballenberghe, V. (1998). Timing and synchrony of parturition in Alaskan moose: Long term versus proximal effects of climate. Journal of Mammalogy, 79, 1332–1344.

9

Cassing, G., Greenber, L.A., & Grzegorz, M. (2006). Moose (Alces alces) browsing in young forest stands in central Sweden: A multiscale perspective. Scandinavian Journal of Forest Research, 21, 221–230. Cederlund, G., Ljungqvist, H., Markgren, G., & Stalfelt, F. (1980). Foods of moose and roedeer at Grimso in Central Sweden — Results of rumen content analyses. Swedish Wildlife Research, 11, 169–247. Cederlund, G., & Sand, H. (1994). Home-range size in relation to age and sex in moose. Journal of Mammalogy, 75, 1005–1012. Cederlund, G.N., Sandegren, F., & Larsson, K. (1987). Summer movement of female moose and dispersal of their offspring. Journal of Wildlife Management, 51(2), 342–352. Chekchak, T., Courtois, R., Ouellet, J.P., Breton, L., & St-Onge, S. (1998). Characteristics of moose (Alces alces) calving sites. Canadian Journal of Zoology, 76, 1663–1670. Clutton-Brock, T.H., Stevenson, I.R., Marrow, P., MacColl, A.D., Houston, A.I., & McNamara, J.M. (1996). Population fluctuations, reproductive costs and life-history tactics in female Soay sheep. Journal of Animal Ecology, 65, 675–689. Coops, N.C., Duffe, J., & Koot, C. (2010). Assessing the utility of lidar remote sensing technology to identify mule deer winter habitat. Canadian Journal of Remote Sensing, 36(2), 81–88. Davies, A.B., & Asner, G.P. (2014). Advances in animal ecology from 3D-LiDAR ecosystem mapping. Trends in Ecology and Evolution, 29(12), 681–691. Dettki, H., Löfstrand, R., & Edenius, L. (2003). Modeling habitat suitability for moose in coastal northern Sweden: Empirical vs. process-oriented approaches. Ambio, 32(8), 549–556. Dussault, C., Ouellet, J.P., Courtois, R., Huot, J., Breton, L., & Larochelle, J. (2004). Behavioural responses of moose to thermal conditions in the boreal forest. Ecoscience, 11, 321–328. Flaherty, S. (2013). Red squirrel habitat mapping using remote sensing. University of Edinburgh (PhD thesis). Gaillard, J.M., Festa-Bianchet, M., Yoccoz, N.G., Loison, A., & Tiogo, C. (2000). Temporal variation in fitness components and population dynamics of large herbivores. Annual Review of Ecology and Systematics, 31, 367–393. Goetz, S.J., Steinberg, D., Betts, M.G., Holmes, R.T., Doran, P.J., Dubayah, R., et al. (2010). Lidar remote sensing variables predict breeding habitat of a neotropical migrant bird. Ecology, 91, 1569–1576. Graf, R.F., Mathys, L., & Bollmann, K. (2009). Habitat assessment for forest dwelling species using LiDAR remote sensing: Capercaillie in the Alps. Forest Ecology and Management, 257, 160–167. Haydn, A. (2012). Calving site selection by moose (Alces alces) along a latitudinal gradient in Sweden. Vienna, Austria: University of Natural Resources and Life Sciences (BOKU) (Master's thesis). Heikkilä, R., & Härkönen, S. (1993). Moose (Alces alces L.) browsing in young Scots pine stands in relation to the characteristics of their winter habitats. Silva Fennica, 27(2), 127–143. Hill, R., Hinsley, S.A., & Broughton, R.K. (2014). Assessing habitats and organism–habitat relationships by airborne laser scanning. In M. Maltamo (Eds.), Forestry Applications of Airborne Laser Scanning (pp. 335–356). Netherlands: Springer. Hjeljord, O., Hövik, N., & Pedersen, H.B. (1990). Choice of feeding sites by moose during summer, the influence of forest structure and plant phenology. Ecography, 13, 281–292. Keech, M.A., Bowyer, R.T., Ver Hoef, J.M., Boertje, R.D., Dale, B.W., & Stephenson, T.R. (2000). Life-history consequences of maternal condition in Alaskan moose. Journal of Wildlife Management, 64, 450–462. Korhonen, K., Ihalainen, A., Miina, J., Saksa, T., & Viiri, H. (2010). Metsänuudistamisen tila Suomessa VMI10:n aineistojen perusteella. Metsätieteen Aikakauskirja, 4(2010), 425–478. Langley, M.A., & Pletscher, D.H. (1994). Calving areas of moose in northwestern Minnesota and southeastern British Columbia. Alces, 30, 127–135. Lavsund, S. (1987). Moose relationships to forestry in Finland, Norway and Sweden. Swedish Wildlife Research, Supplement, 1, 229–244. Lindenmayer, D.B., Margules, C.R., & Botkin, D.B. (2000). Indicators of biodiversity for ecologically sustainable forest management. Conservation Biology, 14(4), 941–950. Lone, K., Loe, L.E., Gobakken, T., Linnel, J.D.C., Odden, J., Remmen, J., et al. (2014a). Living and dying in a multi-predator landscape of fear: Roe deer are squeezed by contrasting pattern of predation risk imposed by lynx and humans. Oikos, http://dx.doi.org/ 10.1111/j.1600-0706.2013.00938.x Lone, K., van Beest, F.M., Mysterud, A., Gobakken, T., Milner, J.M., Ruud, H. -P., et al. (2014b). Improving broad scale forage mapping and habitat selection analyses with airborne laser scanning: The case of moose. Ecosphere, 5(11), 1–22. Lowe, S.J., Patterson, B.R., & Schaefer, J.A. (2010). Lack of behavioral responses of moose (Alces alces) to high ambient temperatures near the southern periphery of their range. Canadian Journal of Zoology, 88, 1032–1041. MacArthur, R.H., & MacArthur, J.W. (1961). On bird species diversity. Ecology, 42, 594–598. Markgren, G. (1974). The moose in Fennoscandia. Le Naturaliste Canadien, 101, 185–194. McInnes, P.F., Naiman, R.J., Pastor, J., & Cohen, Y. (1992). Effects of moose browsing on vegetation and litter of the boreal forest, Isle Royal, Michigan USA. Ecology, 73(6), 2059–2075. Melin, M., Matala, J., Mehtätalo, L., Tiilikainen, R., Tikkanen, O. -P., Maltamo, M., et al. (2014). Moose (Alces alces) reacts to thermal stress by utilising thermal shelters in boreal forests — An analysis based on airborne laser scanning of the canopy structure at moose locations. Global Change Biology, 20, 1115–1125. Melin, M., Packalen, P., Matala, J., Mehtätalo, L., & Pusenius, J. (2013). Assessing and modelling moose (Alces alces) habitats with airborne laser scanning data. International Journal of Applied Earth Observations and Geoinformation, 23, 389–396. METLA (2013). Finnish National Forest Inventory. http://www.metla.fi/ohjelma/vmi/vmimoni-en.htm. Last accessed 01/2013. Miller, B.K., & Litvaitis, J.A. (1992). Habitat segregation by moose in boreal forest ecotone. Acta Theriologica, 37(1–2), 41–50.

