Skip to main content
Download PDF
Download ePub

Remote sensing reveals the role of forage quality and quantity for summer habitat use in red deer

Movement Ecology volume 12, Article number: 80 (2024) Cite this article

Abstract

Background

The habitat use of wild ungulates is determined by forage availability, but also the avoidance of predation and human disturbance. They should apply foraging strategies that provide the most energy at the lowest cost. However, due to data limitations at the scale of movement trajectories, it is not clear to what extent even well-studied species such as red deer (Cervus elaphus) trade-off between forage quality and quantity, especially in heterogeneous alpine habitats characterized by short vegetation periods.

Methods

We used remote sensing data to derive spatially continuous forage quality and quantity information. To predict relative nitrogen (i.e. forage quality) and biomass (i.e. forage quantity), we related field data to predictor variables derived from Sentinel-2 satellite data. In particular, our approach employed random forest regression algorithms, integrating various remote sensing variables such as reflectance values, vegetation indices and optical traits derived from a radiative transfer model. We combined these forage characteristics with variables representing human activity, and applied integrated step selection functions to estimate sex-specific summer habitat selection of red deer in open habitats within and around the Swiss National Park, an alpine Strict Nature Reserve.

Results

The combination of vegetation indices and optical traits greatly improved predictive power in both the biomass (R2 = 0.60, Root mean square error (RMSE) = 88.55 g/m2) and relative nitrogen models (R2 = 0.34, RMSE = 0.28%). Both female and male red deer selected more strongly for biomass (estimate = 0.672 ± 0.059 SE for normalised values for females, and 0.507 ± 0.061 for males) than relative nitrogen (estimate = 0.124 ± 0.062 for females, and 0.161 ± 0.061 for males, respectively). Females showed higher levels of use of the Swiss National Park.

Conclusions

Red deer in summer habitats select forage quantity over quality with little difference between sexes. Females respond more strongly to human activities and thus prefer the Swiss National Park. Our results demonstrate the capability of satellite data to estimate forage quality and quantity separately for movement ecology studies, going beyond the exclusive use of conventional vegetation indices.

Background

Animals face a trade-off between foraging and safety needs [33, 63]. The latter drive prey species to modify their habitat use in response to potential predators [51], with humans usually also perceived as a predation risk due to hunting [27, 68]. Diurnal behaviour can be altered to avoid humans during daytime [1, 20, 25, 33], and protected areas without hunting are preferred over non-protected areas [34, 58]. Foraging needs result in strategies that provide the most energy at the lowest cost, under the condition that animals can move freely and have sufficient information about their environment [71, 82]. Another trade-off exists between forage quality and quantity [28, 82]. Herbivore species specialize along a continuum ranging from the extremes of selectively feeding on energy-rich buds to mainly consuming mature, high-fiber and low-moisture leaves and wood [41, 54]. Based on comparisons of their digestive systems, ruminants have thus been classified into a continuum of three main feeding types: a) concentrate selectors (browsers), b) roughage eaters (grazers), and c) intermediate, opportunistic feeders [42]. Red deer (Cervus elaphus) as intermediate feeders [53] generally meet their high metabolic requirements by high daily forage intake which is matched with availability and quality in the summer diet [4, 98]. Results from studies that tested red deer preference on quality versus quantity simultaneously with separate variables have provided equivocal results in that they found either higher preferences for quantity [78, 81] or for quality [97]. In addition, within-population differences may arise from different seasonal selection patterns. Hence, analyses on tracking phenological green-up in spring considered varying selection between residents and migrants, with migrating individuals gaining better access to high-quality forage than residents [40, 81]. In closely related elk/wapiti (Cervus canadensis), residents selected for higher quantity, but lower quality, and migrating animals for intermediate quantity, but higher quality, respectively [40]. However, it is not clear to what extent red deer trade-off between forage quality and quantity, especially in heterogeneous alpine habitats characterized by short vegetation periods. This is due to two main reasons: because a) suitable and available variables specifically to describe quality versus quantity are difficult to approximate and can only be derived with substantial field effort, and b) studies have often focused on either quality or quantity (but not both), as they are difficult to disentangle [44, 50, 93].

A disadvantage of studies based on field sampling of biomass and nitrogen is that it can merely provide a snapshot in time and over a limited spatial scale. Various remote sensing instruments measuring the reflected sunlight at different wavelengths with varying spatial resolution, revisit times, and number of spectral bands have instead been used in movement ecology of ungulates (Additional file 1: Supplementary Table 1). The spectral information of space- or airborne data has commonly been ground-truthed with in-situ data, i.e. field samples such as the analysis of fecal samples, plant species and biomass [37, 40, 46, 69] to predict the respective variables over larger areas. Former studies predominantly relied on vegetation indices (VIs) related to green vegetation, particularly the Normalized Difference Vegetation Index (NDVI (Additional file 1: Supplementary Table 1); [8, 21, 66, 67]). The popularity of NDVI stemmed from the historical constraints of satellite data, which lacked bands in the red-edge area of the electromagnetic spectrum, such as the thematic mapper (Landsat missions), or had only the red and near-infrared bands available at higher spatial resolutions, e.g., Moderate-resolution Imaging Spectroradiometer (MODIS).

NDVI has proven a suitable proxy for green-up selection of, e.g., migrating elk/wapiti and red deer [9, 40, 64, 81], making it a suitable biomarker of the ecosystem: to approximate ground vegetation biomass [12, 38], but also primary productivity [3, 91], to correlate with fecal crude protein [37], chlorophyll concentration [17, 47] or dry matter digestibility of forage [30], and vegetation structure [44]. However, as NDVI is related to many vegetation properties [44], this method is less suitable for differentiating to which extent red deer prioritize forage quality versus quantity.

Imaging spectrometers, commonly referred to as hyperspectral sensors, measuring the reflected sunlight from a surface in hundreds of narrow spectral bands, have the potential to deliver both relative plant nitrogen content and plant biomass [79, 89]. However, they are currently limited to airborne sensors or satellite precursor missions (e.g., [15, 19]), thus lacking the necessary revisit time needed to study habitat use of wild ungulates over the entire plant growing season. Currently, an operational mission offering the necessary spatial and temporal resolution is the European Space Agency (ESA) Copernicus Sentinel-2 mission, providing reflectance data with 10 spectral bands at a spatial resolution of 10 to 20 m, and a revisit time of 5 days [23]. Despite the lower number of bands compared to an imaging spectrometer, Sentinel-2 sensors have proven their capabilities to derive forage quality and quantity [74]. The spectral bands of Sentinel-2 sensors allow not only to derive different VIs, but also to use physical-based radiative transfer models (RTMs) to estimate biophysical and biochemical plant traits (further referred to as optical traits) likely associated with forage quality and quantity [70, 76]. RTMs use physical laws to describe surface reflectance as a function of canopy, leaf and soil traits [43].

Determining whether to use surface reflectance, VIs, optical traits, or their combination for estimating forage quantity and quality is not trivial, as each approach has its advantages and disadvantages. For example, using reflectance values allows retaining the full spectral information and avoiding losing potentially valuable information, although raw reflectance may be less directly related to vegetation properties. VIs, while being easily derivable and robust in reducing artefacts arising from atmospheric correction processes [16], exhibit limitations like saturation at high biomass values [61] and a lack of specificity for directly estimating forage quality and quantity. The use of optical traits can enhance prediction transferability and account for viewing geometry, but their retrieval is ill-posed, meaning that there can be multiple solutions for a single spectral signature,dependent on ancillary data, and model assumptions are violated in structurally complex environments [77].

In this study, we used Sentinel-2 reflectance data, from which we derived VIs and optical traits to model dynamic parameters of forage quality and quantity at high spatial and temporal resolution. We thereby quantified the importance of surface reflectance, VIs and optical traits in predicting forage quality and quantity. Finally, we tested for the importance of forage quality and quantity on red deer habitat selection in summer using radio-collar GPS data and also accounted for the effects of topography and human disturbance. We expected red deer to select primarily for forage quantity followed by quality [78, 81] with males showing stronger preferences for quality than females [29]. We further expected both sexes to avoid humans [1, 20, 25, 33, 81]. Since the study is conducted in the region of the Swiss National Park (SNP), a protected area with strict restrictions on human use, we also predicted red deer to prefer the SNP over non-protected surroundings [34, 58].

