Procedure to automate the classification and mapping of the vegetation density in arid aeolian sedimentary systems

Abstract

The temporal variation in vegetation cover in aeolian sedimentary systems, especially those in arid regions, provides an indication of environmental change. Based on this, the objective of this paper is to design a simple method for classifying the vegetation density of arid aeolian sedimentary systems through the digital processing of aerial images. The green band of a high resolution orthophoto of La Graciosa island (Canary Islands, Spain) is used as an example. The pixels identified as vegetation were vectorized to point geometry, the vegetation density was then calculated, and a digital vegetation density model (DVDM) thereby obtained. Both spatial and statistical analyses were performed to find the optimal procedure to achieve the objective. Speed, objectivity, cost, and the possibility of working with historical records are discussed, and the proposed method is compared with others based on visual analysis or digital remote sensing. The importance of this method for countries with less research funding or low GDP per capita is also discussed. The proposed procedure opens up future lines of research for comparison of results across various environmental and anthropogenic variables. In addition, vegetation density can be used as a variable in computational fluid dynamic modeling of vegetation in arid dune systems.

Keywords

Dune vegetation arid aeolian sedimentary systems density geographical information technologies method

I. Introduction

Globally, in recent decades, there have been environmental problems related to the instability of coastal dune systems (Jackson and Nordstrom, 2011; Nordstrom, 1994). Regardless of the environmental effects of climate change on them, it has been suggested that their sedimentary activity is a “geo-indicator” of environmental changes (Berger and Iams, 1996).

The presence of vegetation in these dune systems leads to a reduction in rates of sediment transport (Lancaster and Baas, 1998; Levin et al., 2008; Okin, 2008; Sherman and Bauer, 1993). This is due, among other things, to plants increasing the roughness of the ground surface, which results in the reduction of wind speed and thereby the reduction of aeolian sediment transport (Hesp, 1981; Moreno-Casasola, 1986). Thus, plants capture sand grains that are transported by saltation, favoring the retention of sediments. There is therefore a direct link between the formation of sand dunes and vegetation. Some aspects of the vegetation, such as cover, density, height, volume, biotype, and conservation, condition the retention levels of sediments (Carter, 1988; Hesp, 1984, 1989, 1991, 2002). Vegetation cover, and its variations, is a determining factor in the generation of landforms and in dune dynamics, given the interactions that take place between vegetation and aeolian sediment transport (García-Romero et al., 2016; Hesp, 2002; Kuriyama et al., 2005; Luna et al., 2011; Tsoar, 2005). In fact, vegetation cover is the most often used factor in the characterization of stabilization processes (García-Romero et al., 2016; Hesp, 2013; Levin et al., 2008). The dynamics of vegetation cover are particularly interesting in coastal sandy systems, because it is easy to identify changes, whether due to natural alteration or induced by human activities over short time periods (especially arid ones), which produce significant environmental modifications such as dune stabilization, erosion, or remobilization (Cabrera-Vega et al., 2013; Hernández-Calvento et al., 2014; Kutiel et al., 2004; Santana-Cordero et al., 2016; Hernández-Cordero et al., 2017).

Most works which focus on the characterization of vegetation in dune systems give special importance to the percentage of vegetation cover in the space. These works use different concepts, such as plant density (Konlechner et al., 2014), vegetation density (Kutiel et al., 2004), vegetation cover (Carlson and Ripley, 1997; Ritzen and Straatsma, 2002; Zehm et al., 2003), plant cover (Floyd and Anderson, 1987), vegetation recovery (Costa et al., 1996; Hayden et al., 1995; Stallins, 2002), or vegetation covering (Qin et al., 2006). Depending on the methodology applied, the techniques used vary from field work (Canfield, 1941; Daubenmire, 1959; Goodall, 1952; Greig-Smith, 1983; Konlechner et al., 2014) to parallel photography and terrestrial laser scanning (Warmink, 2007), while special mention should be given to techniques that use airborne information sources (Bannari et al., 1995; Hugenholtz et al., 2012).

Aerial photographs are important airborne information sources for the study of the vegetation and its dynamics in dune systems (Carmel and Kadmon, 1998; Delgado-Fernández and Mir-Gual, 2015; Dunn et al., 1990; Green et al., 1993; Hellesen and Levin, 2014; Kadmon and Harari-Kremer, 1999; Okeke and Karnieli, 2006). Traditionally, studies of the vegetation (especially vegetation cover) with aerial photography have been made from visual analysis (Akashi and Mueller-Dombois, 1995; Coppedge et al., 2001; Delgado-Fernández and Mir-Gual, 2015; Dirzo and García, 1992; García et al., 2012; García-Romero et al., 2016; Gaylord and Stetler, 1994; Harrington and Sanderson, 1994; Hernández-Cordero et al., 2015b; Johnston and Naiman, 1990; Miller et al., 1995; Simpson et al., 1994; Turner et al., 1996). Among the limitations of this method, the problem of subjectivity is of particular importance (Morgan et al., 2010). Alternative approaches to visual photo interpretation of aerial photographs and other information sources involve the application of automated methods (Avery and Berlin, 1992; Hugenholtz et al., 2012; Jauhiainen et al., 2007; Okeke and Karnieli, 2006; Tsoar and Blumberg, 2002). Some of these methods classify the vegetation using multispectral and hyperspectral images, captured by satellite or airborne sensors. The technique implemented usually entails the development of vegetation indices. One of the most commonly used indices is the normalized difference vegetation index (NDVI) (Carlson and Ripley, 1997; Chen et al., 2012; De Lange et al., 2004; Elmore et al., 2000; Hall et al., 1998; Lambin and Ehrlich, 1997; Schmidt and Karnieli, 2000; Underwood et al., 2003; Woodcock et al., 1994; Zeng et al., 2013). However, in areas with high soil reflectivity values, the soil-adjusted vegetation index (SAVI) produces better results (Huete, 1988; Levin et al., 2006, 2008). These vegetation indices can also be combined (NDVI + SAVI) (Soria et al., 2007) or used in relation with LiDAR data (Collin, Long and Archambault, 2010, 2012).

In arid aeolian sedimentary systems, independently of soil brightness (Huete, 1988), the use of vegetation indices or other methods based on near-infrared bands is not feasible because the permanent vegetation (formed usually by bush specimens) endures almost constant water stress due to low rainfall and high wind speeds, especially in summer. In these circumstances, the vegetation displays different morphological adaptations, such as the total or partial loss of leaves (Figure 1). Thus, the vegetation exhibits a low spectral response to infrared bands, even though the plants are alive and, therefore, intervening in the aeolian sedimentary dynamics.

Figure 1.

Water stressed vegetation in arid aeolian sedimentary systems forming nebkhas.

