Albedo estimated from remote sensing correlates with ecosystem multifunctionality in global drylands

Abstract

Monitoring ecosystem functioning is a significant step towards detecting changes in ecosystem attributes that could be linked to land degradation and desertification in drylands worldwide. Remote sensing-based vegetation indices (VIs) and land surface albedo are two favorite indicators to monitor desertification process due to their close relationship with ecosystem status and to their increasing applicability over multiple spatiotemporal scales. While VIs are routinely used to monitor ecosystem attributes and functions such as vegetation cover and productivity, no previous study has evaluated whether remote sensing-measured albedo is related to the simultaneous provision of multiple ecosystem functions (multifunctionality) in global drylands. In this study, we evaluated the correlation of six albedo metrics (shortwave black-sky albedo, shortwave white-sky albedo, visible black-sky albedo, visible white-sky albedo, near-infrared black-sky albedo and near-infrared white-sky albedo) and two VIs (Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI)) with multifunctionality indices related to carbon, nitrogen and phosphorus cycling measured in 61 dryland ecosystems from all continents except Antarctica. We found a negative relationship between land surface albedo and multifunctionality. Black-sky albedo had a stronger correlation with multifunctionality than white-sky albedo. Visible black-sky albedo showed the strongest correlation with multifunctionality (MUL, -0.314), as well as with functions related to carbon (CCY, -0.216) and nitrogen cycling (NCY, -0.410), while near-infrared (-0.339) and shortwave black-sky albedo (-0.325) showed stronger correlations with functions related to phosphorus cycling (PCY) than visible black-sky albedo (-0.233) did. VIs showed significant positive correlations with MUL, CCY, and NCY, and the magnitudes were higher than those observed between albedo metrics and the multifunctionality indices. However, VIs were not correlated with PCY, which had significant correlations with both shortwave and near-infrared albedo. Though the magnitudes of the correlations observed were not high, which may result from the wide variability in soil and vegetation types in our dataset, our findings indicate that remotely sensed albedo correlates to multifunctionality, which has been linked to alternative states in global drylands. As such, albedo has the potential to monitor changes in dryland ecosystem functioning, which can inform us about the onset of desertification in these areas.

1. Introduction

Desertification is a major global environmental problem, as it is estimated to affect 10-20% of drylands worldwide (Reynolds et al., 2007a), which occupy ~45% of Earth’s land area (Prăvălie et al., 2016) and host about 38% of the global population (Reynolds et al., 2007a). The establishment of monitoring programs is an effective way to detect early changes in ecosystem variables that could inform us about the onset of desertification processes before they are irreversible (Fernández et al., 2002; Reynolds et al., 2007b). The United Nation’s Conference on Environment and Development defines desertification as land degradation in arid, semi-arid and dry sub-humid areas (i.e., drylands) resulting from various factors, including climatic variations and human activities (UN, 1992). Despite being widely used by researchers, policy makers and organizations (UNCCD, 1994; Dawelbait and Morari, 2012; Lamchin et al., 2015), this definition does not explicitly specify the processes (e.g., soil erosion, loss of vegetation cover) nor causes (e.g., over-grazing, over-cultivation, and wood collection for fuel) leading to desertification, which are often considered through the lens of particular research disciplines (Rasmussen et al., 2001). As such, the ambiguous nature of this definition has led to the use of multiple indicators to assess and monitor desertification status (Dawelbait and Morari, 2012). Many of these indicators are related to soil and vegetation attributes (Rubio and Bochet, 1998; Helldén and Tottrup, 2008; Symeonakis et al., 2016), as these are strongly linked to ecosystem processes such as carbon sequestration, productivity, and nutrient cycling. These processes underlie the provision of key ecosystem functions and services, which loss is linked to land degradation and desertification (Millennium Ecosystem Assessment, 2005; Tarrasón et al., 2016).

Multiple field studies have evaluated the suitability of ground-based measurements of vegetation (e.g., cover, spatial pattern, species composition) and soil (e.g., nutrient content, texture, surface features) properties to assess changes in ecosystem functions along environmental gradients, and to detect non-linear changes in these ecosystem attributes, such as those typically characterizing desertification processes (Kéfi et al., 2007; Maestre and Escudero, 2009; Lin et al., 2010; Berdugo et al., 2017). Significant advances have also been made to develop easy-to-measure soil and vegetation indicators to assess changes in ecosystem functions in the field (Tongway 1994; Tongway and Hindley, 2004; Herrick et al. 2005), which are being successfully calibrated against laboratory-based estimations of these functions and have been used to compare the functional status of dryland ecosystems in multiple continents (e.g., Maestre and Puche, 2009; Gaitán et al., 2018; Ata Rezaei et al., 2006). However, the collection of these ground-based indicators is usually time-consuming and labor-intensive, and may not be used at all because of logistic and economic reasons when large and remote geographic areas need to be monitored. Currently, remote sensing provides a realistic alternative to obtain land surface information across broad regions and over increasingly longer periods at a low cost. It also allows the development of easy-to-measure proxies of ecosystem functions for monitoring the status of drylands. In this direction, remote sensing-based vegetation indices (VIs), such as the Normalized Difference Vegetation Index (NDVI), have been successfully employed to predict ecosystem functions in drylands from Australia (Ong et al., 2009), Spain (García-Gómez and Maestre, 2011), Argentina (Gaitán et al. 2013), China (Yan et al., 2005; Pan et al., 2013), USA (Kunkel, 2011) and Kenya (Kurgat et al., 2014), among others. The ability of VIs to predict these functions mainly results from the close links existing between vegetation growth and soil carbon (C) and nitrogen (N) cycling (Chase et al., 2000; Pan et al., 2013; Kunkel, 2011). However, soil phosphorus (P) are also fundamental to the provision of key ecosystem functions and services (Finzi et al., 2011). Forecasted increases in aridity with climate change may uncouple the C, N and P cycles in drylands (Delgado-Baquerizo et al., 2013). Therefore, it is important to explore whether remote sensing-based indicators can be proxies of soil functions related C, N and P cycling in drylands.