Additional file 1: Supplementary Table 1 (letter landscape page as additional file: table_1_landscape_page.docx).

Methods

Study area

The study area is located in inner alpine valleys of eastern Switzerland (Fig. 1) including the SNP and its surrounding areas in the canton of Grisons (Switzerland), Tyrol (Austria), the Autonomous province of Bolzano – South Tyrol (Italy) and the province of Sondrio (Italy). Elevation ranges from 1000 to 3200 m a.s.l. and the tree line is at approximately 2200 m a.s.l. [36]. The climate is dry and cool, with an annual mean precipitation of 825 mm between the two weather stations Buffalora (1971 m a.s.l.) and Scuol (1303 m a.s.l.), and a mean summer temperature of 13.2 °C (monthly means from June to August) in the study years 2017 to 2021 [55]. Long winters are characteristic (~ 154 days with snow cover between October and May). Villages are located in the valley bottoms between 1000 and 1700 m a.s.l. The main agricultural land use consists of pastures with cattle and/or sheep. With 554′000 overnight stays from May to October [6] there are high levels of summer tourism [59, 94]. In the Swiss part of the study area winter feeding is prohibited. Red deer are hunted outside of hunting ban areas in September; in the neighboring countries Austria and Italy, the open season ranges from May to December. Where hunting is allowed, it is conducted during daytime only. The SNP (170 km2) provides year-round protection to the animals. In this International Union for Conservation of Nature (IUCN) category Ia protected area, i.e. a Strict Nature Reserve, all human use is prohibited except for scientific studies and hiking on trails.

Fig. 1
figure 1

Map of the study area with red deer GPS locations of 45 adult females and 21 males from June to August 2017 to 2021. To visualise the intensity of use, we calculated revisitation information within a radius of 500 m (visually determined as the optimum radius) ranging from a minimum of one revisitation (blue) to a maximum of 479 revisitations (yellow) based on the recurse function [13] in R. The border of the Swiss National Park is represented by red lines and country borders by grey lines

Full size image

Red deer data

From 2017 to 2021, wildlife officials of the canton of Grisons and the SNP captured 70 adult red deer either using dart guns (n = 63) or corral traps (n = 7) in 12 spatially separated marking areas to account for potential migration in groups. All captures were conducted under permit by the federal and the cantonal governments (GR2017-12F, GR2020-08F), and in compliance with Swiss animal welfare laws. Telemetry collars (VECTRONIC Aerospace GmbH, Berlin, Germany) recorded GPS locations every 3 h over a period of 1 to 3 years. We removed inaccurate locations following Bjørneraas et al. [10] and excluded individuals with less than 80% fix rate success per month. Since we only analysed locations in open summer habitats, we filtered location data to June to August and ESA WorldCover raster [96] categories grassland, cropland, bare/sparse vegetation, and moss/lichen (Fig. 1). This resulted in a sample size of 45 adult females and 21 males with similar age distributions between sexes (Additional file 1: Table A1). For day and night comparisons, we defined daytime as the time between sunrise and sunset using the R package suncalc [85], and the opposite, including civil twilight, as nighttime.

Table 1 AICC based ranking of glmmTMBs per sex explaining model selection
Full size table

Explanatory variables

Biomass and relative nitrogen

We estimated absolute grassland canopy foliar biomass (hereafter referred to as biomass) and relative canopy nitrogen content (hereafter referred to as relative nitrogen) for each recorded GPS location, using two separate random forest regression models in combination with variables derived from freely available satellite images.

Ground reference data

To train the random forest models, we used field measurements of biomass and relative nitrogen (distribution of values in Additional file 2: Figure A2) collected during the growing season in 322 plots for the SNP (2011 to 2013; [78]), as well as biomass collected in two additional studies across 52 plots in the SNP and its surroundings (2016 to 2018; [76, 77]). From 2011 to 2016, biomass in g/m2 was clipped 1 cm above the ground on 1 m2 per plot followed by drying at 65 °C and weighing. In 2018, dry biomass was weighed for an area of 0.2 m2 and scaled to 1 m2 (i.e., divided by 0.2). Chemical analysis for relative nitrogen was conducted in a previous study on one third of the samples with standard laboratory methods (TruSpec CN analyser Leco Corp., St Joseph, MI, USA; Fibre Analyser 200, Ankom Technology, NY, USA). For the remainder of the samples, a laboratory infrared reflectance spectrometer was used to predict relative nitrogen with a predictive accuracy of R2 = 0.93 (detailed description of the method can be found in Schweiger et al. [78]). All plots were georeferenced with a high-precision Global Navigation Satellite System (GNSS) receiver with an expected accuracy of < 0.10 m.

Fig. 2
figure 2

Distribution of distances travelled per day (a) and home range sizes (b) by sex, and elevations covered (c) and step lengths (d) divided by sex and inside/outside the Swiss National Park (SNP)

Full size image
Predictor variables

We used three types of remotely sensed predictor variables in our random forest regression models: (1) ten Sentinel-2 spectral band reflectances [31], (2) VIs calculated from the Sentinel-2 spectral bands, namely NDVI, MERIS Terrestrial Chlorophyll Index (MTCI), Triangular Greenness Index (TGI) and Cellulose Absorption Index (CAI), and (3) optical traits (i.e. canopy structure and plant leaf traits) including leaf area index (LAI), chlorophyll content (CHL) and equivalent water thickness (EWT) obtained through the inversion of the PROSAIL RTM [45] with Sentinel-2 data. We further added dry matter (Cm) and relative nitrogen content (PROT_per), both calculated from the PROSAIL derived protein content (PROT) and non-protein carbon based constituents (CBC) to the biomass and relative nitrogen models, respectively.

Besides the commonly used NDVI, the three additional VIs were chosen since they cover different spectral regions (i.e., visible, near infrared and short-wave infrared, respectively) and provide complementary information relevant to biomass and nitrogen estimations. Specifically, MTCI is sensitive to nitrogen content in grasslands [18], TGI responds to total pigment content, which often shows a strong positive correlation with nitrogen and biomass content [86] and CAI is useful for detecting dry and non-photosynthetic vegetation contributing to the total biomass content [88].

Similarly, we chose PROSAIL optical traits related to the variables of interest: the product of Cm and LAI can serve as an estimate of biomass [72], PROT is equivalent to the nitrogen content [95], and the choice of CHL and EWT aligns with that of TGI and CAI.

To match the remote sensing data with field measurements collected previous to the launch of Sentinel-2, we resampled hyperspectral airborne surface reflectance data, acquired within days of the field data collection, to the Sentinel-2 spectral resolution. Detailed information about the derivation of all predictor variables for the GPS locations and validation data is included in Additional file 3.

Model selection

We used the R package randomForest [52] to train separate random forest regression models for biomass and relative nitrogen. To select the most important predictor variables and identify optimal model parametrization (i.e. number of trees, minimum and maximum size of terminal nodes, and number of variables randomly sampled as candidates at each split), we used the recursive feature elimination (RFE) algorithm of the R package caret [49] followed by a hyperparameter optimization. Both RFE and hyperparameter optimization were performed using a fivefold cross-validation with three repeats against field-measured biomass and relative nitrogen. In doing so, we cross-validated 3456 different combinations of hyperparameters, each three times. We selected the model with the highest coefficient of determination (R2) and also reported the root mean square error (RMSE) for the best models.

Variable importance

We calculated the variable importance (i.e., increase in RMSE) of groups of predictor variables for the best random forest models using a permutation-based strategy [83]. We conducted two analyses: the first involved grouping variables with high correlation (> 0.75; Additional file 4: Figure A4), and the second involved grouping variables that belong to the same predictor type (i.e., reflectance, VIs, or optical traits). We permuted the variables from a specific predictor group for the cross-validation samples and fed them into the best random forest model to recompute the RMSE. This process was repeated 1000 times for each different group of predictors. Subsequently, we averaged the increase in RMSE compared to the baseline RMSE values across the 1000 runs.