In short, the most commonly used methods for the characterization of vegetation in dune systems through airborne information sources are either qualitative, usually based on the identification of vegetation cover through visual interpretation, or quantitative, based on the identification of the state of the vegetation, from the data provided by the infrared channels of digital images. In the case of arid dune systems, the first option is valid and widely used when the objective is to characterize plant species, or perform an analysis of the space-time dynamics of the vegetation cover in relation with aeolian geomorphology. However, the second option lacks accuracy, because certain adaptations of the vegetation to these environments (such as the loss of leaves) prevent accurate records of the surfaces actually covered by vegetation.

A possible alternative to quantitatively characterize the vegetation in these arid systems would be to calculate its density, instead of its coverage, since this calculation not only provides data on the amount of vegetation per unit area, but also the mean distance between the vegetation individuals, which is in direct relation with aeolian sedimentary dynamics (Hesp, 1991, 2002). In this context, the concept of vegetation density has been defined as the number of plant individuals in an area (Hernández et al., 2000). In addition, a second definition includes the mean distance between plant individuals in the characterization of vegetation density (Mostacedo and Frederiksen, 2000). This second concept is more appropriate in aeolian sedimentary dynamics, because the distance between plant individuals influences the wind flows, the sedimentary transport and the formation of aeolian landforms (Hesp, 1983). The vegetation density shows the possibility that the wind flows act or not, depending on the density, and therefore generate dune landforms. In this sense, the vegetation density could be the next step in computational fluid dynamic (CFD) modeling of the vegetation, at least at regional scale, which, according to Smyth (2016), is a perspective in the future.

Calculating the vegetation density of dune systems has been performed through the processing of digital images, though not analogic ones. One of the most commonly used methods is to employ automatic classifications using near-infrared bands (Hugenholtz et al., 2012). Another method involves statistical reclassification of the digital levels of color aerial photographs, related to aspects such as intensity and tone (as HSI) or texture (Barry and Bowman, 2006; Hudak and Wessman, 1998; Nurminen et al., 2015), correlated with field data. Object-oriented methods have also been applied to digital images (Laliberte et al., 2004; Platt and Schoennagel, 2009). Finally, new sensors have been introduced, as well as the use of synthetic-aperture radar (SAR) imagery (Rozenstein et al., 2016).

However, to date no method has been developed to calculate actual values of the vegetation density using the visible bands of digital orthophotos applied to black and white historical aerial photographs, enabling spatiotemporal studies about vegetation changes and their relationship with aeolian sedimentary dynamics.

Considering this background, this paper proposes a simple automated procedure to calculate the vegetation density in arid aeolian sedimentary systems. For this purpose, a small island called La Graciosa (Canary Islands) was used as a laboratory. The developed method uses visible channels of current digital orthophotos to facilitate a comparison with the results obtained after its application to black and white aerial historical photos. The specific objectives are the following: (a) characterization of the vegetation density in a 200 × 200 m plot in La Graciosa using both empirical and automated methods; (b) calculation of the optimal sampling of the automated method, considering the overlaying with the results obtained when using the empirical method and by statistical analysis; (c) application of the optimal sampling to the 200 × 200 m plot to verify its feasibility when using historical aerial photographs in black and white; and finally (d) application of the optimal sampling to all the current aeolian sedimentary systems of the island.

1. Study area

The island of La Graciosa (27.05 km²) is located in the northeastern area of the Canary Archipelago (Figure 2), on a Quaternary marine platform, formed as a result of the emission of volcanic material during the Upper Pleistocene and Holocene (De la Nuez et al., 1997). In parallel, the Quaternary sedimentary processes gave rise to soil and sand deposits that are currently slightly cemented (aeolianite) (Meco et al., 2006; Ortiz et al., 2006). Marine terraces and beachrock deposits were formed in the coastal areas.

Figure 2.

Location of the study and pilot area. In yellow color with transparency (left), the aeolian sedimentary systems.

In the Holocene, a significant part of the island (13.96 km²) has been covered by aeolian sedimentary deposits of variable thickness. The sands are medium-coarse organogenic (>75% of bioclastic grains, with abundant fragments of Corallinaceae algae mesh, and with a carbonate content of >84%) (Mangas et al., 2012). Although there is currently aeolian remobilization, the landforms showing substantial levels of sediment transport, such as ripples, are fairly insignificant, so semi-stabilized aeolian sand sheets and nebkha dunes predominate, stabilized by vegetation specimens. The dominant vegetation is made up of shrub and herbs of halophilous, psammophilous and xerophilous species, such as Traganum moquinii, Salsola vermiculata, Launaea arborescens, Ononis hesperia, Polycarpaea nivea, Euphorbia paralias, and Ononis tournefortii (González et al., 1996; Hernández-Cordero et al., 2015a; Pérez-Chacón et al., 2010). Vegetation cover, in turn, is significant, and in some winters covers more than half of the surface area of the island’s aeolian systems.

This vegetation is adapted to the arid-like conditions of the island’s climate. Mean annual precipitation barely tops 100 mm, distributed over an average of just 32 days of rain a year. This rain is torrential and very irregular, both by season and year, with the rainiest months being between November and February (Pérez-Chacón et al., 2010). Temperatures are mild, with a mean annual temperature of 19.7°C and mean variation over the year of 5.7°C (Pérez-Chacón et al., 2010). The prevailing winds are the trade winds, blowing mostly in NE, ENE, and NNE directions at an average speed of 22 km/h. These winds, which are effective for aeolian sedimentary transport, are more significant in January, March and April, while in September they blow less frequently (Pérez-Chacón et al., 2010). In winter, south-westerly storms are customary, generating erosional processes on the island’s southern coast and temporarily changing the direction of aeolian sedimentary transport and marine dynamics (Pérez-Chacón et al., 2010). From an anthropic point of view, the island is sparsely populated, with just 700 inhabitants in 2014 according to the Spanish National Institute of Statistics (Instituto Nacional de Estadística, 2014). However, over recent decades there has been a notable increase in the number of buildings and port infrastructures, as well as a change in the economic model of the island from one based on traditional farming and fishing to one based on tourism. La Graciosa is legally protected by numerous national and international forms of environmental protection: Biosphere Reserve, Special Area of Conservation (SAC), Special Protection Area for Birds (SPAs), Nature Park and Geopark.

II. Materials and methods

A digital orthophoto was used for the development of the procedure described in this paper. The orthophoto was taken in February 2009 with a spatial resolution of 10 cm and using four spectral bands, three in the visible spectrum and one near-infrared. For the criteria of using only visible bands, so that the method could also be applied to historical images (normally in black and white and without infrared bands), the green band was chosen for this study, since it provides the most vegetation cover information of the bands in the visible domain (Chuvieco, 2010). This choice was favored by the fact that the vegetation was in an optimal state of vigor at the time the orthophoto images were produced. In addition, for the application of the procedure in the observation plot (Figure 9), historical aerial photographs were used (Table 1) taken in the winter season (the rainiest months according to Pérez-Chacón et al., 2010). In this way, it was hoped to show that the method described in the present paper is applicable to historical aerial photographs.

