- Email: [email protected]
NDVI as an indicator for changes in water availability to woody vegetation
Ecological Indicators 23 (2012) 290–300
Contents lists available at SciVerse ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
NDVI as an indicator for changes in water availability to woody vegetation Cristina Aguilar a,∗ , Julie C. Zinnert b,c,1 , María José Polo a , Donald R. Young c a b c
Fluvial Dynamics and Hydrology Research Group, IISTA-University of Cordoba, Rabanales Campus, Leonardo Da Vinci Building, 14071 Córdoba, Spain US Army ERDC, Fluorescence Spectroscopy Lab, 7701 Telegraph Road, Alexandria, VA 22315, USA Department of Biology, Virginia Commonwealth University, Richmond, VA 23284, USA
a r t i c l e
i n f o
Article history: Received 15 November 2011 Received in revised form 29 March 2012 Accepted 3 April 2012 Keywords: NDVI Precipitation Freshwater lens Water table depth Plant stress Airborne Landsat TM
a b s t r a c t Barrier islands shrub thickets, the dominant woody community of many Atlantic coast barrier islands, are very sensitive to changes in the freshwater lens and thus, constitute a strong indicator of summer drought. NDVI was computed from airborne images and multispectral images on Hog Island (VA, USA) to evaluate summer growing season changes in woody communities for better predictions of climate change effects. Patterns of NDVI were compared year to year and monthly relative to precipitation and water table depth at the appropriate temporal scale. The highest absolute values of NDVI as well as the larger surface covered by woody vegetation (NDVI > 0.5) occurred in the wet year (2004) with a bimodal distribution of NDVI values (around 0.65 and 0.9) while both dry years (2007 and 2008) showed similar values in maximum, mean and standard deviation and unimodal distributions (around 0.75) of NDVI values. Positive linear adjustments were obtained between maximum (r2 > 0.9) and mean NDVI (r2 > 0.87) and the accumulated rainfall in the hydrological year and the mean water table depth from the last rainfall event till the date of the image acquisition. The spatial variations revealed that water table depth behaved different in wet and dry years. In dry years there was a remarkable increase in mean and maximum values linearly related to water table depth. The highest slope of the adjustment in 2007 indicated a sharp response of vegetation in the driest year. Monthly series of NDVI showed the major role of lack of precipitation through July and August in 2007 with missing classes of NDVI above 0.8 and unimodal distributions in mid-late summer. Best linear fits (r2 close to 1) were obtained with rainfall at different temporal scales: accumulated rainfall in the hydrological year 2004 and accumulated rainfall in the last month previous to the date of 2007 image. Thus, in dry years productivity is closely related to water available from recent past as opposed to over the year for wet years. Good fits (r2 values higher than 0.88) were obtained between monthly decrease in water table depth and NDVI variables just in the dry year. These results demonstrate the important feedback between woody vegetation response to changes in the freshwater lens using empirical data and could apply to other systems with strong directional gradients in resources. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction Coastal system ecological processes are closely coupled to both atmospheric and oceanic drivers, all of which may be influenced by climate change. These ecosystems may be the most sensitive indicators of changing climate (Feagin et al., 2010). Sandy soils typical of North American Atlantic coast barrier islands have minimal water holding capacity which affect distribution and primary production of terrestrial communities. These communities are dependent on access to the soil freshwater lens, which varies temporally and spatially across the landscape. As summer evapotranspiration increases, freshwater capacity in the soil is reduced and may
∗ Corresponding author. Tel.: +34 957212662; fax: +34 957212097. E-mail address: [email protected] (C. Aguilar). 1 Julie C. Zinnert was formerly Julie C. Naumann. 1470-160X/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ecolind.2012.04.008
lead to associated drought stress and limited growth. In addition to evapotranspiration, rainfall, and limited groundwater recharge during the summer months, spatial variations in water availability are subject to plant water use. These effects are exacerbated by microtopography, with dune crests most susceptible to water stress than the lower elevation swales. Frequency and intensity of summer droughts are expected to increase in response to predicted shifts in global climate patterns (Karl et al., 2009). Shrub thickets represent the dominant woody community of many Atlantic coast barrier islands (Young et al., 2007). The primary species is Myrica cerifera, an evergreen, actinorhizal nitrogen fixing shrub. Interestingly, Myrica is characterized by rapid growth but also responds quickly to both salinity and drought stress (Naumann et al., 2007; Young et al., 1994; Zinnert et al., in press). Distribution is limited to low elevation swales with access to the freshwater lens. Myrica thickets are a strong indicator of summer drought relative to other terrestrial plant communities as these communities are most
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
291
Fig. 1. Study site and location of wells (MB 2, MB 3, MB 5, S4, S3 and S2).
sensitive to changes in the freshwater lens. Methods for indicating areas most vulnerable to extended droughts have not been explored in coastal ecosystems. Ability to assess how environmental changes affect dynamics of vegetation is increasingly important for better predictions of climate change effects. Popularity of the application of the Normalized Difference Vegetation Index (NDVI) in ecological studies has enabled quantification and mapping of green vegetation with the goal of estimating aboveground net primary productivity (ANPP) and other landscape-level fluxes (Pettorelli et al., 2005; Wang et al., 2003). NDVI is based on differences in reflectance in the red region (due to pigment absorption) and maximum reflectance in the nearinfrared (caused by cellular structure); it is the most widely used index in remote sensing. It is closely related to a range of intercorrelated biomass variables such as leaf area index (LAI), leaf cover, chlorophyll per unit ground area, green biomass or green vegetation factor (Filella et al., 2004; Gamon et al., 1995). NDVI saturates easily and is not considered a good estimator of high LAI (Asner et al., 2000; Brantley et al., 2011); however, NDVI still retains ecological relevance as an indicator of green biomass change (Wang et al., 2003). NDVI can be a useful tool to couple climate and vegetation distribution and performance at large spatial and temporal scales (Pettorelli et al., 2005). Because vegetation vigor and productivity are related to hydrological variables (rainfall, evapotranspiration, etc.), NDVI serves as a surrogate measure of these factors at the landscape scale (Groeneveld and Baugh, 2007; Wang et al., 2003). The linear response of vegetation NDVI to rainfall for regions with low vegetation cover and rainfall is well documented (Groeneveld and Baugh, 2007; Ji and Peters, 2003; Kawabata et al., 2001; Malo and Nicholson, 1990; Wang et al., 2003; Yang et al., 1998). Other studies have also analyzed trends of NDVI with other variables such as temperature (Wang et al., 2003), and evapotranspiration (Groeneveld, 2008). However, as relationships between NDVI and climatic factors are location-dependent, more detailed analyses are needed (Wang et al., 2003). Further, the sandy soils
on barrier islands have a limited water holding capacity, thus the response of vegetation NDVI should be related to the groundwater table depth in the growing season. Our goal was to determine the relevance of NDVI as an indicator of water availability to coastal woody vegetation. Specifically, we quantified responses to seasonal changes in precipitation and water availability through the groundwater lens by linking landscape level varitations in relative greeness of woody vegetation to past precipitation and hydrology. Throughout the summer growing season changes in woody productivity were assessed by using NDVI calculated from remote sensing imagery. We compared year to year variations in these patterns relative to variations in precipitation and water table depth at the appropiate temporal scale.
2. Materials and methods 2.1. Study site The study is focused on Hog Island (37◦ 40 N; 75◦ 40 W), a barrier island located on the Eastern shore of VA, USA (Fig. 1). The northern end of the island is broad, with a series of dune lines separated by swales and ponds forming a chronosequence (Hayden et al., 1991). The primary woody vegetation is dense thickets of M. cerifera. Upland grasslands are dominated by Spartina patens and Ammophila breviligulata. Extensive marshes on the lagoon side of the island are dominated by Spartina alterniflora. Site records indicate a mean annual temperature around 15 ◦ C and considerable inter annual variability of precipitation between 850 and 1400 mm per year. This variability renders summer drought a high probability of occurrence (Van Cleve and Martin, 1991). The study period (2004–2008) was determined by the availability of spectral data in the summertime for both wet and dry years.
292
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 2. Monthly rainfall (mm/month) measured at Hog Island, monthly evapotranspiration (mm/month) computed from Hargreaves equation with recorded temperature datasets and mean monthly water table depth (m.a.s.l) registered at well MB 3.
2.2. Hydrological data
2.3. Spectral data
Rainfall and temperature data were obtained on a subdaily basis from a meteorological station on Hog Island (Krovetz et al., 2011). Data records were treated to generate daily and monthly values. Fig. 2 shows the huge differences in monthly rainfall that occurred between the wet year (2004) and the dry years (2007 and 2008), especially in the summer period. From the daily records of maximum and minimum temperatures, daily reference evapotranspiration was computed from Hargreaves equation (Hargreaves et al., 1985) in order to assess the evaporative potential according to the meteorological conditions in the area. Unlike rainfall, reference evapotranspiration did not experience significant changes among the years and just during July and August the rate was lower in the wet year (2004) as expected due to the higher frequency of rainy days that cause saturated conditions in the air (Fig. 2). Similarly, water table depth was assessed on a subdaily basis from datasets recorded at three different wells on Hog Island (Fig. 1). Two out of the three wells were located in the middle of two thickets (MB 3 and MB 5; Fig. 1) and the third one on a high marsh close to another thicket (MB 2; Fig. 1). Data available for 2007 and 2008 from three additional wells (S2, S3 and S4; Fig. 1) were used to broaden the spatial range of water levels in both dry years. These wells are located in swales partially surrounded by thickets. Water table depth from well MB 3 is considered to represent the mean behavior of the water table depth in the island affected by woody communities. Mean monthly values of the water table depth in MB 3 showed a considerable soil moisture deficit in the summer periods of dry years, especially in 2007 (Fig. 2) when rainfall is not able to sustain the evapotranspiration needs. The sandy soils in the area have high hydraulic conductivities, experience rapid drainage and may result in very dry conditions if rainfall in summer is insufficient as in 2007.
Two sources of spectral data were available for this study: airborne and multispectral images for the 2004–2008 period. The first source consists of three airborne missions flown in summer, and the data are used to describe the interannual variations. The second source was Landsat TM5, for the analysis of seasonal changes from selected images during the growing season in 2004 and 2007.