Prediction of GPS locations and outlier removal

We assigned each GPS location to the surface reflectance data of the temporally closest Sentinel-2 image, resulting in an average discrepancy of 7.7 days between the two datasets and a standard deviation of 9.34 days. Subsequently, we used the best-performing models to predict biomass and relative nitrogen from the Sentinel-2 data for all GPS locations. Machine learning models such as random forest regression models have difficulties extrapolating predictions to data that differ from the training data. However, the dissimilarity index (DI) as proposed by Meyer & Pebesma [57] can be used to quantify the similarity between a data point to be predicted and the training data. We used the R package CAST [56] to calculate the DI for all GPS locations and removed DI outliers based on the interquartile range, thus only including values lower than \({Q}_{3}+1.5*\left({Q}_{3}- {Q}_{1}\right)\), with Q1 and Q3 representing the first and the third quartile, respectively.

Further habitat variables

Close proximity to forest, as well as avoiding trails, are considered avoidance responses to humans [81]. The same applies to steep slopes, as human activity tends to be higher in flat terrain. Based on the input digital elevation model [65] we calculated slope using the terrain function of the R package raster [24]. We extracted the category tree cover from ESA WorldCover [96] in ArcGIS Pro (version 3.0.3, ESRI) and then calculated path distance to forest for all red deer locations (i.e. of all ESA WorldCover categories).

For the Swiss parts of the study area, we extracted trails from the streets layer of the Swiss Topographic Landscape Model [84] by filtering for trail categories of up to 2 m in width. For the Austrian and Italian parts of the study area, we extracted Open Street Map data using Protomaps (https://protomaps.com/downloads/osm/18343d18-d905-440e-9b45-6bcc41608e16). We post-edited lacking trails in ArcGIS Pro with Swisstopo’s reference map 1:25′000. Combined with the trails in Switzerland, we also calculated path distance of red deer GPS locations to trails.

Step length

Step length, i.e. the distance between two consecutive GPS locations, serves as a measure of movement intensity. It was calculated while processing integrated step selection functions (iSSFs; see section Modelling habitat selection). The inclusion of step length into the model reduces potential biases due to the variability in red deer’s individual movement behaviour [26].

Modelling habitat selection

iSSFs [5] were applied to estimate habitat preferences. In a first step, we used the R package amt [80] to generate individual trajectories. Based on the distribution of step lengths and turning angles, we estimated 25 random locations per observed location. Since we had to restrict the analyses of GPS locations to open habitats for reliable biomass and relative nitrogen values, we restricted our dataset to steps in open habitats with at least three random locations following Sigrist et al. [81]. We scaled all explanatory variables to mean zero and standard deviation 0.5 combining the female and male datasets [32], and then ran sex-specific generalized linear mixed models using Template Model Builder (glmmTMB,[60]). Individual was included as a random effect to correct for individual-specific variation in habitat selection and sample size. Model selection was applied using the R package MuMIn [7],Table 1). All analyses were conducted in R version 4.2.3 [73].

Results

Movements of red deer

Divided into the categories of the ESA WorldCover raster [96], 74% of red deer summer locations in open habitats were in grassland, 24% in moss and lichen, 2% in bare / sparse vegetation and 0.07% in cropland. Daily distances travelled, home range sizes (100% Multiple Convex Polygons) and step lengths inside and outside the SNP, respectively, were similar for both sexes (Fig. 2). Elevations covered outside the SNP were also similar, but males inside the SNP stayed at higher elevations than females. Since step lengths from movements between inside and outside the SNP accounted for only 2% they had little influence on the results for either sex.

Modelling biomass and relative nitrogen

We observed a strong performance of the best biomass model (R2 = 0.60, RMSE = 88.55 g/ m2, ntree = 1000, nodesize = 5, maxnodes = 25, mtry = 8). Tuning the hyperparameters resulted in only slight changes to the outcomes, as we found a mean R2 of 0.578 with a standard deviation of ± 0.01 across all models. Based on the RFE analysis, we found eight variables to be sufficient in predicting biomass (Fig. 3a), with MTCI resulting as the most important variable followed by the group including NDVI, CAI and LAI. The optical trait CHL was also an important predictor.

Fig. 3
figure 3

Average variable importance and their standard deviation over 1000 permutations of the random forest regression models for biomass grouped by correlated variables (a) and by variable type (b), and accordingly for relative nitrogen in (c) and (d). We included three types of variables: reflectance values of Sentinel-2 bands, represented by their central wavelength in nanometers (green), and derived from them vegetation indices (orange) and optical traits derived using the radiative transfer model PROSAIL (blue)

Full size image

Moderate predictive accuracies were found for relative nitrogen (best model: R2 = 0.34, RMSE = 0.28%, ntree = 500, nodesize = 7, maxnodes = 23, mtry = 18) with an average R2 of 0.31 and standard deviation of ± 0.01 across all models. All 18 variables were retained for nitrogen prediction based on the RFE analysis (Fig. 3b). We found the group including spectral bands in the visible part of the spectrum as well as NDVI, CAI and LAI, as the most important variable, followed by CHL (Fig. 3c).

VIs were the most important variables in both biomass and relative nitrogen models. However, optical traits also contributed additional value, particularly in the biomass models (Fig. 3b). While spectral bands appeared to be significant in the models for relative nitrogen, they did not provide additional value compared to VIs and optical traits, and were marginal in the biomass models (Fig. 3d).

Habitat selection of red deer

The best supported model for female red deer in open summer habitats corresponded to the full model, while the best model for males excluded distance to trails and slope (Table 1). Red deer of both sexes selected habitats with high biomass and high relative nitrogen (Fig. 4, Table 2), with biomass having a greater effect than relative nitrogen. The comparison between the sexes showed that females tended to select habitats more strongly for biomass than males, but males selected habitats somewhat more strongly for relative nitrogen than females. However, a statement on significant differences between the sexes is not possible because we calculated sex-specific models.

Fig. 4
figure 4

Habitat selection by red deer in summer with respect to biomass, relative nitrogen, distance to forest, distance to trails and slope, as well as step length, with 95% confidence intervals. f[u]/f[a] refers to the frequency ratio between used and available locations. Values > 0 indicate preference, values < 0 avoidance. Models were run separately per sex. Note the different scales in the plots. Only significant results are shown

Full size image
Table 2 Results of the best supported glmmTMBs for female and male red deer in open summer habitats. SE = standard error, CI = confidence interval. Significance levels for p values * < 0.05, ** < 0.01, *** < 0.001
Full size table

Regarding the variables indicating safety needs, females preferred to stay close to the forest and to trails, and in flat terrain. Males selected for short distances to forest. Both sexes showed high step lengths indicating more movement, with females moving longer distances than males.

For both sexes, GPS locations in open summer habitats were relatively evenly distributed between day and night (Additional file 5: Table A5): For females, 47% of the locations were recorded during the day and 51% for males. While 82% of the females' locations were within the SNP, this applied to only 46% of the males' locations (Fig. 5a). Of these, 89% of female locations were inside the SNP during the day and 75% at night, while for males 59% were inside the SNP during the day and 32% at night (Fig. 5b).

Fig. 5
figure 5

Proportions of summer GPS locations per sex inside (green) and outside the Swiss National Park (SNP; blue) (a), and the respective proportions divided into day and night (b)

Full size image

Discussion

Using iSSFs, we identified summer habitat preferences of free-ranging red deer in a heterogeneous open alpine landscape. We detected effects of variables indicating human disturbance and vegetation characteristics estimated based on remote sensing and field samples (Fig. 4, Table 2). The method combined simultaneous estimation of individual movement and resource selection parameters and thus enabled a likelihood-based inference of resource selection within a mechanistic movement model [5].