Land degradation tends to increase the amount of light reflected from the land in the range 0.28-6.0 μm because of reductions in vegetation cover, soil organic carbon and soil moisture, and by the increase in soil erosion rates in drylands worldwide (Tripathy et al., 1996; Robinove et al., 1980; Otterman, 1977; Zhang and Huang, 2004; Karnieli et al., 2014; Cierniewski et al., 2014). Thus, land surface albedo (defined as the ratio of irradiance reflected to the irradiance received by the land surface) should be negatively related to ecosystem functions and, as such, could be used as a proxy to monitor them (Tripathy et al., 1996; Karnieli et al., 2014). Albedo not only has commonly been used to monitor and assess dryland desertification (Liu et al., 2016; Lamchin et al., 2015; Karnieli et al., 2014; Tripathy et al., 1996; Robinove et al., 1980), but also has long been recognized as a primary controlling factor for the surface energy budget, and thus has been used to explain the climate change such as drought in drylands (Charney et al., 1975; Courel et al., 1984; Green et al., 2017; Yu et al., 2017). However, no study has, to the best of our knowledge, attempted to evaluate whether albedo is related to the provision of multiple ecosystem functions simultaneously (multifunctionality) in global drylands. We aimed to do so by assessing the correlation of albedo metrics with multifunctionality indices related to carbon, nitrogen and phosphorus cycling measured in 61 dryland ecosystems from all continents except Antarctica. Ecosystems are valued for the multiplicity of functions they perform and the services they provide (Cardinale et al., 2011; Reiss et al., 2009; Zavaleta et al., 2010), and thus focusing on multifunctionality can provide insights when monitoring changes in ecosystem functioning that cannot be detected when focusing on single functions (Berdugo et al., 2017; Lefchek et al. 2015).

While the total energy reflected by the Earth's surface in the shortwave domain is characterized by the shortwave (0.3-5.0μm) broadband albedo, the visible (0.3-0.7μm) and near-infrared (0.7-5.0μm) broadband albedos are often also of interest due to the marked difference in the reflectance of vegetation in these two spectral regions (Lucht et al., 2000). Because of this, we evaluated the correlation of multiple albedo metrics (shortwave black-sky albedo, shortwave white-sky albedo, visible black-sky albedo, visible white-sky albedo, near-infrared black-sky albedo and near-infrared white-sky albedo) with multifunctionality. Our study has two primary objectives: (i) to assess whether land surface albedo estimated from remote sensing correlates with multifunctionality measured on the ground at the global scale, and (ii) to determine the albedo metric that best correlates with multifunctionality.

2. Materials and methods

2.1. Study sites

Maestre et al. (2012) compiled a dataset of ecosystem structural and functional attributes from 224 dryland ecosystems (located in areas with aridity index between 0.05 and 0.65) from all continents except Antarctica between February 2006 and December 2010. Field data are available from 30 m × 30 m plots, while the moderate-resolution imaging spectroradiometer (MODIS) MCD43A1 BRDF/Albedo Model Parameters products are available at a resolution of 500 m×500 m (NASA LP DAAC, 2017a). Hence, and to minimize the scale mismatch between field and remote sensing data we only used those sites from the Maestre et al. (2012) dataset that had a high/very high degree of homogeneity, as visually interpreted by using high-resolution images in Google Earth 7 (Google Earth, 2014). Sites were labeled as having a ‘Very high’ and ‘High’ homogeneity if they had the same image tone, land cover, and terrain (i.e., no apparent elevation changes) in a circle with a radius of 500 m and 250 m centered around the plot surveyed in the field, respectively. This reduced the number of sites that could be used in our study to 61, of which 12 (20%), 45 (74%) and 4 (6%) were located in arid, semi-arid and dry sub-humid areas respectively (Fig. 1). We further checked the homogeneity of the sites selected by exploring their soil and vegetation characteristics according to the ISRIC SoilGrids (250 m × 250 m resolution, http://www.isric.org/explore/soilgrids) and ESA GlobCover2009 (300 m × 300 m resolution, http://due.esrin.esa.int/page_globcover.php) databases, respectively. For doing so, we counted the number of soil and vegetation types present in a circle of 250 m radius centered in each field site using ArcGIS 10.2 (see Table S1). We found that most of the selected 61 sites had only a single soil (89%) and vegetation (85%) type. Grasslands and shrublands dominated vegetation at the 61 selected sites. Their mean annual precipitation and temperature varied from 130 mm to 1219 mm, and from -1.8 ºC to 27.7 ºC, respectively. Their elevation ranged between 69 m and 4524 m, and their slopes varied between 0.5º and 8.6º. All sites faced southeast-southwest (Northern hemisphere) and northeast-northwest (Southern hemisphere). See Maestre et al. (2012) for additional details on the study sites.