.

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

10

M. Melin et al. / Remote Sensing of Environment xxx (2015) xxx–xxx

Müller, J., & Vierling, K. (2014). Assessing biodiversity by airborne laser scanning. In M. Maltamo (Eds.), Forestry Applications of Airborne Laser Scanning (pp. 357–374). Netherlands: Springer. Nikula, A., Heikkinen, S., & Helle, E. (2004). Habitat selection of adult moose Alces alces at two spatial scales in central Finland. Wildlife Biology, 10, 121–135. Palminteri, S., Powell, G.V.N., Asner, G.P., & Peres, C.A. (2012). LiDAR measurements of canopy structure predict spatial distribution of a tropical mature forest primate. Remote Sensing of Environment, 127, 98–105. Poole, K.G., Serrouya, R., & Stuart-Smith, K. (2007). Moose calving strategies in interior Montane ecosystems. Journal of Mammalogy, 88(1), 139–150. R Development Core Team (2015). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. RKTL (2011). Data about Finland's moose populations. Available at: http://www.rktl.fi/ riista/riistavarat/hirvi_vuonna.html (Site visited 1/2013) Saether, B.E., Andersen, R., Hjeljord, O., & Heim, M. (1996). Ecological correlates of regional variation in life history of the moose Alces alces. Ecology, 77(5), 1493–1500. Shepard, D. (1968). A two-dimensional interpolation function for irregularly-spaced data. Proceedings of the 1968 ACM National Conference (pp. 517–524). New York, NY: ACM.

Singh, N.J., Börger, L., Dettki, H., Bunnefeld, N., & Ericsson, G. (2012). From migration to nomadism: Movement variability in a northern ungulate across its latitudinal range. Ecological Applications, 22(7), 2007–2020. Trembley, J.P., Solberg, E.J., Sæther, B.E., & Heim, M. (2007). Fidelity to calving areas in moose (Alces alces) in the absence of natural predators. Canadian Journal of Zoology, 85, 902–908. van Beest, F., Rivrud, I.M., Loe, L.E., Milner, J.M., & Mysterud, A. (2011). What determines variation in home range size across spatiotemporal scales in a large browsing herbivore? Journal of Animal Ecology, http://dx.doi.org/10.1111/j.1365-2656.2011.01829.x Vierling, K.T., Vierling, L.A., Gould, W.A., Martinuzzi, S., & Clawges, R.M. (2008). Lidar: Shedding new light on habitat characterization and modeling. Frontiers in Ecology and the Environment, 6(2), 90–98. Work, T. T., St Onge, B., & Jacobs, J. M. (2011). Response of female beetles to LIDAR derived topographic variables in Eastern boreal mixedwood forests (Coleoptera, Carabidae). ZooKeys, 147, 623–639. WRAM (Wireless Remote Animal Monitoring) (2013). http://www.slu.se/WRAM/ (Last accessed 21.9.2010).

Please cite this article as: Melin, M., et al., Ecological dimensions of airborne laser scanning — Analyzing the role of forest structure in moose habitat use within a year, Remote Sensing of Environment (2015), http://dx.doi.org/10.1016/j.rse.2015.07.025

.