Habitat selection of red deer

Red deer face a trade-off between forage quality and quantity and are under pressure to employ foraging strategies that provide the most energy for the lowest cost [28, 71, 82]. Several studies have shown that in seasonal and predictable habitats, red deer prefer higher altitudes in summer than in winter due to greater availability of high-quality forage [9, 30, 40, 64, 69, 81, 97]. We found that red deer selected more strongly for biomass than for relative nitrogen (Fig. 4, Table 2). Thus, in our study they preferred forage quantity over quality. This is largely in line with results from Hebblewhite et al. (40, 78]. The comparison with Schweiger et al. [78] who examined forage preferences earlier in the season also indicates that preferences for biomass over relative nitrogen in our study were not caused by a beginning of brown-down in August or by early depletion of nutrient-rich forage. Different findings by Zweifel-Schielly et al. [97] may be explained by different elevational ranges between their and our study: Their study was conducted down to a minimum elevation of 470 m a.s.l. (versus 1000 m a.s.l. in ours), and also included nutrient-poor vegetation on the forest floor. As expected, differences in sex-specific selection of forage quality and quantity [9, 11] were weak. However, the fact that females in summer tended to select patches with higher quantity but lower quality than males was in line with patterns found for spring and autumn diets [29], which is not unusual for intermediate feeders [62].

Since our study area is characterized by strict protection regulations inside the SNP, where no human activity is permitted except for research and hiking on trails, it provides a good basis to analyse the trade-off between foraging and safety needs of red deer [33, 58, 63]. Predators were absent during the study period, except for one single resident female wolf in part of the study area in the years 2017 to 2022. Red deer preference of areas close to forest may be due to reasons of thermoregulation (e.g. [2], for chamois) or indicates their need for safety [81], which we have identified for both sexes. As previous studies showed, female red deer, especially those with offspring, choose safe habitats even at the expense of forage quality [9, 11]. We know from summer counts inside the SNP that 48% of females (two years and older) had calves over the study period (SNP, unpublished data) and assume that this rate also applies to tagged animals. The increased frequency of GPS locations of females within the SNP during the day (89%) compared to night-time (75%) indicates the overriding need for safety (Fig. 5, Additional file 5: Table A5). Males also showed this pattern, but at a lower level (59% during the day and 32% at night), which indicates clear sexual differences concerning the need for safety, and differences between day and night. Thus, males left the SNP at night about twice as often as females, which could also be interpreted as avoidance of high red deer density. As the SNP is surrounded by alpine pastures [76], which are grazed by livestock (cattle and sheep) and therefore offer lower biomass but higher relative nitrogen, this may have influenced the result of males selecting lower forage quantity but higher quality than females. Our results are applicable to relatively undisturbed summer habitats at higher elevations (> = 1000 m a.s.l.), which are characterised by short vegetation periods and overall low biomass. However, in more intensively cultivated landscapes with high biomass, e.g. agricultural lands, relative preferences may differ.

Sexual differences in habitat selection were also indicated by the different structure of the best-supported models (Table 1): for males, the variables distance to trails and slope were not included. Since avoiding trails and seeking out steep terrain are considered responses to human activity, it is fitting that these should not affect males which appear to be more tolerant to disturbance than females. The preference by females to both proximity to paths and flat terrain (i.e. areas with more human activity), appears to contradict current theory. However, this might be explained by females’ pronounced preference for the SNP, where human activity is strictly limited to trails and visitor flows are predictable: the most frequently used areas were inside the SNP, but in close proximity to trails (Fig. 1).

Modelling biomass and relative nitrogen

We modelled biomass and relative nitrogen at a spatial resolution of 50 m, and thus took into account the accuracy of the GPS locations (11.3 m ± 4.7 m as measured in Schweiger et al. [78]). Compared to that previous study in the same region based on a single remote sensing dataset [78], our study included weekly spectral information from June to August, and a higher sample size of red deer (n = 66 versus n = 2), but a coarser spatial scale (250 m2 versus 36 m2). Despite the different remote sensing approaches used, the predictive accuracies of forage quality and quantity between Sentinel-2 and hyperspectral data [78] were comparable. While our biomass model performed slightly better over multiple years (ΔR2 = 0.03), our relative nitrogen model performed slightly worse (ΔR2 = −0.09). Red deer selection patterns of biomass and relative nitrogen were consistent across both studies. This also corroborates selection patterns of forage quality and quantity in open habitats using a spatial resolution of 100 m2 [81], although the remotely sensed instantaneous rate of green-up as a measure for forage quality and NDVI for quantity was used in their study.

A major limitation of our study and remote sensing of forage quality and quantity in general is predicting relative nitrogen, reflected in the moderate accuracy and complexity, i.e., retention of all predictor variables by our model for relative nitrogen. Similar studies relied on the use of vegetation indices as approximations of forage quality without validating them against ground reference data [9, 58, 64, 81]. Therefore, while our model explained only a third of the variance in relative nitrogen, the inclusion of ground reference data marks a significant improvement and raises questions about the reliability of certain vegetation indices as proxies for relative nitrogen content.

The reflectance of plant communities is mainly determined by the absolute content of nitrogen (g/m2) in leaves and not its concentration relative to other leaf constituents [48]. The problematic use of concentration measurements for nitrogen is enhanced in areas where the vegetation cover is uncorrelated to the nitrogen content, a scenario possible in vegetation-poor areas. Therefore, from a remote sensing point of view, using the absolute nitrogen content should be preferred, despite its strong correlation with biomass (Additional file 6: Table A6, [93]), rendering a comparison of the selection of quality and quantity difficult. To partially mitigate the uncertainties associated with using remotely sensed relative nitrogen, we have used a substantial number of field samples to train the remote sensing models and excluded data points from the habitat selection analysis with ranges unseen during model training. Other mitigation strategies could involve the estimation of vegetation cover and non-photosynthetic vegetation from remote sensing data to actively exclude vegetation-poor areas from the analysis or improve the relative nitrogen models with such additional information. In this regard, future operational and spaceborne hyperspectral sensors [14, 75] promise more accurate predictions of forage quality.

As shown in other studies [35, 74], the inclusion of VIs in predicting forage quality and quantity improved the prediction accuracy. VIs remain the most suitable choice in forage quality and quantity estimations in grasslands due to their simplicity and strong relevance. MTCI, which uses bands from the red-edge and near-infrared region, was particularly useful. Red-edge and near-infrared regions are strongly related to chlorophyll and nitrogen content in plants [18, 22]. Overall, our results suggest to use additional VIs alongside NDVI, particularly given that NDVI experiences saturation issues in densely vegetated areas.

Similar to VIs, we found optical traits to improve the forage quality and quantity prediction. In particular, the biomass models strongly profited from the inclusion of optical traits. PROSAIL is highly sensitive to LAI, which is directly related to biomass [90]. While VIs are commonly used in predictive models, optical traits are rarely used or have shown no added value for grassland quality and quantity indicators [74]. However, Raab et al. [74] used a hybrid inversion of RTMs [92] implemented in the Sentinel Application Platform (SNAP) to derive optical traits, which differed from our look-up table (LUT) inversion. As shown by Hauser et al. [39], an optimised trait retrieval from RTMs with a LUT outperformed the SNAP application. Furthermore, hybrid approaches without further fine tuning can be ineffective in predicting optical traits in heterogeneous grassland. Overall, our results suggest that the inclusion of optical traits derived with an LUT approach increases the predictive power of forage quality and quantity models, but also highlight the challenges of relying solely on them, underpinned by the marginal importance of PROT_per in the relative nitrogen model. Nevertheless, the use of such optical traits next to VIs is motivated by the underlying physical foundation and therefore higher transferability of the models to unseen data [87]. This could especially be the case when a limited amount of field data is available to calibrate the models.

Conclusions

Red deer in open summer habitats selected forage quantity over quality. Centred on an alpine Strict Nature Reserve, their habitat selection was also strongly influenced by their need for safety: Females responded more strongly to human activities than males and therefore preferred the SNP.

Remote sensing data has proven to be valuable for the estimation of forage quality and quantity in open habitats over large areas and multiple years and months. In particular, the inclusion of vegetation indices and rarely used optical traits in regression models for biomass and relative nitrogen has increased prediction accuracies. While biomass can be mapped with high accuracy, certain ambiguities remain in remote sensing of relative nitrogen. Future studies could leverage on forthcoming hyperspectral satellite sensor to take into account soil and non-photosynthetic vegetation cover, potentially enhancing nitrogen prediction from space.