2.3.1. Airborne images Airborne images were used to describe interannual changes. Three airborne hyperspectral images were available in the study area (Brantley et al., 2011; Naumann et al., 2009a,b). The first one corresponded to the spectral mission flown on August 24, 2004 on Hog Island using the Portable Hyperspectral Imager for Lowlight spectroscopy (PHILLS) with a spatial resolution of 4 m/pixel and covering 384–1000 nm. The second image was flown on September 13, 2007 with a ProSpecTIR VIS hyperspectral imaging spectrometer (SpectIR) covering 450–2450 nm and a spatial resolution of 2 m/pixel. The same sensor was used in the third flight on August 19, 2008 providing a spatial resolution of 1 m/pixel. All the images were collected under cloud-free conditions and data were post-processed to minimize geometric and radiometric effects. Calibration was performed using the empirical line calibration method within ENVI (RSI, Inc.) with ground reflectance measurements. Transformation of the PHILLS data was accomplished using the SpectIR data as a reference to normalize the bands. NDVI at landscape-level was calculated on each pixel of the study site from Eq. (1) (Rouse et al., 1974): NDVI =
800 − 670 800 + 670
(1)
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
293
Table 1 Basic statistics of NDVI for the areas with woody communities (NDVI higher than 0.5) and hydrological variables: accumulated rainfall (mm) in the hydrological year (from 1st October to date of the image) and mean water table depth (m.a.s.l) registered at well MB 3 from last rainfall event till the day before the image acquisition. Date
Max
Mean
Stand. Dev.
Total area of woody communities (ha)
Accumulated rainfall (mm)
Water depth (m.a.s.l)
08/24/2004 09/13/2007 08/19/2008
0.97 0.82 0.82
0.75 0.67 0.66
0.13 0.08 0.08
462 179 254
1075.88 482.09 610.76
1.04 0.48 0.63
where 800 and 670 are the reflectance values centered at 800 and 670 nm respectively. 2.3.2. Multispectral images Multispectral imagery was analyzed in order to describe different patterns within the growing season. Images used in this study consisted of scenes from Landsat TM5 (path/row: 14/34) downloaded at the New Earth Explorer Interface for each month in the summer period. Four scenes were acquired for 2004 and 2007, respectively. Cloud-free scenes were selected in order to minimize uncertainties due to heterogeneous atmospheric conditions; however, radiometric calibration and atmospheric correction were applied (Yuan and Niu, 2008). Radiometric calibration was achieved through the equations and rescaling factors in Chander et al. (2009). Atmospheric correction was conducted to retrieve surface reflectance by use of MODTRAN4 as implemented into ENVI FLAASH (FLAASH Module, 2009). Atmospheric properties were defined in terms of aerosol type, water vapor and visibility. In this study the maritime aerosol type and 40 km of visibility were considered for all the images as suggested by ENVI FLAASH. Water vapor content values registered at Wallops site (35◦ N; 75◦ W) from the AERONET network were applied (FLAASH Module, 2009) when no data was available. All scenes were subset to form the image of the study area. NDVI values were computed from Eq. (2) which represents the difference between reflectance in the near-infrared and red bands (bands 4 and 3), 4 and 3 respectively, normalized by the sum of both of them (Tucker, 1979). NDVI =
4 − 3 4 + 3
(2)
were estimated by fitting functions through the data using simple least-squares regression. To show the influence of water availability on vegetation productivity for each date, spatial variations of NDVI were analyzed within each image. Regions were defined in a 20 m buffer area around each well that reflected a homogeneous NDVI value of the thicket around the well. As the wells are free from outside impacts, fluctuations in the water table depth were considered to be linked to natural groundwater use by plants and related to the mean NDVI value in the thicket. Finally, in order to separate effects of soil characteristics and microtopography from the different availability to freshwater in the wet and dry years, a dimensionless NDVI map was calculated dividing the NDVI value of each pixel by the mean NDVI value for the image. 2.4.2. Monthly variations within the growing season A similar methodology was applied with the Landsat 5 TM images. Frequency distribution function for the surfaces of different NDVI classes, cumulative probability distribution function and basic statistics of NDVI values (maximum, mean and standard deviation) were calculated for each image. These computations helped to assess changes in productivity along the growing season within the wet and the dry years. Similarly, adjustments between NDVI and hydrological variables were made for the images within each growing season. Careful consideration in the temporal integration of hydrological variables was taken into account in order to reflect the short-term effects of water availability over the status of the woody cover. 3. Results
2.4. Relations between NDVI and hydrological variables 3.1. Interannual variations This study was restricted to woody communities on the island due to their quick response to groundwater availability. Supervised classification was performed in a previous study using maximum likelihood methodology into five classes (woody, bare sand, grassland, water, and marsh) (Zinnert et al., 2011). Based on NDVI values obtained from Zinnert et al. (2011), 0.5 was used as the NDVI threshold between woody vegetation and other classes. 2.4.1. Interannual variations analysis The area covered with woody communities was classified to describe the frequency distribution function of NDVI for the three images. Surface instead of number of pixels was used in order to compensate different spatial resolution of the three images. Cumulative probability distribution function and basic statistics (maximum, mean and standard deviation) were computed for the sample of pixels. Maximum and mean NDVI values reflected overall and mean productivity and biomass, respectively (Pettorelli et al., 2005), whereas the standard deviation represented a measure of the spatial variability in productivity. In order to quantify the dependence of NDVI values with hydrological regimen, different adjustments with hydrological variables were computed. NDVI trends in terms of the basic statistics were related to variations in precipitation and water table depth. Trends
For 2004, 2007 and 2008, the highest values of NDVI and the largest surface area covered by woody vegetation occurred in the wet year compared to both dry years (Table 1). For pixels with NDVI above 0.5, both dry years showed similar values in the basic statistics considered (maximum, mean and standard deviation) (Table 1). Fig. 3 shows the histogram in terms of surface area with NDVI values higher than 0.5 and the cumulative probability distribution function. In general, a very different distribution can be seen in the wet year when all the classes of NDVI are represented with two modes in NDVI values around 0.65 and 0.9 (Fig. 3). To contrast, in both dry years the surface covered by NDVI above 0.75 is minimal, with just a few pixels with NDVI values higher than 0.8. Compared to 2007 (the driest year), 2008 had more surface area covered in every NDVI class. The same ideas can be drawn from the cumulative distribution function as 2004 is more tilted, indicating higher NDVI values for a certain percentile compared to 2007 and 2008 (e.g., 0.93 in 2004 against around 0.75 in 2007 and 2008 for the 90th percentile). Once relationships between NDVI values and the wet/dry character of the year were demonstrated, NDVI trends were related to hydrological variables. After several trials, the variables considered were accumulated rainfall in the hydrological year (from
294
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 3. Histograms and cumulative probability distribution functions of NDVI areas above 0.5 for each airborne image in 2004, 2007 and 2008.
October 1 of the preceding calendar year to the day before the sensor overpass) and mean water table depth registered at well MB 3 from the last rainfall event before the date of the acquisition (Table 1). The best adjustments were obtained with linear relationships between hydrological variables considered and absolute maximum NDVI and mean NDVI for the area (Fig. 4). Coefficients of determination (r2 ) values above 0.9 were obtained between NDVI maximum values and both variables. Similarly, r2 values above 0.87 were obtained for mean NDVI values. Considering both hydrological variables, slightly better fits resulted with rainfall than with ground water table depth. We assume that the linear trend of adjustments is determined by the availability of 3 years data. Thus, future availability of more hyperspectral images including medium and extreme years in terms of rainfall, would allow inclusion of more than three points in the fits and so, other
types of fits (e.g., exponential, power, etc.) could be tested as well. Spatial variation in NDVI was considered in the area around each well (Table 2). There was a remarkable increase in mean and maximum values moving from west to east of the island in the dry years (2007 and 2008) in the most northern wells (MB 2, MB 3 and MB 4). Mean and maximum NDVI values were linearly related to water table depth during the dry years (Fig. 5) with best fits for the mean values (r2 values above 0.85). The slope of adjustments was highest in 2007 (Fig. 5) which shows the sharp response of vegetation in the driest year. Maximum NDVI values in the dry years could be better approximated to a nonlinear function, as inferred from the trend in Fig. 5. However, the sample of wells is a limiting factor, since conditions around wells S2, S3, and S4 are too similar. Thus, this apparent nonlinear trend should be validated
Fig. 4. Linear adjustments between maximum and mean NDVI of the sample of pixels with NDVI above 0.5 with the accumulated rainfall in the hydrological year (mm) and with the mean water table depth at well MB 3 from the last rainfall event till the date of the airborne image.
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
295
Table 2 Mean and maximum NDVI values in 20 m around the well and mean water table depth (m.a.s.l) from the last rainfall event till the day before the image acquisition. Well
2004
2007
NDVI
MB 2 MB 3 MB 5 S4 S3 S2
Water depth (m.a.s.l)
Mean
Max
0.72 0.87 0.89 0.79 0.90 0.81
0.77 0.94 0.95 0.90 0.95 0.93
1.17 1.04 1.31 No data No data No data
2008
NDVI
Water depth (m.a.s.l)
Mean
Max
0.50 0.66 0.73 0.62 0.67 0.66
0.64 0.75 0.80 0.79 0.77 0.78
against datasets from a denser a more representative network of wells for a complete spatial analysis. Fig. 6 represents pixels with an NDVI value over the mean value in the study site for each year respectively. Dimensionless NDVI maps indirectly isolate the effects of the different rainfall regimen from the slope, soil type and depth, etc. The difference between both figures can be explained by the different rainfall occurrence. Surface with NDVI above the mean value decreased from 221 ha in 2004 to 107 ha in 2007 and 144 ha in 2008. The highest NDVI values in 2004 were about 30% higher of the mean NDVI compared to 23% in 2007 and 25% in 2008. 3.2. Monthly variations within the growing season NDVI monthly series show the role of the absence of precipitation in the months of July and August of 2007 (Fig. 7). As the amount of rainfall is not able to sustain actual evapotranspiration, water table depth falls (Fig. 2) and some woody species are excluded. This is apparent in Fig. 8, where classes above 0.8 are missing in August 2007 and the bimodal distribution is replaced by the typical distribution of months in early summer. The cumulative distribution functions of NDVI showed a very different behavior between both years as well (Fig. 8). In 2004, the function with the lowest NDVI values for a certain percentile corresponds to the first image of the growing season (July), while in 2007 this happens in August, demonstrating the shortage of rainfall during the previous months.