Figure 3.

Removing the areas without aeolian sedimentary systems (step 1) and resampling the green band (step 2).

Figure 4.

Categories used to estimate the percentage of vegetation cover per area.

Figure 5.

(a) Result of the reclassification of the green band using the Jenks natural breaks method. (b) Pixel vectorization with vegetation cover to points. (c) Different vegetation cover densities obtained.

Figure 6.

Results for each neighborhood sampling used.

Figure 7.

Categories of vegetation density obtained from the automated (neighboring sample 9×9) and empirical methods, considering the degree of vegetation cover and the distance between pixels with vegetation. Results from the photo interpretation at a scale of 1:1000. Categories: 1. <25%, 2. 25%–50%, 3. 50%–75% and 4. >75%.

Figure 8.

Results of the application of 9 × 9 m sampling to the observation plot in 1954, 1987, and 2009.

Figure 9.

Results of the application of 9 × 9 m neighborhood sampling to all aeolian sedimentary systems of La Graciosa in 2009.

Table 1.

Sources used in this paper.

Sources	Date	Scale	Spatial resolution (m)	RMS (m)
Historical aerial photography	12/1954	1:18,000	1	1.05–2.01
Historical aerial photography	01/1987	1:5000	1	0.55–1.04
Orthophoto	02/2009	*	0.1	*

In order to carry out the spatial and statistical analyses required to define the method, we worked first with a 200 × 200 m pilot area or observation plot located in the south of the island (Figure 2). This area is considered representative of the whole of the island, in the context of aeolian sedimentary systems, as it presents the different degrees of vegetation coverage identified in the island and also the surfaces characterized by the presence of sand sheets without vegetation cover in the recent past.

1. Procedure for calculating the vegetation density

1.1. Preparation of the green band of the orthophoto

Firstly, we proceeded to eliminate those areas outside the scope of the study, extracting from the processing, following Kutiel et al. (2004), volcanic cones, lava flows, buildings, etc. and focusing the analysis on the aeolian sedimentary systems defined by Pérez-Chacón et al. (2010) (Figure 2). The following step was related to spatial resolution and involved resampling the pixel size to 1 m. Its aim was to facilitate the rest of the procedure and to make this information source comparable with the historical information sources also used (Figure 3).

1.2. Empirical method and categories of vegetation density

The empirical method was based on a visual analysis of different vegetation densities, according to their degree of coverage per area. For this, four categories of vegetation cover were manually defined in the observation plot at a scale of 1:1000 (Figure 6), following the method developed by Shanmugam and Barnsley (2002). To scan the areas with a similar vegetation density, an adaptation of the graphs for visual appreciation of coverages by Folk and Ward (1957) was used, selecting the following coverage intervals per area: 0–25%, 25–50%, 50–75%, and >75% (Figure 4).

2. Automated method for digital vegetation density model (DVDM)

2.1 Reclassification

A reclassification in the resampled green band of the orthophoto (February 2009) was performed, using the Jenks natural breaks method. Each class obtained was characterized by similar values, maximizing the differences between classes. The result (Figure 5(a)) was a classification of the green band which identifies those grey values that represent vegetation. As a result of the previous resampling, it was noted that the vegetation detected normally corresponded to semi-shrubby and shrubby plants, while small seasonal herbaceous species, such as therophytes, were not detected.

2.2. Measuring the vegetation density

The vegetation density calculation was based on the method developed by Mostacedo and Frederiksen (2000) to calculate the density of trees per hectare because the shrub vegetation in arid aeolian sedimentary systems has usually a random distribution in the landscape, similar to trees, and shows as follows

Dh = \frac{10000}{{(\bar{D})}^{2}}

where Dh is the density by hectare and $\bar{D}$ is the average distance between central points.

In our case, the central points are every pixel representing vegetation values obtained in the reclassification, vectorized to point geometry (Figure 5(b)).

The vegetation density could then be estimated based on the vectorization process. This density can be understood as the degree of vegetation coverage, also considering that the smaller the distance between points, the higher the vegetation density, and vice versa. A DVDM was thus obtained following all the previous steps. This model was reclassified in the same four vegetation density categories used in the empirical method, using GIS tools. One example with different conditions with respect to quantity and distance is shown in Figure 5(c).

3. DVDM: optimal sampling

The calculation of density started with neighborhood sampling around each point. This operation can be carried out using any free or commercial GIS software which normally incorporates these tools. The next step was estimation of the optimal distance to sample and calculate densities around each point. The first sampling was performed using the 1 × 1 m² option. Then, multiples of 3 units (meters) were applied, as Hudak and Wessman (1998) proposed, that is, 3 × 3, 6 × 6, 9 × 9, 12 × 12, 15 × 15, 18 × 18, 21 × 21, and 24 × 24 m. The results are shown in Figure 7 and Table 2. Two types of analyses were performed to determine the optimum neighborhood sampling. First, the results obtained from the automated samplings and those obtained from the visual analysis (empirical method) of the image were compared. This comparison was made through spatial analysis performed with geoprocessing tools (GIS) to determine the percentage of coincidence. Second, a statistical analysis was performed to determine the best neighborhood sampling that separates the four categories obtained from the automated method.

Table 2.

Mean occupancy percentage of vegetation cover in the area studied, for each neighborhood sampling used and mean distance between shrubby individuals.

Neighborhood sampling	1		2		3		4
Neighborhood sampling	MO	MD	MO	MD	MO	MD	MO	MD
1×1	34.48	1.03	—	—	—	—	88.65	1.00
3×3	25.88	2.07	34.50	1.12	82.50	1.00	99.78	1.00
6×6	19.53	1.51	19.50	1.21	60.13	1.02	90.07	1.00
9×9	10.65	2.23	22.35	1.79	49.26	1.31	84.25	1.00
12×12	17.77	2.03	22.84	1.19	49.97	1.02	86.13	1.00
15×15	8.82	1.46	19.89	1.26	43.47	1.05	72.62	1.00
18×18	2.53	1.57	26.18	1.62	44.15	1.03	78.47	1.00
21×21	4.14	1.70	43.81	1.28	55.92	1.02	75.17	1.21
24×24	4.64	1.91	24.62	1.20	45.52	1.06	64.64	1.42

MO: mean percentage of occupation of the vegetation by category studied (%); MD: mean distance between shrubby individuals (m).

3.1. Spatial analysis

To perform the spatial analysis, a comparison was performed between the results obtained applying the DVDM in the observation plot (Figure 6) and those provided by the empirical method (Figure 7, right). The comparison between the various results obtained with the different neighborhood sampling sizes and the result obtained through the empirical method was designed to determine the percentage of vegetation cover coincident with the visual analysis (Cva) for each neighborhood sampling size. For this, a spatial analysis using GIS geoprocessing tools (overlaying) was performed to determine which sample, among those obtained automatically, presented the closest relationship to the results obtained through the empirical method.