Availability of data and materials

The dataset supporting the conclusions of this article is available from the corresponding author on reasonable request.

Abbreviations

APEX:

Airborne prism experiment

AVHRR:

Advanced very high resolution radiometer

CAI:

Cellulose absorption index

CBC:

Non-protein carbon based constituents

CHL:

Chlorophyll content

Cm:

Dry matter

DI:

Dissimilarity index

ESA:

European space agency

EWT:

Equivalent water thickness

glmmTMB:

Generalized linear mixed models using template model builder

GNSS:

Global navigation satellite system

iSSFs:

Integrated step selection functions

IUCN:

International union for conservation of nature

LAI:

Leaf area index

LUT:

Look-up table

MODIS:

Moderate-resolution imaging spectroradiometer

MTCI:

MERIS terrestrial chlorophyll index

NDVI:

Normalized difference vegetation index

PROT:

Protein content

PROT_per:

Relative nitrogen content

RFE:

Recursive feature elimination

RMSE:

Root mean square error

RTM:

PROSAIL radiative transfer model

SNAP:

Sentinel application platform

SNP:

Swiss national park

TGI:

Triangular greenness index

VIs:

Vegetation indices

References

  1. Allen AM, Månsson J, Jarnemo A, Bunnefeld N. The impacts of landscape structure on the winter movements and habitat selection of female red deer. Eur J Wildl Res. 2014;60(3):411–21. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/s10344-014-0797-0.

    Article  Google Scholar 

  2. Anderwald P, Buchmann S, Rempfler T, Filli F. Weather-dependent changes in habitat use by Alpine chamois. Mov Ecol. 2024. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-024-00449-x.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Anderwald P, Haller RM, Filli F. Heterogeneity in primary productivity influences competitive interactions between red deer and Alpine chamois. PLoS ONE. 2016;11(1):e0146458. https://doiorg.publicaciones.saludcastillayleon.es/10.1371/journal.pone.0146458.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Arnold W, Beiglböck C, Burmester M, Guschlbauer M, Lengauer A, Schröder B, Wilkens M, Breves G. Contrary seasonal changes of rates of nutrient uptake, organ mass, and voluntary food intake in red deer (Cervus elaphus). Am J Physiol Regul Integr Comp Physiol. 2015;309(3):R277–85. https://doiorg.publicaciones.saludcastillayleon.es/10.1152/ajpregu.00084.2015.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Avgar T, Potts JR, Lewis MA, Boyce MS. Integrated step selection analysis: bridging the gap between resource selection and animal movement. Methods Ecol Evol. 2016;7(5):619–30. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/2041-210X.12528.

    Article  Google Scholar 

  6. Backhaus, N., Buser, C., Butticaz, M., Jorio, D., & Speich, M. (2013). Wirtschaftliche Auswirkungen des Sommertourismus im UNESCO Biosphärenreservat Val Müstair Parc Naziunal. Schriftenreihe Humangeographie 27. http://www.geo.uzh.ch/fileadmin/files/content/abteilungen/human/Publikationen/Wertschoepfung_SNP-BVM_2013.pdf

  7. Bartoń K. MuMIn: multimodal inference. R Package Vers. 2023;1(47):5.

    Google Scholar 

  8. Beck HE, McVicar TR, van Dijk AIJM, Schellekens J, de Jeu RAM, Bruijnzeel LA. Global evaluation of four AVHRR-NDVI data sets: Intercomparison and assessment against Landsat imagery. Remote Sens Environ. 2011;115(10):2547–63. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2011.05.012.

    Article  Google Scholar 

  9. Bischof R, Loe LE, Meisingset EL, Zimmermann B, van Moorter B, Mysterud A. A migratory northern ungulate in the pursuit of spring: jumping or surfing the green wave? Am Nat. 2012;180(4):407–24. https://doiorg.publicaciones.saludcastillayleon.es/10.1086/667590.

    Article  PubMed  Google Scholar 

  10. Bjørneraas K, Van Moorter B, Rolandsen CM, Herfindal I. Screening global positioning system location data for errors using animal movement characteristics. J Wildl Manag. 2010;74(6):1361–6. https://doiorg.publicaciones.saludcastillayleon.es/10.2193/2009-405.

    Article  Google Scholar 

  11. Bonenfant C, Loe LE, Mysterud A, Langvatn R, Stenseth NC, Gaillard JM, Klein F. Multiple causes of sexual segregation in European red deer: enlightenments from varying breeding phenology at high and low latitude. Proceed Royal Soc B Biol Sci. 2004;271(1542):883–92. https://doiorg.publicaciones.saludcastillayleon.es/10.1098/rspb.2003.2661.

    Article  Google Scholar 

  12. Borowik T, Pettorelli N, Sönnichsen L, Jędrzejewska B. Normalized difference vegetation index (NDVI) as a predictor of forage availability for ungulates in forest and field habitats. Eur J Wildl Res. 2013;59:675–82. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/s10344-013-0720-0.

    Article  Google Scholar 

  13. Bracis, C. (2024). recurse. R Package Version 1.3.0.

  14. Cawse-Nicholson K, Townsend PA, Schimel D, Assiri AM, Blake PL, Buongiorno MF, Campbell P, Carmon N, Casey KA, Correa-Pabón RE, Dahlin KM, Dashti H, Dennison PE, Dierssen H, Erickson A, Fisher JB, Frouin R, Gatebe CK, Gholizadeh H, Zhang Q. NASA’s surface biology and geology designated observable: a perspective on surface imaging algorithms. Remote Sens Environ. 2021. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2021.112349.

    Article  Google Scholar 

  15. Chabrillat, S., Brell, M., Segl, K., Foerster, S., & Guanter, L. (2021). The EnMAP satellite - mission status and science preparatory activities. International geoscience and remote sensing symposium (IGARSS), 123–125.

  16. Chraibi E, de Boissieu F, Barbier N, Luque S, Féret JB. Stability in time and consistency between atmospheric corrections: Assessing the reliability of Sentinel-2 products for biodiversity monitoring in tropical forests. Int J Appl Earth Observation Geoinform. 2022. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.jag.2022.102884.

    Article  Google Scholar 

  17. Christianson D, Creel S. Fecal chlorophyll describes the link between primary production and consumption in a terrestrial herbivore. Ecol Appl. 2009;19(5):1323–35. https://doiorg.publicaciones.saludcastillayleon.es/10.1890/08-0987.1.

    Article  PubMed  Google Scholar 

  18. Clevers JGPW, Gitelson AA. Remote estimation of crop and grass chlorophyll and nitrogen content using red-edge bands on sentinel-2 and-3. Int J Appl Earth Obs Geoinf. 2013;23(1):344–51. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.jag.2012.10.008.

    Article  Google Scholar 

  19. Cogliati S, Sarti F, Chiarantini L, Cosi M, Lorusso R, Lopinto E, Miglietta F, Genesio L, Guanter L, Damm A, Pérez-López S, Scheffler D, Tagliabue G, Panigada C, Rascher U, Dowling TPF, Giardino C, Colombo R. The PRISMA imaging spectroscopy mission: overview and first performance analysis. Remote Sens Environ. 2021. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2021.112499.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Coppes J, Burghardt F, Hagen R, Suchant R, Braunisch V. Human recreation affects spatio-temporal habitat use patterns in red deer (Cervus elaphus). PLoS ONE. 2017;12(5):1–19. https://doiorg.publicaciones.saludcastillayleon.es/10.1371/journal.pone.0175134.

    Article  CAS  Google Scholar 

  21. Cracknell AP. The exciting and totally unanticipated success of the AVHRR in applications for which it was never intended. Adv Space Res. 2001;28(1):233–40. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/S0273-1177(01)00349-0.

    Article  Google Scholar 

  22. Dash J, Curran PJ. The MERIS terrestrial chlorophyll index. Int J Remote Sens. 2004;25(23):5403–13. https://doiorg.publicaciones.saludcastillayleon.es/10.1080/0143116042000274015.