0 0.48 1.10 0.54 0.66 0.34
NDVI
Water depth (m.a.s.l)
Mean
Max
0.55 0.62 0.68 0.61 0.64 0.61
0.68 0.77 0.77 0.76 0.76 0.77
−0.15 0.63 1.51 0.67 0.70 0.60
Linear fits were obtained with rainfall but the best fits were obtained at different temporal scales: the accumulated rainfall in the hydrological year for the wet year and the accumulated rainfall in the last month previous to the date of the image for the dry year. r2 values were very close to 1 in every adjustment of surface area with NDVI higher than a threshold (Fig. 9) and NDVI basic statistics (Fig. 10). Linear fits with accumulated rainfall in the last month previous to the date of the image were carried out for the wet year, however r2 values were relatively low, between 0.25 and 0.7 for all variables considered. Adjustments between NDVI variables (surface of NDVI thresholds and basic statistics) and mean water table depth from the previous rainfall event were poor in the wet year (r2 values between 0.40 and 0.62). However, for the dry year very good fits were obtained between the monthly decrease in water table depth and NDVI variables with r2 values higher than 0.93 for the surface of NDVI thresholds, and from 0.88 to 0.95 for the basic statistics (Fig. 11). 4. Discussion Our results show that NDVI of woody vegetation can be used as an indicator for groundwater availability and precipitation inputs in coastal systems. The drivers of interannual variability in ANPP are of interest for projections of ecosystem responses to climate change (Fang et al., 2001). Knapp and Smith (2001) suggested that fluctuations in precipitation were not related to interannual ANPP at the large-scale (across North America), while Fang et al. (2001) found positive relationships between ANPP derived from NDVI and precipitation. Coastal ecological systems are ideal for quantifying differences in primary productivity due to spatial and temporal variations in climate based stressors (i.e., water availability, salinity) as they are primarily influenced by directional environmental gradients (Ehrenfeld, 1990). Studies that demonstrate the use of remote sensing for deriving ANPP and relationships with climatic variables can allow for rapid, large-scale assessment of ecosystem processes. Our results are notable in documenting the response of high productivity woody vegetation to annual and monthly fluctuations in both precipitation and groundwater availability via NDVI. Additionally, our results are significant for showing these relationships in a mesic environment since numerous studies have shown direct relationships in arid and semiarid ecosystems (Jobbágy et al., 2002; Jobbágy and Sala, 2000; Paruelo et al., 1997), thus extending the capabilities of using NDVI as a metric for water availability at the landscape level. 4.1. Interannual variations
Fig. 5. Linear adjustment between the mean (solid markers) and maximum (empty markers) NDVI values in 20 m around the well and mean water table depth (m.a.s.l) the day before the image acquisition.
In many systems, the relationships between interannual variations in NDVI are more closely linked to interannual variations in temperature over those in precipitation (Wang et al., 2008). In this study, annual changes in NDVI were coupled to changes in
296
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 6. Area in blue with NDVI value above the mean NDVI value per year (Table 1) in the study site. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
precipitation. This change is reflected in lower NDVI values in 2007 and 2008 and in the linear relationships between accumulated rainfall and mean water table depth with the maximum and mean NDVI values as indicators of overall and mean productivity and biomass, respectively. Green leaf area varies with vegetation water content, and since NDVI is an indicator of green leaf area, NDVI decreases as water becomes less available (Claudio et al., 2006; Gamon et al., ˜ 1995; Penuelas et al., 1997; Sims and Gamon, 2003). In the barrier island system, using NDVI as an indicator of groundwater was dependent upon rainfall as water table depth behaved different in wet and dry years. NDVI was a good indicator of groundwater availability in dry years as the freshwater lens became
the main source of water for woody vegetation. Because of low elevation and high transpiration rates of the dominant woody species, M. cerifera, the woody community on Hog Island is closely coupled to the freshwater lens (Shao et al., 1995), which is recharged through precipitation. During wet years, water table depth did not change very much as water uptake by the plants was compensated by rainfall-infiltration events. 4.2. Monthly variations The model MCHOG was developed by Shao et al. (1995) to simulate transpiration of woody vegetation on barrier islands and
Fig. 7. NDVI from Landsat TM5 images along the growing season in 2004 and 2007. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
297
Fig. 8. Histograms and cumulative probability distribution function of NDVI areas above 0.5 for each day considered in the growing season in 2004 and 2007.
Fig. 9. Linear adjustments between surface with NDVI values above a threshold of the sample of pixels with NDVI above 0.5 with the accumulated rainfall in the hydrological year (mm) till the date of the image for 2004 and with the mean rainfall of the previous month (mm/month) for 2007.
298
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 10. Linear adjustments between NDVI statistics of the sample of pixels with NDVI above 0.5 with the accumulated rainfall in the hydrological year (mm) till the date of the image for 2004 and with the mean rainfall of the previous month (mm/month) for 2007.
demonstrated the importance of soil water (via precipitation) to productivity. Their results indicated that the high LAI of M. cerifera and low water holding capacity of sandy soils lead to a rapid decrease in soil water content and reduced transpiration during extended periods without precipitation. To further investigate the coupling between vegetation productivity and water availability considering the low water holding capacity of soils, we examined monthly variations in NDVI to hydrological variables over the growing season.
For barrier island woody vegetation, monthly variability in NDVI was dependent upon amount of precipitation and the temporal scale at which it was analyzed. During wet years, linear fits were found between NDVI and accumulated rainfall from the hyrdological year. This shifted during dry years to rainfall accumulated in the previous month. This is consistent with our findings in interannual variability in NDVI as well as the importance for soil water content to transpiration of M. cerifera (Shao et al., 1995). Woody vegetation uses water efficiently so that in a dry year, when evapotranspirative
Fig. 11. Linear adjustments between the surface with NDVI values above a threshold and NDVI statistics of the sample of pixels with NDVI above 0.5 with the monthly water table depth decrease (m/month) in 2007.
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
processes become relevant, productivity is closely related to water available to the system from the recent past (as opposed to over the year). Monthly relationships between NDVI and groundwater were seen only during dry years when rainfall was scarce and evapotranspiration high. As seen interannually, the freshwater lens became the major source of water for plants. These results demonstrate the important feedback between woody vegetation response to changes in the freshwater lens as shown in the Shao et al. (1995) model, using empirical data. These results could apply to other systems, especially where there are strong directional gradients in resources (i.e., other coastal systems), where porous soils constrain productivity. Spatial variability in NDVI across the island was observed and likely related to age of soils and woody vegetation. Accretion on the ocean side of the island over the past hundred years has formed a chronosequence of soils (Hayden et al., 1991). The soils become progressively older from the ocean beach towards the island interior. Dense thickets of M. cerifera form in the protected swales along the chronosequence and vary in ANPP based on age (Young, 2007). In dry years, this spatial variability was more severe as water table depth was reduced, compared to the wet year when water depth remained steady and >1 m at sealevel. Again, this demonstrates that water table depth was linearly related to NDVI in dry years when water availability in the soil became the only source of water intake for the plants and was a limiting factor. Because of this close coupling of NDVI and water availability, it is very important to consider timing of remote sensing when applying indices to determine productivity. For climate change research, responses of various communities are important to differentiate as drivers of productivity may vary among ecosystems. In addition to understanding ANPP dynamics, these results are important for understanding woody vegetative dynamics in barrier island systems, as woody vegetation is naturally expanding on all Virginia barrier islands that are not experiencing erosion (Young et al., 2007; Zinnert et al., 2011). According to Zinnert et al. (2011), increased woody cover on Hog Island was a function of increased island area and climatic variables, both of which directly affect the size of the freshwater lens (Shao et al., 1995; Young et al., 2007). The decrease in precipitation seen over the last 30 years enabled woody species to colonize lower elevation sites that were formerly flooded providing an evapotranspirative feedback to the freshwater lens in the summer months. The high LAI (up to 12; Brantley and Young, 2010) and ANPP (1500 g m−2 , Knapp et al., 2008) of M. cerifera causes high summer evapotranspiration (Shao et al., 1995). As M. cerifera continues to expand, the demand for water increases. Our results show the importance of the freshwater lens to enabling woody expansion even in months with low precipitation. Although NDVI values declined in dry years indicating moisture stress, these values were still relatively high compared to other systems (Gamon et al., 1995; Jobbágy et al., 2002; Wang et al., 2003) and high compared to NDVI of woody communities on Hog Island prior to 1993 when woody area occupied less total island area (Zinnert et al., 2011). The strong relationship between NDVI and hydrological variables indicates that simple metrics can be used to make future predictions of vulnerable coastal areas to climate change effects. Barrier island shrub thickets represent relatively stable portions of the landscape, and can serve as sentinels of climate change where increased summer droughts can constrict the freshwater lens, reducing water availability (Young et al., 2007). Woody communities may serve both as sentinels of climate change (both positive and negative) when distributions expand due to associated atmospheric effects of climate change or when woody vegetation is lost due to erosion or evapotranspirative demand exceeds precipitation and groundwater availability. Remote sensing technology
299
allows for rapid determination of land conversion over multiple timescales to quantify this phenomenon. 5. Conclusions Annual changes in NDVI were coupled to changes in precipitation. This change was reflected in lower NDVI values (0.75 over 0.9) in the dry years and in the linear relationships between accumulated rainfall and mean water table depth with maximum and mean NDVI values as indicators of overall and mean productivity and biomass. For a certain date a strong dependence of the vegetation status to groundwater was found in dry years when the freshwater lens is the main source of water for the woody vegetation. Monthly variations revealed the different behavior of the woody community within the growing season according to hydrology as in the dry year, water table depths falls and the histogram of NDVI classes gets narrower in the middle of the season. Linear fits with r2 close to 1 were found between NDVI values and accumulated rainfall in the hydrological year. This shifted during dry years to rainfall accumulated in the previous month showing the quick response of vegetation to less availability of the water in the system from the recent past. Monthly relationships between NDVI and groundwater were seen only during dry years when rainfall was scarce and evapotranspiration high and the freshwater lens became the major source of water for plants. These results demonstrate the important feedback between woody vegetation response to changes in the freshwater lens as shown in the Shao et al. (1995) model, using empirical data. Because of this dependence of NDVI to water availability, it is important to consider timing of remote sensing when applying indices to determine productivity. The strong relationship between NDVI and hydrological variables indicates that simple metrics can be used as indicators of vulnerable coastal areas to climate change effects. These results can be a reference to estimate local changes in the vegetation dynamics under the different climatic change scenarios proposed by the Intergovernmental Panel on Climate Change (IPCC) in terms of rainfall. Acknowledgments Some data used in this publication was provided by the Virginia Coast Reserve LTER project, which was supported by National Science Foundation grants BSR-8702333-06, DEB-9211772, DEB9411974, DEB-0080381 and DEB-0621014. The authors would like to thank the people involved in AERONET and USGS-Earth Explorer teams for providing atmospheric and remote sensing data used for this study. References ˜ Asner, G.P., Townsend, A.R., Braswell, B.H., 2000. Satellite observation of El Nino effects on Amazon forest phenology and productivity. Geophys. Res. Lett. 27, 981–984. Brantley, S.T., Young, D.R., 2010. Linking light attenuation sunflecks, and canopy architecture in mesic shrub thickets. Plant Ecol. 206, 225–236. Brantley, S.T., Zinnert, J.C., Young, D.R., 2011. Application of hyperspectral vegetation indices to detect variations in leaf area index in high LAI temperate canopies. Remote Sens. Environ. 115, 514–523. Chander, G., Markham, B.L., Helder, D.L., 2009. Summary of current radiometric calibration coefficients for Landsat MSS TM, ETM+, and EO-1 ALI sensors. Remote Sens. Environ. 113, 893–903. Claudio, H.C., Cheng, Y., Fuentes, D., Gamon, J., Luo, H., Oechel, W., Qiu, H.-L., Rahman, A., Sims, D.A., 2006. Monitoring drought effects on vegetation water content and fluxes in chaparral with the 970 nm water band index. Remote Sens. Environ. 103, 304–311. Ehrenfeld, J.G., 1990. Dynamics and processes of barrier island vegetation. Rev. Aquat. Sci. 2, 437–480. Fang, J., Piao, S., Tang, Z., 2001. Interannual variability in net primary production and precipitation. Science 293, 1723.