3.2. Statistical analysis

The aim of the statistical analysis was to determine which sampling best separated the four vegetation density categories obtained by the automated method. For this, we calculated the coefficient of variation of the distances between each points obtained in area resulting from the different neighboring samples taken in the observation plot. Because density results have no direct relationship with the surface, the coefficients of variation were calculated inversely proportional to the surface areas of the resulting samples tested. To interpret the results obtained for the coefficient of variation (Cv), the starting point is that the lesser the Cv, the greater the homogeneity. Data preparation for this statistical analysis was performed with GIS tools for distance calculation, overlaying and statistical summaries using the vegetation vector points and their proximity and overlapping in the area resulting from the different neighboring samples.

C v = \frac{σ d i s t}{\bar{x} d i s t} / a r e a

where Cv is the coefficient of variation, $σ d i s t$ is the standard deviation distance between central points in the resulting areas, and $\bar{x} d i s t$ is the mean distance between central points in the resulting areas.

3.3 Final index

The results of the two analyses performed (spatial and statistical) were integrated into a final index (S). This index should contain, first, a greater uniformity between the distances between vegetation points (i.e. a low Cv) and, second, the highest percentage of occupation coincident with the visual analysis (Cva). The difference between the two variables should determine an optimal neighborhood sampling, corresponding to that which obtained the lowest Cv and the highest level of coincidences with visual analysis (Cva). This index is defined as follows

S = C v a - C v

where S is the final index of the sampling, Cv is the coefficient of variation of the sampling, and Cva is the agreement with the visual analysis.

The optimal sampling vegetation density is applied finally to the rest of the aeolian sedimentary systems of La Graciosa and to the pilot area to see if it is possible to apply this method to historical aerial photographs.

III. Results

1. Empirical method result

Figure 7 (right) shows the results of the empirical method applied to the 200 × 200 m pilot plot. As can be seen, category 1 corresponds to low densities (bare sand sheet and isolated individual plants). The intermediate categories (2 and 3) correspond to areas where the distribution of vegetation is more concentrated. The difference between the two is the distance between shrubs, a variable used to calculate the vegetation density. Finally, the category of high density (4) includes areas completely covered by vegetation, without bare sands.

2. Automated method: results of the different neighboring samples

Figure 6 shows how the results of the different DVDM vary with respect to the neighboring sample used, since each of the four defined vegetation density categories share different surfaces, depending on the neighborhood sampling. To understand the information provided by these results, an occupancy rate of vegetation located in the interior of each resulting area was calculated for each area obtained as a result of the samplings. Finally, an average occupancy rate was obtained for each category. In the same way, the mean distance between shrubby individuals through vector points was also calculated.

Table 2 shows the mean percentage occupied by vegetation per area in each category obtained for the neighborhood sampling used and the mean distance between shrubby individuals. To make the results easier to interpret, the vegetation density values (number of individuals per area) have been converted into percentage of vegetation cover.

3. Spatial and statistical analysis

The comparison between the results obtained with the empirical and the automated methods was performed using geoprocessing tools, integrated in GIS (Figure 7). The highest percentage of matching surfaces classified in the same category in both methods (Cva column, Table 3), corresponds to a neighborhood sampling of 9x9, and was 66% (Cva column, row 4, Table 3).

Table 3.

Coefficient of variation inversely proportional to the surface (Cv), percent coincidence with the visual analysis method (Cva), and final index of each sampling performed (S).

Sampling	Cva	Cv	S
1×1	0.2912419	0.0000995	0.2911424
3×3	0.5405697	0.0021797	0.5383900
6×6	0.6433146	0.0154515	0.6278631
9×9	0.6601342	0.0132641	0.6468701
12×12	0.6446270	0.0223336	0.6222934
15×15	0.6199838	0.0271803	0.5928035
18×18	0.6113266	0.0219582	0.5893684
21×21	0.6008308	0.0091786	0.5916522
24×24	0.5891559	0.1927782	0.3963777

The results of the statistical analysis based on the coefficient of variation inversely proportional to the surface (Cv column, Table 3) show the lowest values for the 1x1 sampling, as no neighborhood analysis was actually performed. The 3×3, 9×9, and 21×21 samples were best suited to the aims of the present study, though the neighborhood samplings of 3×3 and 21×21 had lower Cva values than the 9×9 sampling. Finally, the results of the final index (S column, Table 3) show that the optimum neighborhood sampling is 9 × 9 m.

4. Application of the method (9×9) to 1954 and 1987 aerial photographs in the observation plot and determination of the evolution of the vegetation density

The application of the automated method described above for the observation plot shows that this procedure can be used to work with historical aerial photographs in white and black (Figure 8), which is very important with respect to the vegetation of arid aeolian sedimentary systems, characterized by low height and coverage. Once the pixels that represent the vegetation are identified through dark colors by the reclassification method used (Jenks natural breaks), and then vectorized to points, as explained above, the procedure is fast, inexpensive and easy to perform with any GIS software (the algorithms used are quite common and high technical skills are not required). As for the results, the optimal DVDM was considered for both dates and reclassified in the four vegetation density categories studied. Although the information shown is the vegetation density, due to the relationship between vegetation and sedimentary dynamics, wind flow, geomorphology and other environmental factors, quantified studies of environmental changes could be made in arid aeolian sedimentary systems. One such example is shown in Figure 8.

The procedure developed also allows us to determine the areas that are being colonized by vegetation, and more importantly, the degree of plant colonization that impacts on the aforementioned environmental factors (sedimentary dynamics, wind flow, geomorphology). In addition, based on the information obtained the changes that the plot has experienced can be observed where vegetation has increased both in terms of coverage and density. And the way highest densities have rolled the location depending on the amount of vegetation. For example, the highest densities at the beginning (1954), were in the north of the pilot area, and currently have covered principally the center of the observation plot.

5 The vegetation density applied to all the aeolian sedimentary systems of La Graciosa in 2009

The estimate of the vegetation density in the aeolian sedimentary systems of La Graciosa, as the result of the application of the optimum neighborhood sampling of 9 × 9 m (Figure 9 and Table 4), shows the following percentage results: the categories high vegetation density (3) and very high vegetation density (4) occupy more than 36%; the category medium vegetation density (2) accounts for more than 41% (this category coincides with areas that are being colonized by vegetation); and, finally, the category of low vegetation density (1) represents just over 22%, lower than what might be expected for the characteristic dry environmental conditions. In short, we can say that the aeolian sedimentary systems of La Graciosa presented, in February 2009, a significant vegetation cover. Considering that the analysis was performed on a winter-based orthophoto, the vegetation cover is near its yearly maximum value.

Table 4.

Percentage occupied by each category as a function of the vegetation density in aeolian sedimentary systems in La Graciosa in 2009.