    Article  Google Scholar 

  23. Drusch M, Del Bello U, Carlier S, Colin O, Fernandez V, Gascon F, Hoersch B, Isola C, Laberinti P, Martimort P, Meygret A, Spoto F, Sy O, Marchese F, Bargellini P. Sentinel-2: ESA’s optical high-resolution mission for GMES operational services. Remote Sens Environ. 2012;120:25–36. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2011.11.026.

    Article  Google Scholar 

  24. Etten, J. Van, Sumner, M., Cheng, J., Baston, D., Bevan, A., Bivand, R., Busetto, L., Canty, M., Fasoli, B., Forrest, D., Golicher, D., Gray, J., Greenberg, J. A., Hiemstra, P., Karney, C., & Mattiuzzi, M. (2023). raster. R Package Version 3.6–14.

  25. Fattebert J, Morelle K, Jurkiewicz J, Ukalska J, Borkowski J. Safety first: seasonal and diel habitat selection patterns by red deer in a contrasted landscape. J Zool. 2019;308(2):111–20. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/jzo.12657.

    Article  Google Scholar 

  26. Forester JD, Im HK, Rathouz PJ. Accounting for animal movement in estimation of resource selection functions: Sampling and data analysis. Ecology. 2009;90(12):3554–65. https://doiorg.publicaciones.saludcastillayleon.es/10.1890/08-0874.1.

    Article  PubMed  Google Scholar 

  27. Frid A, Dill L. Human-caused disturbance stimuli as a form of predation risk. Ecol Soc. 2002. https://doiorg.publicaciones.saludcastillayleon.es/10.5751/es-00404-060111.

    Article  Google Scholar 

  28. Fryxell JM. Forage quality and aggregation by large herbivores. Am Nat. 1991;138(2):478–98. https://doiorg.publicaciones.saludcastillayleon.es/10.1086/285227.

    Article  Google Scholar 

  29. Garcia F, Ruckstuhl K, Neuhaus P, Coelho C, Wang M, Sousa P, Alves J. Differences in the diets of female and male red deer: The meaning for sexual segregation. Biology. 2023;12:1–13.

    Article  Google Scholar 

  30. Garroutte EL, Hansen AJ, Lawrence RL. Using NDVI and EVI to map spatiotemporal variation in the biomass and quality of forage for migratory elk in the Greater Yellowstone Ecosystem. Remote Sens. 2016. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/rs8050404.

    Article  Google Scholar 

  31. Gascon F, Bouzinac C, Thépaut O, Jung M, Francesconi B, Louis J, Lonjou V, Lafrance B, Massera S, Gaudel-Vacaresse A, Languille F, Alhammoud B, Viallefont F, Pflug B, Bieniarz J, Clerc S, Pessiot L, Trémas T, Cadau E, Fernandez V. Copernicus Sentinel-2A calibration and products validation status. Remote Sens. 2017;9:584. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/rs9060584.

    Article  Google Scholar 

  32. Gelman A. Scaling regression inputs by dividing by two standard deviations. Stat Med. 2008;27:2865–73. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/sim.3107.

    Article  PubMed  Google Scholar 

  33. Godvik IMR, Loe LE, Vik JO, Veiberg V, Langvatn R, Mysterud A. Temporal scales, trade-offs, and functional responses in red deer habitat selection. Ecology. 2009;90(3):699–710. https://doiorg.publicaciones.saludcastillayleon.es/10.1890/08-0576.1.

    Article  PubMed  Google Scholar 

  34. Grignolio S, Merli E, Bongi P, Ciuti S, Apollonio M. Effects of hunting with hounds on a non-target species living on the edge of a protected area. Biol Conserv. 2011;144(1):641–649. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.biocon.2010.10.022

  35. Guerini Filho M, Kuplich TM, Quadros FLFD. Estimating natural grassland biomass by vegetation indices using Sentinel 2 remote sensing data. Int J Remote Sens. 2020;41(8):2861–76. https://doiorg.publicaciones.saludcastillayleon.es/10.1080/01431161.2019.1697004.

    Article  Google Scholar 

  36. Haller, H. (2002). Der Rothirsch im Schweizerischen Nationalpark und dessen Umgebung. In: Eine alpine population von Cervus elaphus zeitlich und räumlich dokumentiert. Nat.park-Forsch. Schweiz 91.

  37. Hamel S, Garel M, Festa-Bianchet M, Gaillard JM, Côté SD. Spring normalized difference vegetation index (NDVI) predicts annual variation in timing of peak faecal crude protein in mountain ungulates. J Appl Ecol. 2009;46(3):582–9. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/j.1365-2664.2009.01643.x.

    Article  Google Scholar 

  38. Hansen BB, Herfindal I, Aanes R, Sæther BE, Henriksen S. Functional response in habitat selection and the tradeoffs between foraging niche components in a large herbivore. Oikos. 2009;118(6):859–72. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/j.1600-0706.2009.17098.x.

    Article  Google Scholar 

  39. Hauser LT, Féret JB, An Binh N, van der Windt N, Sil ÂF, Timmermans J, Soudzilovskaia NA, van Bodegom PM. Towards scalable estimation of plant functional diversity from Sentinel-2: In-situ validation in a heterogeneous (semi-)natural landscape. Remote Sens Environ. 2021. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2021.112505.

    Article  Google Scholar 

  40. Hebblewhite M, Merrill E, McDermid G. A multi-scale test of the forage maturation hypothesis in a partially migratory ungulate population. Ecol Monogr. 2008;78(2):141–66. https://doiorg.publicaciones.saludcastillayleon.es/10.1890/06-1708.1.

    Article  Google Scholar 

  41. Hofmann, R. R. (1973). The ruminant stomach. In: Stomach structure and feeding habits of East African game ruminants (East Afr Monogr Biol (ed.); Vol 2). E.A. Lit. Bureau. https://api.semanticscholar.org/CorpusID:82493565

  42. Hofmann RR. Evolutionary steps of ecophysiological adaptation and diversification of ruminants: a comparative view of their digestive system. Oecologia. 1989;78(4):443–57. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/BF00378733.

    Article  CAS  PubMed  Google Scholar 

  43. Houborg R, Fisher JB, Skidmore AK. Advances in remote sensing of vegetation function and traits. Int J Appl Earth Obs Geoinf. 2015;43:1–6. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.jag.2015.06.001.

    Article  Google Scholar 

  44. Huete A, Didan K, Miura T, E.P., R., Gao, X., & Fereira, L. G. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens Environ. 2002;83:195–213. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/S0020-1693(00)85959-9.

    Article  Google Scholar 

  45. Jacquemoud S, Verhoef W, Baret F, Bacour C, Zarco-Tejada PJ, Asner GP, François C, Ustin SL. PROSPECT + SAIL models: A review of use for vegetation characterization. Remote Sens Environ. 2009;113(SUPPL. 1):S56–66. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2008.01.026.

    Article  Google Scholar 

  46. Johnson HE, Gustine DD, Golden TS, Adams LG, Parrett LS, Lenart EA, Barboza PS. NDVI exhibits mixed success in predicting spatiotemporal variation in caribou summer forage quality and quantity. Ecosphere. 2018. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/ecs2.2461.

    Article  PubMed  PubMed Central  Google Scholar 

  47. Jones CL, Weckler PR, Maness NO, Jayasekara R, Stone ML, Chrz D. Remote sensing to estimate chlorophyll concentration in spinach using multi-spectral plant reflectance. Trans ASABE. 2007;50(6):2267–73. https://doiorg.publicaciones.saludcastillayleon.es/10.13031/2013.24079.

    Article  CAS  Google Scholar 

  48. Kattenborn T, Schiefer F, Zarco-Tejada P, Schmidtlein S. Advantages of retrieving pigment content [μg/cm2] versus concentration [%] from canopy reflectance. Remote Sens Environ. 2019;230:111195. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2019.05.014.

    Article  Google Scholar 

  49. Kuhn, M., Wing, J., Weston, S., Williams, A., Keefer, C., Engelhardt, A., Cooper, T., Mayer, Z., Kenkel, B., & Team, R. C. (2020). caret. R Package Version 6.0–93.

  50. Langvatn R, Hanley TA. Feeding-patch choice by red deer in relation to foraging efficiency. Oecologia. 1993;95(2):164–70. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/BF00323486.