300
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Feagin, R.A., Smith, W.K., Psuty, N.P., Young, D.R., Martinez, M.L., Carter, G.A., Lucas, K.L., Gibeat, J.C., Gemma, J.N., Koske, R.E., 2010. Barrier islands: coupling anthropogenic stability with ecological sustainability. J. Coastal Res. 26, 987–992. ˜ Filella, I., Penuelas, J., Llorens, L., Estiarte, M., 2004. Reflectance assessment of seasonal and annual changes in biomass and CO2 uptake of a Mediterranean shrubland submitted to experimental warming and drought. Remote Sens. Environ. 90, 308–318. FLAASH Module, 2009. Atmospheric Correction Module: QUAC and FLAASH User’s Guide, Version 4.7. ITT Visual Information Solutions, Boulder, CO, 44 pp. ˜ Gamon, J.A., Field, C.B., Goulden, M.L., Griffin, K.L., Hartley, A.E., Joel, G., Penuelas, J., Valentini, R., 1995. Relationships between NDVI canopy structures, and photosynthesis in three Californian vegetation types. Ecol. Appl. 5, 28–41. Groeneveld, D.P., Baugh, W.M., 2007. Correcting satellite data to detect vegetation signal for eco-hydrologic analyses. J. Hydrol. 344, 135–145. Groeneveld, D.P., 2008. Remotely sensed groundwater evapotranspiration from alkali scrub vegetation affected by declining water table. J. Hydrol. 358, 294–303. Hargreaves, G.L., Hargreaves, G.H., Riley, J.P., 1985. Irrigation water requirements for Senegal river basin. J. Irrig. Drain. Eng. 111, 265–275. Hayden, B.P., Deuser, R.D., Callahan, J.T., Shugart, H.H., 1991. Long-term research at the Virginia Coast Reserve. Bioscience 41, 310–318. Jobbágy, E.G., Sala, O.E., 2000. Controls of grass and shrub aboveground production in the Patagonian Steppe. Ecol. Appl. 10, 541–549. Jobbágy, E.G., Sala, O.E., Paruelo, J.M., 2002. Patterns and controls of primary production in the Patagonian Steppe: a remote sensing approach. Ecology 83, 307–319. Ji, L., Peters, A.J., 2003. Assessing vegetation response to drought in the northern Great Plains using vegetation and drought indices. Remote Sens. Environ. 87, 85–98. Karl, T.R., Melillo, J.M., Peterson, T.C., 2009. Global Climate Change Impacts in the United States. Cambridge University Press, New York. Kawabata, A., Ichii, K., Yamaguchi, Y., 2001. Global monitoring of the interannual changes in vegetation activities using NDVI and its relationships to temperature and precipitation. Int. J. Remote Sens. 22, 1377–1382. Knapp, A.K., Smith, M.D., 2001. Variation among biomes in temporal dynamics of aboveground primary production. Science 291, 481–484. Knapp, A.K., Briggs, J.M., Collins, S.L., Archer, S.R., Bret-Harte, M.S., Ewers, B.E., Peters, D.P., Young, D.R., Shaver, G.R., Pendall, E., Cleary, M.B., 2008. Shrub encroachment in North American grasslands: shifts in growth form dominance rapidly alters control of ecosystem carbon inputs. Global Change Biol. 14, 615–623. Krovetz, D.K., Porter, J.H., Spitler, J.R., Smith, P.H., 2011. Meteorological Measurements of the Virginia Coast Reserve LTER. Data of the Virginia Coast Reserve Long-term Ecological Research Project VCR97018. Malo, A.R., Nicholson, S.E., 1990. A study of rainfall and vegetation dynamics in the African Sahel using normalized difference vegetation index. J. Arid Environ. 19, 1–24. Naumann, J.C., Young, D.R., Anderson, J.E., 2007. Linking leaf chlorophyll fluorescence properties to physiological responses for detection of salt and drought stress in coastal plant species. Physiol. Plant 131, 422–433. Naumann, J.C., Young, D.R., Anderson, J.E., 2009a. Spatial variations in salinity stress across a coastal landscape using vegetation indices derived from hyperspectral imagery. Plant Ecol. 202, 285–297. Naumann, J.C., Anderson, J.E., Young, D.R., 2009b. Linking physiological responses, chlorophyll fluorescence and hyperspectral imagery to detect salinity stress
using the physiological reflectance index in the coastal shrub, Myrica cerifera. Remote Sens. Environ. 112, 3865–3875. Paruelo, J.M., Epstein, H.E., Lauenroth, W.K., Burke, I.C., 1997. ANPP estimates from NDVI for the Central Grassland Region of the United States. Ecology 78, 953–958. ˜ ˜ Penuelas, J., Llusia, J., Pinol, J., Filella, I., 1997. Photochemical reflectance index and leaf photosynthetic radiation-use efficiency assessment in Mediterranean trees. Int. J. Remote Sens. 18, 2863–2868. Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.M., Tucker, C.J., Stenseth, N.C., 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol. Evol. 20, 503–510. Rouse, J.W., Haas, R.H., Schell, J.A., Deering, D.W., Harlan, J.C., 1974. Monitoring the vernal advancements and retrogradation of natural vegetation. NASA/GSFC, Final Report, Greenbelt, MD, USA, pp. 1–137. Shao, G., Shugart, H.H., Young, D.R., 1995. Simulation of transpiration sensitivity to environmental changes for shrub (Myrica cerifera) thickets on a Virginia barrier island. Ecol. Model. 78, 235–248. Sims, D.A., Gamon, J.A., 2003. Estimation of vegetation water content and photosynthetic tissue area from spectral reflectance: a comparison of indices based on liquid water and chlorophyll absorption features. Remote Sens. Environ. 84, 526–537. Tucker, C., 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8, 127–150. Van Cleve, K., Martin, S., 1991. Long-Term Ecological Research in the United States: A Network of Research Sites 1991. Long-Term Ecological Research Network Office, University of Washington, College of Forest Resources, AR-10 Seattle, Washington 98195, pp. 158–165. Wang, J., Rich, P.M., Price, K.P., 2003. Temporal responses of NDVI to precipitation and temperature in the central Great Plains, USA. Int. J. Remote Sens. 24, 2345–2364. Wang, J., Meng, J.J., Cai, Y.L., 2008. Assessing vegetation dynamics impacted by climate change in the southwestern karst region of China with AVHRR NDVI and AVHRR NPP time-series. Environ. Geol. 54, 1185–1195. Yang, L., Wylie, B.K., Tieszen, L.L., Reed, B.C., 1998. An analysis of relationships among climate forcing and time-integrated NDVI of grasslands over the US northern and central Great Plains. Remote Sens. Environ. 65, 25–37. Young, D.R., Erickson, D.L., Semones, S.W., 1994. Salinity and the small-scale distribution of three barrier island shrubs. Can. J. Bot. 72, 1365–1372. Young, D.R., 2007. Estimating aboveground net primary production in shrubdominated ecosystems. In: Fahey, T.J., Knapp, A.K. (Eds.), Principles and Standards for Measuring Primary Production. Oxford University Press, New York, pp. 49–62. Young, D.R., Porter, J.H., Bachmann, C.M., Shao, G., Fusina, R.A., Bowles, J.H., Korwan, D., Donato, T.F., 2007. Cross-scale patterns in shrub thicket dynamics in the Virginia barrier complex. Ecosystems 10, 854–863. Yuan, J., Niu, Z., 2008. Evaluation of atmospheric correction using FLAASH. In: International Workshop on Earth Observation and Remote Sensing Applications, pp. 1–6, http://dx.doi.org/10.1109/EORSA.2008.4620341. Zinnert, J.C., Nelson, J.D., Hoffman, A.M. Effects of salinity on physiological responses and the photochemical reflectance index in two co-occurring coastal shrubs. Plant Soil, in press. Zinnert, J.C., Shiflett, S.A., Vick, J.K., Young, D.R., 2011. Using remote sensing imagery to detect woody vegetative cover dynamics in response to recent climate change on an Atlantic Coast barrier island. Geocarto Intern. 26, 595–612.