Vegetation density categories	Area (%)
1 (MO:10.65; MD: 2.23)	22.17
2 (MO:22.35; MD: 1.79)	41.79
3 (MO:49.26; MD: 1.31)	27.37
4 (MO:84.25; MD: 1.00)	8.67

MO: mean percentage of occupation of the vegetation by category studied (%); MD: mean distance between shrubby individuals (m).

IV. Discussion

The proposed method to classify the vegetation density of arid aeolian sedimentary systems has significant advantages over other existing methods, as well as some limitations. The three main advantages are as follows.

First, as with any automated method, the aim is to obtain results not conditioned by subjectivity, as is the case of visual analysis (Kadmon and Harari-Kremer, 1999; Morgan et al., 2010). So, when a photographic interpretation is used to analyze certain elements, the analyst must make decisions to delimit each degree of density. The probability that another analyst would interpret the reality differently is quite high. In this respect, the training and experience of the analyst determines the accuracy of the analysis and, therefore, of the results obtained (Morgan et al., 2010). In the method developed in this paper, density calculations are made from simple algorithms implemented in any GIS (free or commercial), providing results which are not conditioned by any analyst.

From a technical point of view, two advantages of the method can be emphasized. Firstly, it can be easily implemented because it can be applied using any GIS software, both commercial and free, which implements simple classification algorithms like those used in this case. Secondly, the advantage of this method compared with other commonly used techniques which are based on the digital analysis of satellite or airborne imagery lies in its degree of applicability. The extent of this applicability is based on the non-requirement of near-infrared bands when using this method, and is centered on four factors: (a) Use of the green band of the visible spectrum allows the use of commonly available images and less costly as digital images with infrared bands; (b) From a global perspective, there are more available images of aeolian sedimentary systems of the planet captured by this type of conventional sensor; (c) The historical record of aerial photography is longer than that of satellite images (Kadmon and Harari-Kremer, 1999; Morgan et al., 2010). In this sense, this method is adapted to the availability of these historical images to be applied to diachronic vegetation studies. To date, this type of vegetation density classification using aerial photographs has been mainly performed from photo interpretation (Shanmugam and Barnsley, 2002), or automatic classifications, which link the digital value to vegetation density with validation in the field but the vegetation density value obtained were relative and not actual (Kutiel et al., 2004); and (d) Finally, we consider that the proposed method has advantages over other commonly used methods which are based on the calculation of vegetation indices, especially NDVI and SAVI. These vegetation indices do not calculate different types of vegetation density, but rather the vigor of the vegetation, according to its reflectivity (Carlson and Ripley, 1997; Elmore et al., 2000; Lambin and Ehrlich, 1997; Schmidt and Karnieli, 2000). In this respect, it should be noted that in arid regions, like the coastal areas of the Canary Islands, the vegetation has a marked seasonality. Thus, during the dry season, the plants show significant variations in biomass as a strategy to prevent water loss through evapotranspiration (partial or total loss of leaves, reduced size, etc.), presenting an aspect of scarce vigor. Therefore, calculation of these vegetation indices in these circumstances cannot detect all the existing vegetation, but only that with chlorophyll in large leaf surfaces, missing those plants with a reduced leaf surface. This type of vegetation, however, has an important role, from the environmental point of view, in the stabilization of sandy deposits and in the formation of aeolian landforms. In view of the above, implementation of this simple method in other regions of the world with similar environmental conditions can be used as the basis for a more comprehensive environmental analysis in which the most important question would be the relationship between vegetation cover and aeolian landforms. This is of considerable interest, because the interaction between vegetation and aeolian sedimentary dynamics is a major explanatory factor for understanding dune systems and the consequences of the changes they undergo as the result of both natural and anthropic causes. In this sense, the DVDM presented in the present paper could open up future lines of research in CFD modeling where precise simulations have not yet been obtained in vegetated landforms. It is important to note that the distance between plant individuals conditions the airflow path: the shorter the distance between plant individuals the greater the disruption to the airflow and, consequently, the greater the likelihood of sedimentary accumulation. The density concept proposed in this paper considers the distance factor and the number of plant individuals that spatially occupy an arid aeolian sedimentary system. Implementing the DVDM with additional variables such as plant height or plant porosity from photogrammetric techniques or LiDAR sensors would add even further to an understanding of how a fluid such as air is conditioned by the vegetation. These are very important aspects because the vegetation represents a roughness for the airflow conditioning sediment transport (Smyth, 2016). These systems can thus be monitored from a historical perspective, given the possibility of identifying natural or induced processes. Such analyses of aeolian sedimentary systems may help to explain some features of global change. From an ecological point of view, it is also possible that the categories obtained may have a relationship with the distribution of certain plant communities of La Graciosa, whose survival strategies follow a pattern directly related to their density. In this respect, there may be communities that are more sensitive to high densities and, conversely, others that need low and medium densities to survive in an arid environment where water resources are scarce.

Another key advantage of the application of this simple method is related to the result of greater source availability, the ease of technical implementation and the low cost required for its application: (a) application of this method is less expensive than that based on traditional visual analysis (Kadmon and Harari-Kremer, 1999), as no lengthy training is required; (b) images with visible bands are cheaper than those with infrared bands, and in many cases, depending on the processing, can achieve greater spatial resolution; (c) algorithms are easily implemented using GIS free software. These aspects are very important because by using this method studies can be undertaken in countries with less research funding or less technical qualification in environmental management. Furthermore, such countries tend to have the highest percentage of occupation of aeolian sedimentary systems. This can be seen in Figure 10 which relates their percentage of occupation (Kelso and Patterson, 2012) with the gross domestic product (GDP) of each country (The World Bank, 2015). Consequently, it can be argued that studies on aeolian sedimentary systems, and especially their vegetation, are relevant for the future in the actual scenario of global change.

Figure 10.

Distribution of countries with aeolian sedimentary systems (percentage of occupation of aeolian sedimentary systems in the countries)/gross domestic product (GDP).

Three limitations have been detected for this method, two of them regarding the use of the green band of the orthophoto and the other related to the use of the tool for calculating densities.

This method is only applicable in sandy systems where the sands have a high reflectance in the visible spectrum. In this respect, we start from the premise that the lowest digital levels correspond to vegetation, while the highest digital levels correspond to sand. Therefore, in areas where other low reflectance coverings appear, as may be the case of rocks, or where the sand presents low reflectance levels (volcanic, for example), it may not be easy to distinguish between the types of covering, at least using only a visible band. In this regard, other complementary analyses, as for example calculations of roughness or texture, could be useful to improve the classification.

Another limitation is the result of the size of herbaceous therophyte vegetation, which is not discriminated by the proposed method. For this reason, in arid dune systems, it is important to remember the role played by seasonal therophytic vegetation in stabilized dunes and nebkha dune fields (Hernández-Cordero et al., 2015b).

The final limitation we have detected is related to the tool that calculates the density of points. As the size of the sampling unit is increased, the results become more generalized. This requires clarification, in each case, of the optimal sampling unit size, although in this work we have already made a more precise approximation.