    Article  PubMed  Google Scholar 

  51. Laundré JW, Hernández L, Ripple WJ. The landscape of fear: ecological implications of being afraid. Open Ecol J. 2010;3:1–7.

    Article  Google Scholar 

  52. Liaw A, Wiener M. randomForest’. R Package Vers. 2022;4(7–1):1.

    Google Scholar 

  53. Mattioli, S., Zachos, F. E., Rossi, L., Lister, A. M., & Corlatti, L. (2022). Red deer Cervus elaphus Linnaeus, 1758. In K. Hackländer & F. E. Zachos (Eds.), Handbook of the Mammals of Europe. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/978-3-319-65038-8_19-1

  54. Mattson, W. J. (1980). Herbivory in relation to plant nitrogen content. Annual review of ecology and systematic, 11, 119–161. http://www.jstor.org/stable/2096905

  55. MeteoSwiss. (2022). IDAweb. Data portal for teaching and research. https://gate.meteoswiss.ch/idaweb/login.do

  56. Meyer, H., Milà, C., Ludwig, M., Linnenbrink, J., Otto, P., Reudenbach, C., Nauss, T., & Pebesma, E. (2020). CAST. R Package Version 0.8.1.

  57. Meyer H, Pebesma E. Predicting into unknown space? Estimating the area of applicability of spatial prediction models. Methods Ecol Evol. 2021;12(9):1620–33. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/2041-210X.13650.

    Article  Google Scholar 

  58. Mikle NL, Graves TA, Olexa EM. To forage or flee: lessons from an elk migration near a protected area. Ecosphere. 2019;10(4):1–15. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/ecs2.2693.

    Article  Google Scholar 

  59. Millhäusler A, Anderwald P, Haeni M, Haller RM. Publicity, economics and weather—changes in visitor numbers to a European national park over 8 years. J Outdoor Recreat Tour. 2016;16(April):50–7. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.jort.2016.09.005.

    Article  Google Scholar 

  60. Muff S, Signer J, Fieberg J. Accounting for individual-specific variation in habitat-selection studies: Efficient estimation of mixed-effects models using Bayesian or frequentist computation. J Anim Ecol. 2020;89(1):80–92. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/1365-2656.13087.

    Article  PubMed  Google Scholar 

  61. Mutanga O, Masenyama A, Sibanda M. Spectral saturation in the remote sensing of high-density vegetation traits: A systematic review of progress, challenges, and prospects. ISPRS J Photogramm Remote Sens. 2023;198:297–309. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.isprsjprs.2023.03.010.

    Article  Google Scholar 

  62. Mysterud A. The relationship between ecological segregation sexual body size dimorphism large herbivores. Oecologia. 2000;124(1):40–54. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/s004420050023.

    Article  CAS  PubMed  Google Scholar 

  63. Mysterud A, Ims RA. Functional responses in habitat use: availability influences relative use in trade-off situations. Ecology. 1998;79(4):1435–41. https://doiorg.publicaciones.saludcastillayleon.es/10.2307/176754.

    Article  Google Scholar 

  64. Mysterud A, Vike BK, Meisingset EL, Rivrud IM. The role of landscape characteristics for forage maturation and nutritional benefits of migration in red deer. Ecol Evol. 2017;7(12):4448–55. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/ece3.3006.

    Article  PubMed  PubMed Central  Google Scholar 

  65. NASA. (2020). NASADEM Merged DEM Global 1 arc second V001. https://doiorg.publicaciones.saludcastillayleon.es/10.5067/MEaSUREs/NASADEM/NASADEM_HGT.001

  66. Neumann W, Martinuzzi S, Estes AB, Pidgeon AM, Dettki H, Ericsson G, Radeloff VC. Opportunities for the application of advanced remotely-sensed data in ecological studies of terrestrial animal movement. Mov Ecol. 2015;3(1):1–13. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-015-0036-7.

    Article  Google Scholar 

  67. Pettorelli N, Ryan S, Mueller T, Bunnefeld N, Jedrzejewska B, Lima M, Kausrud K. The Normalized Difference Vegetation Index (NDVI): Unforeseen successes in animal ecology. Climate Res. 2011;46(1):15–27. https://doiorg.publicaciones.saludcastillayleon.es/10.3354/cr00936.

    Article  Google Scholar 

  68. Proffitt KM, Grigg JL, Hamlin KL, Garrott RA. Contrasting effects of wolves and human hunters on elk behavioral responses to predation risk. J Wildl Manag. 2009;73(3):345–56. https://doiorg.publicaciones.saludcastillayleon.es/10.2193/2008-210.

    Article  Google Scholar 

  69. Proffitt KM, Hebblewhite M, Peters W, Hupp N, Shamhart J. Linking landscape-scale differences in forage to ungulate nutritional ecology. Ecol Appl. 2016;26(7):2156–74. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/eap.1370.

    Article  PubMed  Google Scholar 

  70. Punalekar SM, Verhoef A, Quaife TL, Humphries D, Bermingham L, Reynolds CK. Application of Sentinel-2A data for pasture biomass monitoring using a physically based radiative transfer model. Remote Sens Environ. 2018;218:207–20. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2018.09.028.

    Article  Google Scholar 

  71. Pyke GH. Optimal foraging theory: a critical review. Annu Rev Ecol Syst. 1984;15:523–75. https://doiorg.publicaciones.saludcastillayleon.es/10.1146/annurev.ecolsys.15.1.523.

    Article  Google Scholar 

  72. Quan X, He B, Yebra M, Yin C, Liao Z, Zhang X, Li X. A radiative transfer model-based method for the estimation of grassland aboveground biomass. Int J Appl Earth Obs Geoinf. 2017;54(February):159–68. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.jag.2016.10.002.

    Article  Google Scholar 

  73. R Core Team. (2018). A language and environment for statistical computing (R foundation for statistical computing (ed.)).

  74. Raab C, Riesch F, Tonn B, Barrett B, Meißner M, Balkenhol N, Isselstein J. Target-oriented habitat and wildlife management: estimating forage quantity and quality of semi-natural grasslands with Sentinel-1 and Sentinel-2 data. Remote Sen Ecol Conser. 2020;6(3):381–98. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/rse2.149.

    Article  Google Scholar 

  75. Rast M, Nieke J, Adams J, Isola C, Gascon F. Copernicus Hyperspectral Imaging Mission for the Environment (Chime). Int Geosci Remote Sens Symp (IGARSS). 2021. https://doiorg.publicaciones.saludcastillayleon.es/10.1109/IGARSS47720.2021.9553319.

    Article  Google Scholar 

  76. Rossi C, Kneubühler M, Schütz M, Schaepman ME, Haller RM, Risch AC. From local to regional: functional diversity in differently managed alpine grasslands. Remote Sens Environ. 2020;236:111415. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.rse.2019.111415.

    Article  Google Scholar 

  77. Rossi C, Kneubühler M, Schütz M, Schaepman ME, Haller RM, Risch AC. Spatial resolution, spectral metrics and biomass are key aspects in estimating plant species richness from spectral diversity in species-rich grasslands. Remote Sens Ecol Conserv. 2022;8(3):297–314. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/rse2.244.

    Article  Google Scholar 

  78. Schweiger AK, Risch AC, Damm A, Kneubühler M, Haller R, Schaepman ME, Schütz M. Using imaging spectroscopy to predict above-ground plant biomass in alpine grasslands grazed by large ungulates. J Veg Sci. 2015;26(1):175–90. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/jvs.12214.

    Article  Google Scholar 

  79. Schweiger AK, Schütz M, Anderwald P, Schaepman ME, Kneubühler M, Haller R, Risch AC. Foraging ecology of three sympatric ungulate species—behavioural and resource maps indicate differences between chamois, ibex and red deer. Mov Ecol. 2015;3(1):6. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-015-0033-x.

    Article  PubMed  PubMed Central  Google Scholar 

  80. Signer J, Fieberg J, Avgar T. Animal movement tools (amt): R package for managing tracking data and conducting habitat selection analyses. Ecol Evol. 2019;9(2):880–90. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/ece3.4823.