Contents lists available at SciVerse ScienceDirect
Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind
NDVI as an indicator for changes in water availability to woody vegetation Cristina Aguilar a,∗ , Julie C. Zinnert b,c,1 , María José Polo a , Donald R. Young c a b c
Fluvial Dynamics and Hydrology Research Group, IISTA-University of Cordoba, Rabanales Campus, Leonardo Da Vinci Building, 14071 Córdoba, Spain US Army ERDC, Fluorescence Spectroscopy Lab, 7701 Telegraph Road, Alexandria, VA 22315, USA Department of Biology, Virginia Commonwealth University, Richmond, VA 23284, USA
a r t i c l e
i n f o
Article history: Received 15 November 2011 Received in revised form 29 March 2012 Accepted 3 April 2012 Keywords: NDVI Precipitation Freshwater lens Water table depth Plant stress Airborne Landsat TM
a b s t r a c t Barrier islands shrub thickets, the dominant woody community of many Atlantic coast barrier islands, are very sensitive to changes in the freshwater lens and thus, constitute a strong indicator of summer drought. NDVI was computed from airborne images and multispectral images on Hog Island (VA, USA) to evaluate summer growing season changes in woody communities for better predictions of climate change effects. Patterns of NDVI were compared year to year and monthly relative to precipitation and water table depth at the appropriate temporal scale. The highest absolute values of NDVI as well as the larger surface covered by woody vegetation (NDVI > 0.5) occurred in the wet year (2004) with a bimodal distribution of NDVI values (around 0.65 and 0.9) while both dry years (2007 and 2008) showed similar values in maximum, mean and standard deviation and unimodal distributions (around 0.75) of NDVI values. Positive linear adjustments were obtained between maximum (r2 > 0.9) and mean NDVI (r2 > 0.87) and the accumulated rainfall in the hydrological year and the mean water table depth from the last rainfall event till the date of the image acquisition. The spatial variations revealed that water table depth behaved different in wet and dry years. In dry years there was a remarkable increase in mean and maximum values linearly related to water table depth. The highest slope of the adjustment in 2007 indicated a sharp response of vegetation in the driest year. Monthly series of NDVI showed the major role of lack of precipitation through July and August in 2007 with missing classes of NDVI above 0.8 and unimodal distributions in mid-late summer. Best linear fits (r2 close to 1) were obtained with rainfall at different temporal scales: accumulated rainfall in the hydrological year 2004 and accumulated rainfall in the last month previous to the date of 2007 image. Thus, in dry years productivity is closely related to water available from recent past as opposed to over the year for wet years. Good fits (r2 values higher than 0.88) were obtained between monthly decrease in water table depth and NDVI variables just in the dry year. These results demonstrate the important feedback between woody vegetation response to changes in the freshwater lens using empirical data and could apply to other systems with strong directional gradients in resources. © 2012 Elsevier Ltd. All rights reserved.
1. Introduction Coastal system ecological processes are closely coupled to both atmospheric and oceanic drivers, all of which may be influenced by climate change. These ecosystems may be the most sensitive indicators of changing climate (Feagin et al., 2010). Sandy soils typical of North American Atlantic coast barrier islands have minimal water holding capacity which affect distribution and primary production of terrestrial communities. These communities are dependent on access to the soil freshwater lens, which varies temporally and spatially across the landscape. As summer evapotranspiration increases, freshwater capacity in the soil is reduced and may
∗ Corresponding author. Tel.: +34 957212662; fax: +34 957212097. E-mail address: [email protected] (C. Aguilar). 1 Julie C. Zinnert was formerly Julie C. Naumann. 1470-160X/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ecolind.2012.04.008
lead to associated drought stress and limited growth. In addition to evapotranspiration, rainfall, and limited groundwater recharge during the summer months, spatial variations in water availability are subject to plant water use. These effects are exacerbated by microtopography, with dune crests most susceptible to water stress than the lower elevation swales. Frequency and intensity of summer droughts are expected to increase in response to predicted shifts in global climate patterns (Karl et al., 2009). Shrub thickets represent the dominant woody community of many Atlantic coast barrier islands (Young et al., 2007). The primary species is Myrica cerifera, an evergreen, actinorhizal nitrogen fixing shrub. Interestingly, Myrica is characterized by rapid growth but also responds quickly to both salinity and drought stress (Naumann et al., 2007; Young et al., 1994; Zinnert et al., in press). Distribution is limited to low elevation swales with access to the freshwater lens. Myrica thickets are a strong indicator of summer drought relative to other terrestrial plant communities as these communities are most
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
291
Fig. 1. Study site and location of wells (MB 2, MB 3, MB 5, S4, S3 and S2).
sensitive to changes in the freshwater lens. Methods for indicating areas most vulnerable to extended droughts have not been explored in coastal ecosystems. Ability to assess how environmental changes affect dynamics of vegetation is increasingly important for better predictions of climate change effects. Popularity of the application of the Normalized Difference Vegetation Index (NDVI) in ecological studies has enabled quantification and mapping of green vegetation with the goal of estimating aboveground net primary productivity (ANPP) and other landscape-level fluxes (Pettorelli et al., 2005; Wang et al., 2003). NDVI is based on differences in reflectance in the red region (due to pigment absorption) and maximum reflectance in the nearinfrared (caused by cellular structure); it is the most widely used index in remote sensing. It is closely related to a range of intercorrelated biomass variables such as leaf area index (LAI), leaf cover, chlorophyll per unit ground area, green biomass or green vegetation factor (Filella et al., 2004; Gamon et al., 1995). NDVI saturates easily and is not considered a good estimator of high LAI (Asner et al., 2000; Brantley et al., 2011); however, NDVI still retains ecological relevance as an indicator of green biomass change (Wang et al., 2003). NDVI can be a useful tool to couple climate and vegetation distribution and performance at large spatial and temporal scales (Pettorelli et al., 2005). Because vegetation vigor and productivity are related to hydrological variables (rainfall, evapotranspiration, etc.), NDVI serves as a surrogate measure of these factors at the landscape scale (Groeneveld and Baugh, 2007; Wang et al., 2003). The linear response of vegetation NDVI to rainfall for regions with low vegetation cover and rainfall is well documented (Groeneveld and Baugh, 2007; Ji and Peters, 2003; Kawabata et al., 2001; Malo and Nicholson, 1990; Wang et al., 2003; Yang et al., 1998). Other studies have also analyzed trends of NDVI with other variables such as temperature (Wang et al., 2003), and evapotranspiration (Groeneveld, 2008). However, as relationships between NDVI and climatic factors are location-dependent, more detailed analyses are needed (Wang et al., 2003). Further, the sandy soils
on barrier islands have a limited water holding capacity, thus the response of vegetation NDVI should be related to the groundwater table depth in the growing season. Our goal was to determine the relevance of NDVI as an indicator of water availability to coastal woody vegetation. Specifically, we quantified responses to seasonal changes in precipitation and water availability through the groundwater lens by linking landscape level varitations in relative greeness of woody vegetation to past precipitation and hydrology. Throughout the summer growing season changes in woody productivity were assessed by using NDVI calculated from remote sensing imagery. We compared year to year variations in these patterns relative to variations in precipitation and water table depth at the appropiate temporal scale.
2. Materials and methods 2.1. Study site The study is focused on Hog Island (37◦ 40 N; 75◦ 40 W), a barrier island located on the Eastern shore of VA, USA (Fig. 1). The northern end of the island is broad, with a series of dune lines separated by swales and ponds forming a chronosequence (Hayden et al., 1991). The primary woody vegetation is dense thickets of M. cerifera. Upland grasslands are dominated by Spartina patens and Ammophila breviligulata. Extensive marshes on the lagoon side of the island are dominated by Spartina alterniflora. Site records indicate a mean annual temperature around 15 ◦ C and considerable inter annual variability of precipitation between 850 and 1400 mm per year. This variability renders summer drought a high probability of occurrence (Van Cleve and Martin, 1991). The study period (2004–2008) was determined by the availability of spectral data in the summertime for both wet and dry years.
292
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 2. Monthly rainfall (mm/month) measured at Hog Island, monthly evapotranspiration (mm/month) computed from Hargreaves equation with recorded temperature datasets and mean monthly water table depth (m.a.s.l) registered at well MB 3.
2.2. Hydrological data
2.3. Spectral data
Rainfall and temperature data were obtained on a subdaily basis from a meteorological station on Hog Island (Krovetz et al., 2011). Data records were treated to generate daily and monthly values. Fig. 2 shows the huge differences in monthly rainfall that occurred between the wet year (2004) and the dry years (2007 and 2008), especially in the summer period. From the daily records of maximum and minimum temperatures, daily reference evapotranspiration was computed from Hargreaves equation (Hargreaves et al., 1985) in order to assess the evaporative potential according to the meteorological conditions in the area. Unlike rainfall, reference evapotranspiration did not experience significant changes among the years and just during July and August the rate was lower in the wet year (2004) as expected due to the higher frequency of rainy days that cause saturated conditions in the air (Fig. 2). Similarly, water table depth was assessed on a subdaily basis from datasets recorded at three different wells on Hog Island (Fig. 1). Two out of the three wells were located in the middle of two thickets (MB 3 and MB 5; Fig. 1) and the third one on a high marsh close to another thicket (MB 2; Fig. 1). Data available for 2007 and 2008 from three additional wells (S2, S3 and S4; Fig. 1) were used to broaden the spatial range of water levels in both dry years. These wells are located in swales partially surrounded by thickets. Water table depth from well MB 3 is considered to represent the mean behavior of the water table depth in the island affected by woody communities. Mean monthly values of the water table depth in MB 3 showed a considerable soil moisture deficit in the summer periods of dry years, especially in 2007 (Fig. 2) when rainfall is not able to sustain the evapotranspiration needs. The sandy soils in the area have high hydraulic conductivities, experience rapid drainage and may result in very dry conditions if rainfall in summer is insufficient as in 2007.
Two sources of spectral data were available for this study: airborne and multispectral images for the 2004–2008 period. The first source consists of three airborne missions flown in summer, and the data are used to describe the interannual variations. The second source was Landsat TM5, for the analysis of seasonal changes from selected images during the growing season in 2004 and 2007.