V. Conclusions

1. Aspects related to the usefulness of the method

A simple procedure to automatically classify vegetation density in arid sandy systems using digital orthophotos has been developed, termed the digital vegetation density model (DVDM). This method allows the use of historical information sources, such as aerial photographs in natural color or black and white, enabling studies to be undertaken on evolution of the vegetation by using the density variable. This method provides an alternative to those used to date, including those which need more expensive and less historical recorded information, as satellite images with near-infrared bands, as well as traditional photo-interpretation methods with a greater workload and heavy reliance on the analyst’s subjectivity.

2. Aspects related to the application of the method

The procedure requires an optimal neighborhood sampling unit size, determined in our case to be 9 × 9 m². This optimal neighborhood sampling size was also applied to two historical aerial photographs to check its effectiveness when using black and white images. The changes in vegetation density between 1954, 1987, and 2009 can be seen and quantified objectively. The use of GIS software, commercial or free, facilitates its application as well as helping to understand the patterns of spatial distribution of density degrees, relating this variable with aeolian sedimentary dynamics.

3. Aspects related to the results and their usefulness

The results allowed characterization of the vegetation density in the aeolian sedimentary systems of La Graciosa (Canary Islands, Spain) through application of the DVDM. Given the relationship between vegetation and aeolian sedimentary processes in dune systems, the results are of particular interest for management of such natural areas, especially to avoid human impact in those areas considered more sensitive.

Footnotes

Acknowledgements

This research has been made thanks to the funding of the Organismo Autónomo Parques Nacionales of Spain (the “Isla de La Graciosa” center), managed by the ULPGC and University Foundation of Las Palmas (FULP), as well as the ULPGC funding in postdoctoral programs. Also is a contribution of the projects CSO2013-43256-R and CSO2016-79673-R of the Plan for R+D+i (innovation) of the Spanish Government co-financed with ERDF funds. L. García-Romero, Ph.D. student in the Doctoral Program in Oceanography and Global Change (IOCAG-ULPGC) is supported by a research contract of the Agency of Research, Innovation and Information Society of the Canary Islands Government also co-financed with ERDF funds.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Projects of the Plan for R+D+i (innovation) of the Spanish Government co-financed with ERDF funds: Caracterización de procesos socio-ecológicos de los sistemas playa-dunas de Canarias como base para su gestión sostenible (CSO2013-43256-R) and Análisis de procesos naturales y humanos asociados a los sistemas playa-duna de Canarias (CSO2016-79673-R).

References

Akashi

Mueller-Dombois

(1995) A landscape perspective of the Hawaiian rain forest dieback. Journal of Vegetation Science 6: 449–464.

Avery

Berlin

(1992) Fundamentals of Remote Sensing and Airphoto Interpretation. 5th ed. Upper Saddle River, NY: Prentice Hall.

Bannari

Morin

Bonn

. (1995) A review of vegetation indices. Remote Sensing Reviews 13: 95–120.

Barry

Bowman

DMJS

(2006) Postcards from the past: charting the landscape-scale conversion of tropical Australian savanna to closed forest during the 20th century. Landscape Ecology 21: 1253–1266.

Berger

Iams

(1996) Geoindicators: Assessing Rapid Environmental Changes in Earth Systems. Rotterdam: A.A. Balkema.

Cabrera-Vega

Cruz

Hernández-Calvento

. (2013) Morphological changes in dunes as an indicator of anthropogenic interferences in arid dune fields. Journal of Coastal Research 65(SI): 1271–1276.

Canfield

(1941) Application of line interception in sampling range vegetation. Journal of Forestry 39: 388–394.

Carlson

Ripley

(1997) On the relation between NDVI, fractional vegetation cover, and leaf area index. Remote Sensing of Environment 62: 241–252.

Carmel

Kadmon

(1998) Computerized classification of Mediterranean vegetation using panchromatic aerial photographs. Journal of Vegetation Science 9: 445–454.

10.

Carter

RWG

(1988) Coastal Environments. An Introduction to the Physical, Ecological and Cultural Systems of Coastlines. London: Academic Press.

11.

Chen

Vaglio Laurin

Battles

. (2012) Integration of airborne LiDAR and vegetation types derived from aerial photography for mapping aboveground live biomass. Remote Sensing of Environment 121: 108–117.

12.

Chuvieco

(2010) Teledetección Ambiental: la observación de la Tierra desde el espacio. Barcelona: Ariel.

13.

Collin

Long

Archambault

(2010) Salt-marsh characterization, zonation assessment and mapping through a dual-wavelength LiDAR. Remote Sensing of Environment 114: 520–530.

14.

Collin

Long

Archambault

(2012) Merging land-marine realms: spatial patterns of seamless coastal habitats using a multispectral LiDAR. Remote Sensing of Environment 123: 390–399.

15.

Coppedge

Engle

Fuhlendorf

. (2001) Landscape cover type and pattern dynamics in fragmented southern Great Plains grasslands, USA. Landscape Ecology 16: 677–690.

16.

Costa

CSB

Cordazzo

Seeliger

(1996) Shore disturbance and dune plant distribution. Journal of Coastal Research 12: 133–140.

17.

De Lange

Van Til

Dury

(2004) The use of hyperspectral data in coastal zone vegetation monitoring. EARSeL eProceedings 3: 143–153.

18.

De la Nuez

Quesada

Alonso

(1997) Los volcanes de los islotes al norte de Lanzarote. Lanzarote, Spain: Fundación César Manrique, Arrecife.

19.

Delgado-Fernández

Mir-Gual

(2015). Incremento de la cobertura vegetal en las dunas de Sefton, NO de Inglaterra. Geo-temas 15: 9–12.

20.

Daubenmire

(1959) A canopy coverage method of vegetation analysis. Northwest Science 33: 43–64.

21.

Dirzo

García

(1992) Rates of deforestation in Los Tuxtlas, a neotropical area in southeast Mexico. Conservation Biology 6: 84–90.

22.

Dunn

Sharpe

Guntenspergen

. (1990) Methods for analyzing temporal changes in landscape pattern. In Turner

Gardner

(eds) Quantitative Methods in Landscape Ecology. New York: Springer-Verlag, 173–198.

23.

Elmore

Mustard

Manning

. (2000) Quantifying vegetation change in semiarid environments: precision and accuracy of spectral mixture analysis and the normalized difference vegetation index. Remote Sensing of Environment 73: 87–102.

24.

Floyd

Anderson

(1987) A comparison of three methods for estimating plant cover. Journal of Ecology 75: 221–228.

25.

Folk

Ward

(1957 Brazos river bar. A study in significance of grain size parameters. Journal of Sedimentology and Petrology 27: 3–26.

26.

García

Pérez-Chacón

Hernández

. (2012) Distribución espacial de la cobertura vegetal en el sistema sedimentario eólico de la isla de La Graciosa (islas Canarias, España). In: Cunill

Pelachs

Pérez-Obiol

. (eds) Las Zonas de Montaña: Gestión y Biodiversidad. Barcelona: Fundació Catalunya Caixa, 139–144.