    Article  PubMed  PubMed Central  Google Scholar 

  81. Sigrist B, Signer C, Wellig SD, Ozgul A, Filli F, Jenny H, Thiel D, Wirthner S, Graf RF. Green-up selection by red deer in heterogeneous, human-dominated landscapes of Central Europe. Ecol Evol. 2022;12(7):1–13. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/ece3.9048.

    Article  Google Scholar 

  82. Stephens DW, Krebs JR. Foraging theory. Princeton University Press; 1986.

    Google Scholar 

  83. Strobl C, Boulesteix AL, Zeileis A, Hothorn T. Bias in random forest variable importance measures: illustrations, sources and a solution. BMC Bioinform. 2007. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/1471-2105-8-25.

    Article  Google Scholar 

  84. Swisstopo. (2015). Objektkatalog swissTLM 1.3. 51.

  85. Thieurmel, B., & Elmarhraoui, A. (2022). suncalc. R Package Version 0.5.1. https://github.com/datastorm-open/suncalc

  86. Ullah S, Si Y, Schlerf M, Skidmore AK, Shafique M, Iqbal IA. Estimation of grassland biomass and nitrogen using MERIS data. Int J Appl Earth Obs Geoinf. 2012;19(1):196–204. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.jag.2012.05.008.

    Article  Google Scholar 

  87. Verrelst J, Camps-Valls G, Muñoz-Marí J, Rivera JP, Veroustraete F, Clevers JGPW, Moreno J. Optical remote sensing and the retrieval of terrestrial vegetation bio-geophysical properties—a review. ISPRS J Photogramm Remote Sens. 2015;108:273–90. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.isprsjprs.2015.05.005.

    Article  Google Scholar 

  88. Verrelst J, Halabuk A, Atzberger C, Hank T, Steinhauser S, Berger K. A comprehensive survey on quantifying non-photosynthetic vegetation cover and biomass from imaging spectroscopy. Ecol Ind. 2023;155: 110911. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ecolind.2023.110911.

    Article  Google Scholar 

  89. Verrelst J, Rivera-Caicedo JP, Reyes-Muñoz P, Morata M, Amin E, Tagliabue G, Panigada C, Hank T, Berger K. Mapping landscape canopy nitrogen content from space using PRISMA data. ISPRS J Photogramm Remote Sens. 2021;178(June):382–95. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.isprsjprs.2021.06.017.

    Article  PubMed  PubMed Central  Google Scholar 

  90. Verrelst J, Vicent J, Rivera-Caicedo JP, Lumbierres M, Morcillo-Pallarés P, Moreno J. Global sensitivity analysis of leaf-canopy-atmosphere RTMs: implications for biophysical variables retrieval from top-of-atmosphere radiance data. Remote Sens. 2019;11(16):1–26. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/rs11161923.

    Article  Google Scholar 

  91. Viana DS, Granados JE, Fandos P, Pérez JM, Cano-Manuel FJ, Burón D, Fandos G, Ángeles M, Aguado P, Figuerola J, Soriguer RC. Linking seasonal home range size with habitat selection and movement in a mountain ungulate. Mov Ecology. 2018. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-017-0119-8.

    Article  Google Scholar 

  92. Weiss, M., Baret, F., & Jay, S. (2020). S2ToolBox Level 2 Products: LAI, FAPAR, FCOVER.(Version 2.0). In Institut national de la recherche agronomique (INRA). https://step.esa.int/docs/extra/ATBD_S2ToolBox_L2B_V1.1.pdf

  93. Wilmshurst JF, Fryxell JM, Hudson RJ. Forage quality and patch choice by wapiti (Cervus elaphus). Behav Ecol. 1995;6(2):209–17. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/beheco/6.2.209.

    Article  Google Scholar 

  94. Wipf S, Michel AH, Walder D, Poelsma F, Anderwald P, Cruickshank SS, Jaroszynska F, Rossi C, Backhaus N. Boost in visitor numbers post COVID-19 shutdown: consequences for an Alpine national park. Mt Res Dev. 2023;43(2):R12–21. https://doiorg.publicaciones.saludcastillayleon.es/10.1659/mrd.2022.00025.

    Article  Google Scholar 

  95. Yeoh HH, Wee YC. Leaf protein contents and nitrogen-to-protein conversion factors for 90 plant species. Food Chem. 1994;49(3):245–50. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/0308-8146(94)90167-8.

    Article  CAS  Google Scholar 

  96. Zanaga, D., Van De Kerchove, R., De Keersmaecker, W., Souverijns, N., Brockmann, C., Quast, R., Wevers, J., Grosu, A., Paccini, A., Vergnaud, S., Cartus, O., Santoro, M., Fritz, S., Georgieva, I., Lesiv, M., Carter, S., Herold, M., Li, L., Tsendbazar, N. E., Arino, O. (2021). ESA WorldCover 10 m 2020 v100 (pp. 1–27). https://doiorg.publicaciones.saludcastillayleon.es/10.5281/zenodo.5571936

  97. Zweifel-Schielly B, Kreuzer M, Ewald KC, Suter W. Habitat selection by an Alpine ungulate: the significance of forage characteristics varies with scale and season. Ecography. 2009;32(1):103–13. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/j.1600-0587.2008.05178.x.

    Article  Google Scholar 

  98. Zweifel-Schielly B, Leuenberger Y, Kreuzer M, Suter W. A herbivore’s food landscape: seasonal dynamics and nutritional implications of diet selection by a red deer population in contrasting Alpine habitats. J Zool. 2012;286(1):68–80. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/j.1469-7998.2011.00853.x.

    Article  Google Scholar 

Download references

Acknowledgements

The authors would like to thank all game keepers and Swiss National Park rangers who were responsible for the capture and handling of the animals, and our veterinarian Marianne Caviezel-Ring. We also want to thank Anna Katharina Schweiger for providing field sample data, Ruedi Haller, director of the Swiss National Park, and Adrian Arquint, head of Grison’s authority for hunting and fisheries, and for their support at all stages of the project. Finally, we want to thank the NUKAHIVA foundation for their financial support.

Funding

Funding was provided by the Hunting and Fisheries Department of the Canton of Grisons, the Swiss National Park and the NUKAHIVA foundation.

Author information

Authors and Affiliations

  1. Department of Integrative Biology and Biodiversity Research, Institute of Wildlife Biology and Game Management, BOKU University, Vienna, Austria

    Thomas Rempfler & Klaus Hackländer

  2. Swiss National Park, Chastè Planta-Wildenberg, Runatsch 124, 7530, Zernez, Switzerland

    Thomas Rempfler, Christian Rossi, Jan Schweizer, Flurin Filli, Sven Buchmann & Pia Anderwald

  3. Research Unit Wildlife Biology and Management, Bavarian State Institute of Forestry, Freising, Germany

    Wibke Peters

  4. Institute of Natural Resource Sciences, Wildlife Management Unit, Zurich University of Applied Sciences, Wädenswil, Switzerland

    Claudio Signer

  5. Canton of Grisons, Hunting and Fisheries Department, Chur, Switzerland

    Hannes Jenny

  6. Deutsche Wildtier Stiftung, Hamburg, Germany

    Klaus Hackländer

Authors
  1. Thomas Rempfler
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Christian Rossi
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Jan Schweizer
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Wibke Peters
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Claudio Signer
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Flurin Filli
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Hannes Jenny
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Klaus Hackländer
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Sven Buchmann
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Pia Anderwald
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

TR, CR, PA and FF had the original idea for the study. JS and SB refined the study design and analytical approach. Fieldwork was organised by TR, FF and HJ. CR and JS performed the statistical analyses on the vegetation characteristics, TR, PA and SB on the integrated step selection functions. TR, CR, JS and PA drafted the manuscript. All authors made important contributions by editing the final manuscript.

Corresponding author

Correspondence to Thomas Rempfler.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Rempfler, T., Rossi, C., Schweizer, J. et al. Remote sensing reveals the role of forage quality and quantity for summer habitat use in red deer. Mov Ecol 12, 80 (2024). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-024-00521-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-024-00521-6

Keywords

Movement Ecology

ISSN: 2051-3933