2.3.1. Airborne images Airborne images were used to describe interannual changes. Three airborne hyperspectral images were available in the study area (Brantley et al., 2011; Naumann et al., 2009a,b). The first one corresponded to the spectral mission flown on August 24, 2004 on Hog Island using the Portable Hyperspectral Imager for Lowlight spectroscopy (PHILLS) with a spatial resolution of 4 m/pixel and covering 384–1000 nm. The second image was flown on September 13, 2007 with a ProSpecTIR VIS hyperspectral imaging spectrometer (SpectIR) covering 450–2450 nm and a spatial resolution of 2 m/pixel. The same sensor was used in the third flight on August 19, 2008 providing a spatial resolution of 1 m/pixel. All the images were collected under cloud-free conditions and data were post-processed to minimize geometric and radiometric effects. Calibration was performed using the empirical line calibration method within ENVI (RSI, Inc.) with ground reflectance measurements. Transformation of the PHILLS data was accomplished using the SpectIR data as a reference to normalize the bands. NDVI at landscape-level was calculated on each pixel of the study site from Eq. (1) (Rouse et al., 1974): NDVI =
800 − 670 800 + 670
(1)
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
293
Table 1 Basic statistics of NDVI for the areas with woody communities (NDVI higher than 0.5) and hydrological variables: accumulated rainfall (mm) in the hydrological year (from 1st October to date of the image) and mean water table depth (m.a.s.l) registered at well MB 3 from last rainfall event till the day before the image acquisition. Date
Max
Mean
Stand. Dev.
Total area of woody communities (ha)
Accumulated rainfall (mm)
Water depth (m.a.s.l)
08/24/2004 09/13/2007 08/19/2008
0.97 0.82 0.82
0.75 0.67 0.66
0.13 0.08 0.08
462 179 254
1075.88 482.09 610.76
1.04 0.48 0.63
where 800 and 670 are the reflectance values centered at 800 and 670 nm respectively. 2.3.2. Multispectral images Multispectral imagery was analyzed in order to describe different patterns within the growing season. Images used in this study consisted of scenes from Landsat TM5 (path/row: 14/34) downloaded at the New Earth Explorer Interface for each month in the summer period. Four scenes were acquired for 2004 and 2007, respectively. Cloud-free scenes were selected in order to minimize uncertainties due to heterogeneous atmospheric conditions; however, radiometric calibration and atmospheric correction were applied (Yuan and Niu, 2008). Radiometric calibration was achieved through the equations and rescaling factors in Chander et al. (2009). Atmospheric correction was conducted to retrieve surface reflectance by use of MODTRAN4 as implemented into ENVI FLAASH (FLAASH Module, 2009). Atmospheric properties were defined in terms of aerosol type, water vapor and visibility. In this study the maritime aerosol type and 40 km of visibility were considered for all the images as suggested by ENVI FLAASH. Water vapor content values registered at Wallops site (35◦ N; 75◦ W) from the AERONET network were applied (FLAASH Module, 2009) when no data was available. All scenes were subset to form the image of the study area. NDVI values were computed from Eq. (2) which represents the difference between reflectance in the near-infrared and red bands (bands 4 and 3), 4 and 3 respectively, normalized by the sum of both of them (Tucker, 1979). NDVI =
4 − 3 4 + 3
(2)
were estimated by fitting functions through the data using simple least-squares regression. To show the influence of water availability on vegetation productivity for each date, spatial variations of NDVI were analyzed within each image. Regions were defined in a 20 m buffer area around each well that reflected a homogeneous NDVI value of the thicket around the well. As the wells are free from outside impacts, fluctuations in the water table depth were considered to be linked to natural groundwater use by plants and related to the mean NDVI value in the thicket. Finally, in order to separate effects of soil characteristics and microtopography from the different availability to freshwater in the wet and dry years, a dimensionless NDVI map was calculated dividing the NDVI value of each pixel by the mean NDVI value for the image. 2.4.2. Monthly variations within the growing season A similar methodology was applied with the Landsat 5 TM images. Frequency distribution function for the surfaces of different NDVI classes, cumulative probability distribution function and basic statistics of NDVI values (maximum, mean and standard deviation) were calculated for each image. These computations helped to assess changes in productivity along the growing season within the wet and the dry years. Similarly, adjustments between NDVI and hydrological variables were made for the images within each growing season. Careful consideration in the temporal integration of hydrological variables was taken into account in order to reflect the short-term effects of water availability over the status of the woody cover. 3. Results
2.4. Relations between NDVI and hydrological variables 3.1. Interannual variations This study was restricted to woody communities on the island due to their quick response to groundwater availability. Supervised classification was performed in a previous study using maximum likelihood methodology into five classes (woody, bare sand, grassland, water, and marsh) (Zinnert et al., 2011). Based on NDVI values obtained from Zinnert et al. (2011), 0.5 was used as the NDVI threshold between woody vegetation and other classes. 2.4.1. Interannual variations analysis The area covered with woody communities was classified to describe the frequency distribution function of NDVI for the three images. Surface instead of number of pixels was used in order to compensate different spatial resolution of the three images. Cumulative probability distribution function and basic statistics (maximum, mean and standard deviation) were computed for the sample of pixels. Maximum and mean NDVI values reflected overall and mean productivity and biomass, respectively (Pettorelli et al., 2005), whereas the standard deviation represented a measure of the spatial variability in productivity. In order to quantify the dependence of NDVI values with hydrological regimen, different adjustments with hydrological variables were computed. NDVI trends in terms of the basic statistics were related to variations in precipitation and water table depth. Trends
For 2004, 2007 and 2008, the highest values of NDVI and the largest surface area covered by woody vegetation occurred in the wet year compared to both dry years (Table 1). For pixels with NDVI above 0.5, both dry years showed similar values in the basic statistics considered (maximum, mean and standard deviation) (Table 1). Fig. 3 shows the histogram in terms of surface area with NDVI values higher than 0.5 and the cumulative probability distribution function. In general, a very different distribution can be seen in the wet year when all the classes of NDVI are represented with two modes in NDVI values around 0.65 and 0.9 (Fig. 3). To contrast, in both dry years the surface covered by NDVI above 0.75 is minimal, with just a few pixels with NDVI values higher than 0.8. Compared to 2007 (the driest year), 2008 had more surface area covered in every NDVI class. The same ideas can be drawn from the cumulative distribution function as 2004 is more tilted, indicating higher NDVI values for a certain percentile compared to 2007 and 2008 (e.g., 0.93 in 2004 against around 0.75 in 2007 and 2008 for the 90th percentile). Once relationships between NDVI values and the wet/dry character of the year were demonstrated, NDVI trends were related to hydrological variables. After several trials, the variables considered were accumulated rainfall in the hydrological year (from
294
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 3. Histograms and cumulative probability distribution functions of NDVI areas above 0.5 for each airborne image in 2004, 2007 and 2008.
October 1 of the preceding calendar year to the day before the sensor overpass) and mean water table depth registered at well MB 3 from the last rainfall event before the date of the acquisition (Table 1). The best adjustments were obtained with linear relationships between hydrological variables considered and absolute maximum NDVI and mean NDVI for the area (Fig. 4). Coefficients of determination (r2 ) values above 0.9 were obtained between NDVI maximum values and both variables. Similarly, r2 values above 0.87 were obtained for mean NDVI values. Considering both hydrological variables, slightly better fits resulted with rainfall than with ground water table depth. We assume that the linear trend of adjustments is determined by the availability of 3 years data. Thus, future availability of more hyperspectral images including medium and extreme years in terms of rainfall, would allow inclusion of more than three points in the fits and so, other
types of fits (e.g., exponential, power, etc.) could be tested as well. Spatial variation in NDVI was considered in the area around each well (Table 2). There was a remarkable increase in mean and maximum values moving from west to east of the island in the dry years (2007 and 2008) in the most northern wells (MB 2, MB 3 and MB 4). Mean and maximum NDVI values were linearly related to water table depth during the dry years (Fig. 5) with best fits for the mean values (r2 values above 0.85). The slope of adjustments was highest in 2007 (Fig. 5) which shows the sharp response of vegetation in the driest year. Maximum NDVI values in the dry years could be better approximated to a nonlinear function, as inferred from the trend in Fig. 5. However, the sample of wells is a limiting factor, since conditions around wells S2, S3, and S4 are too similar. Thus, this apparent nonlinear trend should be validated
Fig. 4. Linear adjustments between maximum and mean NDVI of the sample of pixels with NDVI above 0.5 with the accumulated rainfall in the hydrological year (mm) and with the mean water table depth at well MB 3 from the last rainfall event till the date of the airborne image.
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
295
Table 2 Mean and maximum NDVI values in 20 m around the well and mean water table depth (m.a.s.l) from the last rainfall event till the day before the image acquisition. Well
2004
2007
NDVI
MB 2 MB 3 MB 5 S4 S3 S2
Water depth (m.a.s.l)
Mean
Max
0.72 0.87 0.89 0.79 0.90 0.81
0.77 0.94 0.95 0.90 0.95 0.93
1.17 1.04 1.31 No data No data No data
2008
NDVI
Water depth (m.a.s.l)
Mean
Max
0.50 0.66 0.73 0.62 0.67 0.66
0.64 0.75 0.80 0.79 0.77 0.78
against datasets from a denser a more representative network of wells for a complete spatial analysis. Fig. 6 represents pixels with an NDVI value over the mean value in the study site for each year respectively. Dimensionless NDVI maps indirectly isolate the effects of the different rainfall regimen from the slope, soil type and depth, etc. The difference between both figures can be explained by the different rainfall occurrence. Surface with NDVI above the mean value decreased from 221 ha in 2004 to 107 ha in 2007 and 144 ha in 2008. The highest NDVI values in 2004 were about 30% higher of the mean NDVI compared to 23% in 2007 and 25% in 2008. 3.2. Monthly variations within the growing season NDVI monthly series show the role of the absence of precipitation in the months of July and August of 2007 (Fig. 7). As the amount of rainfall is not able to sustain actual evapotranspiration, water table depth falls (Fig. 2) and some woody species are excluded. This is apparent in Fig. 8, where classes above 0.8 are missing in August 2007 and the bimodal distribution is replaced by the typical distribution of months in early summer. The cumulative distribution functions of NDVI showed a very different behavior between both years as well (Fig. 8). In 2004, the function with the lowest NDVI values for a certain percentile corresponds to the first image of the growing season (July), while in 2007 this happens in August, demonstrating the shortage of rainfall during the previous months.