27.

García-Romero

Hernández-Cordero

Fernández-Cabrera

. (2016) Urban-touristic impacts on the aeolian sedimentary systems of the Canary Islands: conflict between development and conservation. Island Studies Journal 11(1): 91–112.

28.

Gaylord

Stetler

(1994) Eolian-climatic thresholds and sand dunes at the Hanford site, south-central Washington, USA. Journal of Arid Environments 28: 95–116.

29.

González

Morín

Acosta

(1996) La Graciosa. Estudio histórico y geográfico. Santa Cruz de Tenerife, Spain: Centro de la Cultura Popular Canaria.

30.

Goodall

(1952) Some considerations in the use of point quadrats for the analysis of vegetation. Australian Journal of Scientific Research 5: 1–41.

31.

Green

Cummins

Right

. (1993) A methodology for acquiring information on vegetation succession from remotely sensed imagery. In: Haines-Young

Green

Cousins

(eds) Landscape Ecology and GIS. London: Taylor and Francis, 111–128.

32.

Greig-Smith

(1983) Quantitative Plant Ecology. Berkeley, CA: University of California Press.

33.

Hayden

Santos

MCFV

Shao

. (1995) Geomorphological controls of coastal vegetation at the Virginia Coast Reserve. Geomorphology 13: 283–300.

34.

Hall

Foster

Verbyla

. (1998) Assessment of snow-cover mapping accuracy in a variety of vegetation-cover densities in Central Alaska. Remote sensing of Environment 66: 129–137.

35.

Harrington

Sanderson

(1994) Recent contraction of wet sclerophyll forest in the wet tropics of Queensland due to invasion by rainforest. Pacific Conservation Biology 1: 319–327.

36.

Hellesen

Levin

(2014) Methodology to estimate loss of semi-natural grasslands due to shrub encroachment in Denmark from 1965 to 2010: a sample-based study using dot grids on aerial photographs. Journal of Land Use Science 9(3): 331–348.

37.

Hernández

Teresa-Serra

Faúndez-Yancas

(2000) Manual de Métodos y Criterios para la Evaluación y Monitoreo de la Flora y la Vegetación. Santiago, Chile: Facultad de Ciencias Forestales, Universidad de Chile.

38.

Hernández-Calvento

Jackson

DWT

Medina

. (2014) Downwind effects on an arid dunefield from an evolving urbanised area. Aeolian Research 15: 301–309.

39.

Hernández-Cordero

Gracia-Prieto

Hernández-Calvento

. (2015a). Proposal for new EU habitats associated with coastal dune fields of the Macaronesian region. A case study in the Canary Islands (Spain). Journal of Coastal Conservation 19: 213–225.

40.

Hernández-Cordero

Hernández-Calvento

Pérez-Chacón Espino

(2017) Vegetation changes as an indicator of impact from tourist development in an arid transgressive coastal dune field. Land Use Policy 64: 479–491.

41.

Hernández-Cordero

Pérez-Chacón Espino

Hernández-Calvento

(2015b) Vegetation, distance to the coast, and aeolian geomorphic processes and landforms in a transgressive arid coastal dune system. Physical Geography 36(1): 60–83.

42.

Hesp

(1981) The formation of shadow dunes. Journal of Sedimentary and Petrology 51(1): 101–112.

43.

Hesp

(1983) Morphodynamics of incipient foredunes. In: Brookfield

Ahlbrandt

(eds) Eolian Sediment and Processes. Elsevier, 325–342.

44.

Hesp

(1984) The formation of sand “beach ridges” and foredunes. Search 15(9–10): 289–291.

45.

Hesp

(1989) A review of biological and geomorphological processes involved in the initiation and development of incipient foredunes. In: Gimmingham

Ritchie

Willetts

. (eds) Coastal Sand Dunes. Edinburgh: Royal Society of Edinburgh, 181–201.

46.

Hesp

(1991) Ecological processes and plant adaptations on coastal dunes. Journal of Arid Environments 21: 165–191.

47.

Hesp

(2002) Foredunes and blowout: initiation, geomorphology and dynamics. Geomorphology 48: 245–268.

48.

Hesp

(2013) Conceptual models of the evolution of transgressive dunefield systems. Geomorphology 199: 138–149.

49.

Hudak

Wessman

(1998) Textural analysis of historical aerial photography to characterize woody plant encroachment in South African savanna. Remote Sensing of Environment 66: 317–330.

50.

Huete

(1988) A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment 25: 295–309.

51.

Hugenholtz

Levin

Barchyn

. (2012) Remote sensing and spatial analysis of aeolian sand dunes: a review and outlook. Earth-Science Reviews 111: 319–334.

52.

Instituto Nacional de Estadística (2014) Gobierno de España. Available at: http://www.ine.es/ (accessed 25 May 2014).

53.

Jackson

Nordstrom

(2011) Aeolian sediment transport and landforms in managed coastal systems: a review. Aeolian Research 3(2): 181–196.

54.

Jauhiainen

Holopainen

Rasinmäki

(2007) Monitoring peatland vegetation by means of digitized aerial photographs. Scandinavian Journal of Forest Research 22: 168–177.

55.

Johnston

Naiman

(1990) The use of a geographic information system to analyze long-term landscape alteration by beaver. Landscape Ecology 4: 5–20.

56.

Kadmon

Harari-Kremer

(1999) Studying long-term vegetation dynamics using digital processing of historical aerial photographs. Remote Sensing of Environment 68: 164–176.

57.

Kelso

Patterson

(2012) World land-based polygon features, 1:10 million (2012). Made with Natural Earth. Available at: http://www.naturalearthdata.com (accessed ?).

58.

Konlechner

Hilton

Arens

(2014) Transgressive dune development following deliberate de-vegetation for dune restoration in the Netherlands and New Zealand. Dynamic Environmental 33: 141–154.

59.

Kuriyama

Mochizuki

Nakashima

(2005) Influence of vegetation on aeolian sand transport rate from a backshore to a foredune at Hasaki, Japan. Sedimentology 52: 1123–1132.

60.

Kutiel

Cohen

Shoshany

. (2004) Vegetation establishment on the southern Israeli coastal sand dunes between the years 1965 and 1999. Landscape and Urban Planning 67: 141–156.

61.

Laliberte

Rango

Havstad

. (2004) Object-oriented image analysis for mapping shrub encroachment from 1937 to 2003 in southern New Mexico. Remote Sensing of Environment 93(2): 198–210.

62.

Lambin

Ehrlich

(1997) Land-cover changes in sub-Saharan Africa (1982–1981): application of a change index based on remotely sensed surface temperature and vegetation indices at a continental scale. Remote Sensing of Environment 61:181–200.

63.

Lancaster

Baas

ACW

(1998) Influence of vegetation cover on sand transport by wind: field studies at Owens Lake, California. Earth Surface Processes and Landforms 23: 69–82.