0 0.48 1.10 0.54 0.66 0.34
NDVI
Water depth (m.a.s.l)
Mean
Max
0.55 0.62 0.68 0.61 0.64 0.61
0.68 0.77 0.77 0.76 0.76 0.77
−0.15 0.63 1.51 0.67 0.70 0.60
Linear fits were obtained with rainfall but the best fits were obtained at different temporal scales: the accumulated rainfall in the hydrological year for the wet year and the accumulated rainfall in the last month previous to the date of the image for the dry year. r2 values were very close to 1 in every adjustment of surface area with NDVI higher than a threshold (Fig. 9) and NDVI basic statistics (Fig. 10). Linear fits with accumulated rainfall in the last month previous to the date of the image were carried out for the wet year, however r2 values were relatively low, between 0.25 and 0.7 for all variables considered. Adjustments between NDVI variables (surface of NDVI thresholds and basic statistics) and mean water table depth from the previous rainfall event were poor in the wet year (r2 values between 0.40 and 0.62). However, for the dry year very good fits were obtained between the monthly decrease in water table depth and NDVI variables with r2 values higher than 0.93 for the surface of NDVI thresholds, and from 0.88 to 0.95 for the basic statistics (Fig. 11). 4. Discussion Our results show that NDVI of woody vegetation can be used as an indicator for groundwater availability and precipitation inputs in coastal systems. The drivers of interannual variability in ANPP are of interest for projections of ecosystem responses to climate change (Fang et al., 2001). Knapp and Smith (2001) suggested that fluctuations in precipitation were not related to interannual ANPP at the large-scale (across North America), while Fang et al. (2001) found positive relationships between ANPP derived from NDVI and precipitation. Coastal ecological systems are ideal for quantifying differences in primary productivity due to spatial and temporal variations in climate based stressors (i.e., water availability, salinity) as they are primarily influenced by directional environmental gradients (Ehrenfeld, 1990). Studies that demonstrate the use of remote sensing for deriving ANPP and relationships with climatic variables can allow for rapid, large-scale assessment of ecosystem processes. Our results are notable in documenting the response of high productivity woody vegetation to annual and monthly fluctuations in both precipitation and groundwater availability via NDVI. Additionally, our results are significant for showing these relationships in a mesic environment since numerous studies have shown direct relationships in arid and semiarid ecosystems (Jobbágy et al., 2002; Jobbágy and Sala, 2000; Paruelo et al., 1997), thus extending the capabilities of using NDVI as a metric for water availability at the landscape level. 4.1. Interannual variations
Fig. 5. Linear adjustment between the mean (solid markers) and maximum (empty markers) NDVI values in 20 m around the well and mean water table depth (m.a.s.l) the day before the image acquisition.
In many systems, the relationships between interannual variations in NDVI are more closely linked to interannual variations in temperature over those in precipitation (Wang et al., 2008). In this study, annual changes in NDVI were coupled to changes in
296
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 6. Area in blue with NDVI value above the mean NDVI value per year (Table 1) in the study site. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
precipitation. This change is reflected in lower NDVI values in 2007 and 2008 and in the linear relationships between accumulated rainfall and mean water table depth with the maximum and mean NDVI values as indicators of overall and mean productivity and biomass, respectively. Green leaf area varies with vegetation water content, and since NDVI is an indicator of green leaf area, NDVI decreases as water becomes less available (Claudio et al., 2006; Gamon et al., ˜ 1995; Penuelas et al., 1997; Sims and Gamon, 2003). In the barrier island system, using NDVI as an indicator of groundwater was dependent upon rainfall as water table depth behaved different in wet and dry years. NDVI was a good indicator of groundwater availability in dry years as the freshwater lens became
the main source of water for woody vegetation. Because of low elevation and high transpiration rates of the dominant woody species, M. cerifera, the woody community on Hog Island is closely coupled to the freshwater lens (Shao et al., 1995), which is recharged through precipitation. During wet years, water table depth did not change very much as water uptake by the plants was compensated by rainfall-infiltration events. 4.2. Monthly variations The model MCHOG was developed by Shao et al. (1995) to simulate transpiration of woody vegetation on barrier islands and
Fig. 7. NDVI from Landsat TM5 images along the growing season in 2004 and 2007. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
297
Fig. 8. Histograms and cumulative probability distribution function of NDVI areas above 0.5 for each day considered in the growing season in 2004 and 2007.
Fig. 9. Linear adjustments between surface with NDVI values above a threshold of the sample of pixels with NDVI above 0.5 with the accumulated rainfall in the hydrological year (mm) till the date of the image for 2004 and with the mean rainfall of the previous month (mm/month) for 2007.
298
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Fig. 10. Linear adjustments between NDVI statistics of the sample of pixels with NDVI above 0.5 with the accumulated rainfall in the hydrological year (mm) till the date of the image for 2004 and with the mean rainfall of the previous month (mm/month) for 2007.
demonstrated the importance of soil water (via precipitation) to productivity. Their results indicated that the high LAI of M. cerifera and low water holding capacity of sandy soils lead to a rapid decrease in soil water content and reduced transpiration during extended periods without precipitation. To further investigate the coupling between vegetation productivity and water availability considering the low water holding capacity of soils, we examined monthly variations in NDVI to hydrological variables over the growing season.
For barrier island woody vegetation, monthly variability in NDVI was dependent upon amount of precipitation and the temporal scale at which it was analyzed. During wet years, linear fits were found between NDVI and accumulated rainfall from the hyrdological year. This shifted during dry years to rainfall accumulated in the previous month. This is consistent with our findings in interannual variability in NDVI as well as the importance for soil water content to transpiration of M. cerifera (Shao et al., 1995). Woody vegetation uses water efficiently so that in a dry year, when evapotranspirative
Fig. 11. Linear adjustments between the surface with NDVI values above a threshold and NDVI statistics of the sample of pixels with NDVI above 0.5 with the monthly water table depth decrease (m/month) in 2007.
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
processes become relevant, productivity is closely related to water available to the system from the recent past (as opposed to over the year). Monthly relationships between NDVI and groundwater were seen only during dry years when rainfall was scarce and evapotranspiration high. As seen interannually, the freshwater lens became the major source of water for plants. These results demonstrate the important feedback between woody vegetation response to changes in the freshwater lens as shown in the Shao et al. (1995) model, using empirical data. These results could apply to other systems, especially where there are strong directional gradients in resources (i.e., other coastal systems), where porous soils constrain productivity. Spatial variability in NDVI across the island was observed and likely related to age of soils and woody vegetation. Accretion on the ocean side of the island over the past hundred years has formed a chronosequence of soils (Hayden et al., 1991). The soils become progressively older from the ocean beach towards the island interior. Dense thickets of M. cerifera form in the protected swales along the chronosequence and vary in ANPP based on age (Young, 2007). In dry years, this spatial variability was more severe as water table depth was reduced, compared to the wet year when water depth remained steady and >1 m at sealevel. Again, this demonstrates that water table depth was linearly related to NDVI in dry years when water availability in the soil became the only source of water intake for the plants and was a limiting factor. Because of this close coupling of NDVI and water availability, it is very important to consider timing of remote sensing when applying indices to determine productivity. For climate change research, responses of various communities are important to differentiate as drivers of productivity may vary among ecosystems. In addition to understanding ANPP dynamics, these results are important for understanding woody vegetative dynamics in barrier island systems, as woody vegetation is naturally expanding on all Virginia barrier islands that are not experiencing erosion (Young et al., 2007; Zinnert et al., 2011). According to Zinnert et al. (2011), increased woody cover on Hog Island was a function of increased island area and climatic variables, both of which directly affect the size of the freshwater lens (Shao et al., 1995; Young et al., 2007). The decrease in precipitation seen over the last 30 years enabled woody species to colonize lower elevation sites that were formerly flooded providing an evapotranspirative feedback to the freshwater lens in the summer months. The high LAI (up to 12; Brantley and Young, 2010) and ANPP (1500 g m−2 , Knapp et al., 2008) of M. cerifera causes high summer evapotranspiration (Shao et al., 1995). As M. cerifera continues to expand, the demand for water increases. Our results show the importance of the freshwater lens to enabling woody expansion even in months with low precipitation. Although NDVI values declined in dry years indicating moisture stress, these values were still relatively high compared to other systems (Gamon et al., 1995; Jobbágy et al., 2002; Wang et al., 2003) and high compared to NDVI of woody communities on Hog Island prior to 1993 when woody area occupied less total island area (Zinnert et al., 2011). The strong relationship between NDVI and hydrological variables indicates that simple metrics can be used to make future predictions of vulnerable coastal areas to climate change effects. Barrier island shrub thickets represent relatively stable portions of the landscape, and can serve as sentinels of climate change where increased summer droughts can constrict the freshwater lens, reducing water availability (Young et al., 2007). Woody communities may serve both as sentinels of climate change (both positive and negative) when distributions expand due to associated atmospheric effects of climate change or when woody vegetation is lost due to erosion or evapotranspirative demand exceeds precipitation and groundwater availability. Remote sensing technology
299
allows for rapid determination of land conversion over multiple timescales to quantify this phenomenon. 5. Conclusions Annual changes in NDVI were coupled to changes in precipitation. This change was reflected in lower NDVI values (0.75 over 0.9) in the dry years and in the linear relationships between accumulated rainfall and mean water table depth with maximum and mean NDVI values as indicators of overall and mean productivity and biomass. For a certain date a strong dependence of the vegetation status to groundwater was found in dry years when the freshwater lens is the main source of water for the woody vegetation. Monthly variations revealed the different behavior of the woody community within the growing season according to hydrology as in the dry year, water table depths falls and the histogram of NDVI classes gets narrower in the middle of the season. Linear fits with r2 close to 1 were found between NDVI values and accumulated rainfall in the hydrological year. This shifted during dry years to rainfall accumulated in the previous month showing the quick response of vegetation to less availability of the water in the system from the recent past. Monthly relationships between NDVI and groundwater were seen only during dry years when rainfall was scarce and evapotranspiration high and the freshwater lens became the major source of water for plants. These results demonstrate the important feedback between woody vegetation response to changes in the freshwater lens as shown in the Shao et al. (1995) model, using empirical data. Because of this dependence of NDVI to water availability, it is important to consider timing of remote sensing when applying indices to determine productivity. The strong relationship between NDVI and hydrological variables indicates that simple metrics can be used as indicators of vulnerable coastal areas to climate change effects. These results can be a reference to estimate local changes in the vegetation dynamics under the different climatic change scenarios proposed by the Intergovernmental Panel on Climate Change (IPCC) in terms of rainfall. Acknowledgments Some data used in this publication was provided by the Virginia Coast Reserve LTER project, which was supported by National Science Foundation grants BSR-8702333-06, DEB-9211772, DEB9411974, DEB-0080381 and DEB-0621014. The authors would like to thank the people involved in AERONET and USGS-Earth Explorer teams for providing atmospheric and remote sensing data used for this study. References ˜ Asner, G.P., Townsend, A.R., Braswell, B.H., 2000. Satellite observation of El Nino effects on Amazon forest phenology and productivity. Geophys. Res. Lett. 27, 981–984. Brantley, S.T., Young, D.R., 2010. Linking light attenuation sunflecks, and canopy architecture in mesic shrub thickets. Plant Ecol. 206, 225–236. Brantley, S.T., Zinnert, J.C., Young, D.R., 2011. Application of hyperspectral vegetation indices to detect variations in leaf area index in high LAI temperate canopies. Remote Sens. Environ. 115, 514–523. Chander, G., Markham, B.L., Helder, D.L., 2009. Summary of current radiometric calibration coefficients for Landsat MSS TM, ETM+, and EO-1 ALI sensors. Remote Sens. Environ. 113, 893–903. Claudio, H.C., Cheng, Y., Fuentes, D., Gamon, J., Luo, H., Oechel, W., Qiu, H.-L., Rahman, A., Sims, D.A., 2006. Monitoring drought effects on vegetation water content and fluxes in chaparral with the 970 nm water band index. Remote Sens. Environ. 103, 304–311. Ehrenfeld, J.G., 1990. Dynamics and processes of barrier island vegetation. Rev. Aquat. Sci. 2, 437–480. Fang, J., Piao, S., Tang, Z., 2001. Interannual variability in net primary production and precipitation. Science 293, 1723.
300
C. Aguilar et al. / Ecological Indicators 23 (2012) 290–300
Feagin, R.A., Smith, W.K., Psuty, N.P., Young, D.R., Martinez, M.L., Carter, G.A., Lucas, K.L., Gibeat, J.C., Gemma, J.N., Koske, R.E., 2010. Barrier islands: coupling anthropogenic stability with ecological sustainability. J. Coastal Res. 26, 987–992. ˜ Filella, I., Penuelas, J., Llorens, L., Estiarte, M., 2004. Reflectance assessment of seasonal and annual changes in biomass and CO2 uptake of a Mediterranean shrubland submitted to experimental warming and drought. Remote Sens. Environ. 90, 308–318. FLAASH Module, 2009. Atmospheric Correction Module: QUAC and FLAASH User’s Guide, Version 4.7. ITT Visual Information Solutions, Boulder, CO, 44 pp. ˜ Gamon, J.A., Field, C.B., Goulden, M.L., Griffin, K.L., Hartley, A.E., Joel, G., Penuelas, J., Valentini, R., 1995. Relationships between NDVI canopy structures, and photosynthesis in three Californian vegetation types. Ecol. Appl. 5, 28–41. Groeneveld, D.P., Baugh, W.M., 2007. Correcting satellite data to detect vegetation signal for eco-hydrologic analyses. J. Hydrol. 344, 135–145. Groeneveld, D.P., 2008. Remotely sensed groundwater evapotranspiration from alkali scrub vegetation affected by declining water table. J. Hydrol. 358, 294–303. Hargreaves, G.L., Hargreaves, G.H., Riley, J.P., 1985. Irrigation water requirements for Senegal river basin. J. Irrig. Drain. Eng. 111, 265–275. Hayden, B.P., Deuser, R.D., Callahan, J.T., Shugart, H.H., 1991. Long-term research at the Virginia Coast Reserve. Bioscience 41, 310–318. Jobbágy, E.G., Sala, O.E., 2000. Controls of grass and shrub aboveground production in the Patagonian Steppe. Ecol. Appl. 10, 541–549. Jobbágy, E.G., Sala, O.E., Paruelo, J.M., 2002. Patterns and controls of primary production in the Patagonian Steppe: a remote sensing approach. Ecology 83, 307–319. Ji, L., Peters, A.J., 2003. Assessing vegetation response to drought in the northern Great Plains using vegetation and drought indices. Remote Sens. Environ. 87, 85–98. Karl, T.R., Melillo, J.M., Peterson, T.C., 2009. Global Climate Change Impacts in the United States. Cambridge University Press, New York. Kawabata, A., Ichii, K., Yamaguchi, Y., 2001. Global monitoring of the interannual changes in vegetation activities using NDVI and its relationships to temperature and precipitation. Int. J. Remote Sens. 22, 1377–1382. Knapp, A.K., Smith, M.D., 2001. Variation among biomes in temporal dynamics of aboveground primary production. Science 291, 481–484. Knapp, A.K., Briggs, J.M., Collins, S.L., Archer, S.R., Bret-Harte, M.S., Ewers, B.E., Peters, D.P., Young, D.R., Shaver, G.R., Pendall, E., Cleary, M.B., 2008. Shrub encroachment in North American grasslands: shifts in growth form dominance rapidly alters control of ecosystem carbon inputs. Global Change Biol. 14, 615–623. Krovetz, D.K., Porter, J.H., Spitler, J.R., Smith, P.H., 2011. Meteorological Measurements of the Virginia Coast Reserve LTER. Data of the Virginia Coast Reserve Long-term Ecological Research Project VCR97018. Malo, A.R., Nicholson, S.E., 1990. A study of rainfall and vegetation dynamics in the African Sahel using normalized difference vegetation index. J. Arid Environ. 19, 1–24. Naumann, J.C., Young, D.R., Anderson, J.E., 2007. Linking leaf chlorophyll fluorescence properties to physiological responses for detection of salt and drought stress in coastal plant species. Physiol. Plant 131, 422–433. Naumann, J.C., Young, D.R., Anderson, J.E., 2009a. Spatial variations in salinity stress across a coastal landscape using vegetation indices derived from hyperspectral imagery. Plant Ecol. 202, 285–297. Naumann, J.C., Anderson, J.E., Young, D.R., 2009b. Linking physiological responses, chlorophyll fluorescence and hyperspectral imagery to detect salinity stress
using the physiological reflectance index in the coastal shrub, Myrica cerifera. Remote Sens. Environ. 112, 3865–3875. Paruelo, J.M., Epstein, H.E., Lauenroth, W.K., Burke, I.C., 1997. ANPP estimates from NDVI for the Central Grassland Region of the United States. Ecology 78, 953–958. ˜ ˜ Penuelas, J., Llusia, J., Pinol, J., Filella, I., 1997. Photochemical reflectance index and leaf photosynthetic radiation-use efficiency assessment in Mediterranean trees. Int. J. Remote Sens. 18, 2863–2868. Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.M., Tucker, C.J., Stenseth, N.C., 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol. Evol. 20, 503–510. Rouse, J.W., Haas, R.H., Schell, J.A., Deering, D.W., Harlan, J.C., 1974. Monitoring the vernal advancements and retrogradation of natural vegetation. NASA/GSFC, Final Report, Greenbelt, MD, USA, pp. 1–137. Shao, G., Shugart, H.H., Young, D.R., 1995. Simulation of transpiration sensitivity to environmental changes for shrub (Myrica cerifera) thickets on a Virginia barrier island. Ecol. Model. 78, 235–248. Sims, D.A., Gamon, J.A., 2003. Estimation of vegetation water content and photosynthetic tissue area from spectral reflectance: a comparison of indices based on liquid water and chlorophyll absorption features. Remote Sens. Environ. 84, 526–537. Tucker, C., 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8, 127–150. Van Cleve, K., Martin, S., 1991. Long-Term Ecological Research in the United States: A Network of Research Sites 1991. Long-Term Ecological Research Network Office, University of Washington, College of Forest Resources, AR-10 Seattle, Washington 98195, pp. 158–165. Wang, J., Rich, P.M., Price, K.P., 2003. Temporal responses of NDVI to precipitation and temperature in the central Great Plains, USA. Int. J. Remote Sens. 24, 2345–2364. Wang, J., Meng, J.J., Cai, Y.L., 2008. Assessing vegetation dynamics impacted by climate change in the southwestern karst region of China with AVHRR NDVI and AVHRR NPP time-series. Environ. Geol. 54, 1185–1195. Yang, L., Wylie, B.K., Tieszen, L.L., Reed, B.C., 1998. An analysis of relationships among climate forcing and time-integrated NDVI of grasslands over the US northern and central Great Plains. Remote Sens. Environ. 65, 25–37. Young, D.R., Erickson, D.L., Semones, S.W., 1994. Salinity and the small-scale distribution of three barrier island shrubs. Can. J. Bot. 72, 1365–1372. Young, D.R., 2007. Estimating aboveground net primary production in shrubdominated ecosystems. In: Fahey, T.J., Knapp, A.K. (Eds.), Principles and Standards for Measuring Primary Production. Oxford University Press, New York, pp. 49–62. Young, D.R., Porter, J.H., Bachmann, C.M., Shao, G., Fusina, R.A., Bowles, J.H., Korwan, D., Donato, T.F., 2007. Cross-scale patterns in shrub thicket dynamics in the Virginia barrier complex. Ecosystems 10, 854–863. Yuan, J., Niu, Z., 2008. Evaluation of atmospheric correction using FLAASH. In: International Workshop on Earth Observation and Remote Sensing Applications, pp. 1–6, http://dx.doi.org/10.1109/EORSA.2008.4620341. Zinnert, J.C., Nelson, J.D., Hoffman, A.M. Effects of salinity on physiological responses and the photochemical reflectance index in two co-occurring coastal shrubs. Plant Soil, in press. Zinnert, J.C., Shiflett, S.A., Vick, J.K., Young, D.R., 2011. Using remote sensing imagery to detect woody vegetative cover dynamics in response to recent climate change on an Atlantic Coast barrier island. Geocarto Intern. 26, 595–612.