64.

Levin

Kidron

Ben-Dor

(2006) The spatial and temporal variability of sand erosion across a stabilizing coastal dune field. Sedimentology 53: 697–715.

65.

Levin

Kidron

Ben-Dor

(2008) A field quantification of coastal dune perennial plants as indicators of surface stability, erosion or deposition. Sedimentology 55(4): 751–772.

66.

Luna MCM de

Parteli

EJR

Durán

. (2011) Model for the genesis of coastal dune fields with vegetation. Geomorphology 129: 215–224.

67.

Mangas

Pérez-Chacón

Hernández-Calvento

. (2012) Procedencia y características texturales de las arenas eólicas costeras de la isla de La Graciosa (Islas Canarias). Geotemas 13: 577–580.

68.

Meco

Ballester

Betancout

. (2006). Paleoclimatología del Neógeno en las islas Canarias. Las Palmas, Spain: Ministerio de Medio Ambiente y ULPGC.

69.

Miller

Schulz

Thompson Hobbs

. (1995) Changes in the landscape structure of a Southeastern Wyoming Riparian Zone following shifts in stream dynamics. Biology Conservation 72: 371–379.

70.

Moreno-Casasola

(1986) Sand movement as a factor in the distribution of plant communities in a coastal dune system. Vegetatio 65: 67–76.

71.

Morgan

Gergel

Coops

(2010) Aerial photography: a rapidly evolving tool for ecological management. BioScience 60: 47–59.

72.

Mostacedo

Fredericksen

(2000) Manual de métodos básicos de muestreo y análisis en ecología vegetal. Available at: http://www.bionica.info/biblioteca/Mostacedo2000EcologiaVegetal.pdf (accessed ?).

73.

Nordstrom

(1994) Beaches and dunes of human-altered coasts. Progress in Physical Geography 18(4): 497–516.

74.

Nurminen

Litkey

Honkavaara

. (2015) Automation aspects for the georeferencing of photogrammetric aerial image archives in forested scenes. Remote Sensing 7: 1565–1593.

75.

Okeke

Karnieli

(2006) Methods for fuzzy classification and accuracy assessment of historical aerial photographs for vegetation change analyses. Part I. Algorithm development. International Journal of Remote Sensing 27: 153–176.

76.

Okin

(2008) A new model of wind erosion in the presence of vegetation. Journal of Geophysical Research 113: F02S10.

77.

Ortiz

Torres

Yanes

. (2006) Climatic cycles inferred from the aminostratigraphy and aminochronology of Quaternary dunes and palaeosols from the eastern islands of the Canary Archipelago. Journal of Quaternary Science 21: 287–306.

78.

Pérez-Chacón

Hernández-Calvento

Fernández-Negrín

. (2010) Caracterización del sistema sedimentario eólico de La Graciosa (Archipiélago canario). Report for the National Parks Body.

79.

Platt

Schoennagel

(2009) An object-oriented approach to assessing changes in tree cover in the Colorado Front Range 1938–1999. Forest Ecology and Management 258(7): 1342–1349.

80.

Qin

Zhu

Zhang

. (2006) Review of vegetation covering and its measuring and calculating method. J Northwest Sci-Tech University of Agriculture and Forestry (Natural Science Edition) 34(9): 163–170.

81.

Ritzen

Straatsma

(2002) Laseraltimetrie en hydraulische vegetatieruwheid van uiterwaarden. Master’s Thesis, Universiteit Utrecht, the Netherlands.

82.

Rozenstein

Siegal

Blumberg

. (2016) Investigating the backscatter contrast anomaly in synthetic aperture radar (SAR) imagery of the dunes along the Israel–Egypt border. International Journal of Applied Earth Observation and Geoinformation 46: 13–21.

83.

Santana-Cordero

Monteiro-Quintana

Hernández-Calvento

. (2016) Long-term human impact on the coast of La Graciosa, Canary Islands. Land Degradation and Development 27: 479–489.

84.

Schmidt

Karnieli

(2000) Remote sensing of the seasonal variability of vegetation in a semi-arid environment. Journal of Arid Environments 45: 43–59.

85.

Shanmugam

Barnsley

(2002) Quantifying landscape-ecological succession in a coastal dune system using sequential aerial photography and GIS. Journal of Coastal Conservation 8: 61–68.

86.

Sherman

Bauer

(1993) Dynamics of beach-dune systems. Progress in Physical Geography 17(4): 413–447.

87.

Simpson

Boerner

REJ

De Mers

. (1994) Forty-eight years of landscape change on two contiguous Ohio landscapes. Landscape Ecology 9: 261–270.

88.

Smyth

TAG

(2016) A review of computational fluid dynamics (CFD) airflow modelling over aeolian landforms. Aeolian Research 22: 153–164.

89.

Soria

Salomón

Abraham

. (2007) Estimación de la cobertura vegetal en el Departamento de Lavalle, Mendoza. Mar del Plata, Argentina. In: XII Congreso de la Asociación Española de Teledetección, Mar del Plata, Argentina, 19-21 September 2007. Valencia, Spain: Asociación Española de Teledetección; Buenos Aires, Argentina: Universidad Nacional del Centro de la Provincia de Buenos Aires.

90.

Stallins

(2002) Dune plant species diversity and function in two barrier island biogeomorphic systems. Plant Ecology 165: 183–196.

91.

Tsoar

(2005) Sand dunes mobility and stability in relation to climate. Physica A: Statistical Mechanics and its Applications 357: 50–56.

92.

Tsoar

Blumberg

(2002) Formation of parabolic dunes from barchan and transverse dunes along Israel’s Mediterranean coast. Earth Surface Processes and Landforms 27: 1147–1161.

93.

Turner

Wong

Chew

. (1996) Rapid assessment of tropical rain forest successional status using aerial photographs. Biology Conservation 77: 177–183.

94.

Underwood

Ustin

Dipietro

(2003) Mapping nonnative plants using hyperspectral imagery. Remote Sensing of the Environment 86: 150–61.

95.

Warmink

(2007) Vegetation Density Measurements using Parallel Photography and Terrestrial Laser Scanning. Utrecht, the Netherlands: Department of Physical Geography, Faculty of Geosciences, Utrecht University.

96.

Woodcock

Collins

Gopal

. (1994) Mapping forest vegetation using Landsat TM imagery and a canopy reflectance model. Remote Sensing of Environment 50: 240–254.

97.

The World Bank (2015) World Development Indicators. London: The World Bank.

98.

Zehm

Nobis

Schwabe

(2003) Multiparameter analysis of vertical vegetation structure based on digital image processing. Flora 198: 142–160.

99.

Zeng

Cao

. (2013) Dynamic monitoring of plant cover and soil erosion using remote sensing, mathematical modeling, computer simulation and GIS techniques. American Journal of Plant Sciences 4: 1466–1493.