Figure 1.

Distribution of the 61 study sites selected from the global dryland database of Maestre et al. (2012).

2.2. Assessing multifunctionality

To quantify multifunctionality, we used 14 soil variables (functions hereafter) related to C (organic carbon, β-glucosidase, phenols, aromatic compounds, hexoses and pentoses), N (total nitrogen, nitrate, ammonium, amino acids, proteins and potential nitrogen transformation rate) and P (available inorganic phosphorus and phosphatase) cycling. These variables, available from Maestre et al. (2012), measure either “true” ecosystem functions (Reiss et al., 2009; e.g. potential N transformation rate) or are key properties/processes (Jax, 2010; e.g. organic C, total N and soil enzymes) linked to nutrient cycling, primary productivity, and buildup of nutrient stocks. These “slow” variables also act as surrogates of important supporting (i.e., services that maintain the conditions for life on Earth) and regulating (i.e., benefits obtained from regulation of ecosystem processes) ecosystem services (Millennium Ecosystem Assessment, 2005; Isbell et al., 2011; Maestre et al., 2012) and have lengthy turnover times (Delgado-Baquerizo et al. 2017), being thus useful for gaining insights into long-term ecosystem changes and resource collapses (Reynolds et al., 2007a). Therefore, they have commonly been used when assessing changes in ecosystem functions along environmental gradients in drylands (Maestre and Escudero, 2009; Delgado-Baquerizo et al., 2013; Parras-Alcántara, 2015; Jiao et al., 2016). At each 30 m × 30 m plot, 10-15 soil samples were collected randomly placed under the canopy of the dominant perennial plants and in open areas devoid of perennial vegetation and were taken to the laboratory for further analysis. See Maestre et al. (2012) for additional details on the field sampling and the laboratory methods used.

To obtain multifunctionality indices, most functions measured were first normalized by using a sqrt-transformation; we then calculated their Z-scores. These were averaged to obtain a multifunctionality index (MUL). This index provides a straightforward and easily interpretable measure of the ability of different communities to sustain multiple ecosystem functions simultaneously (Byrnes et al., 2014). It is also statistically robust (Maestre et al., 2012), and is being increasingly used when assessing multifunctionality (e.g., Quero et al., 2013; Wagg et al., 2014; Valencia et al. 2015). Similarly, Z-score transformed variables from C, N and P cycling were averaged to form similar indices for carbon (CCY), nitrogen (NCY) and phosphorus (PCY) cycling. More details about how these indices were obtained can be found in Maestre et al. (2012). The primary statistical information of the 14 functions measured, and of the different indices evaluated (MUL, CCY, NCY, and PCY), were shown in Supplementary Table S2.

2.3. Land surface albedo

The actual land surface albedo depends on the atmospheric conditions besides surface reflective properties (Lucht et al., 2000; Wang et al., 2017). To avoid the influence of atmospheric conditions, two intrinsic land surface albedos, i.e., black-sky (BSA) and white-sky (WSA) albedos, were employed. BSA (directional hemispherical reflectance) albedo is defined as albedo in the absence of a diffuse component, and WSA (bi-hemispherical reflectance) albedo is defined as albedo in the absence of a direct component when the diffuse component is isotropic. Compared to the actual land surface albedo, which is interpolated between these two as a function of the fraction of diffuse skylight (itself a function of the aerosol optical depth; Lucht et al., 2000), both BSA and WSA are purely properties of the land surface, and do not depend on the state of the atmosphere (Strahler et al., 1999). They can be calculated as (Lucht et al., 2000):

$${\alpha }_{bs}(\theta ,b)=f_{iso}(b)(g_{0iso}+g_{1iso}{\theta }^2+g_{2iso}{\theta }^3)+f_{vol}(b)(g_{0vol}+g_{1vol}{\theta }^2+g_{2vol}{\theta }^3)+f_{geo}(b)(g_{0geo}+g_{1geo}{\theta }^2+g_{2geo}{\theta }^3)$$ (1)

$${\alpha }_{ws}(b)=f_{iso}(b)g_{iso}+f_{vol}(b)g_{vol}+f_{geo}(b)g_{geo}$$ (2)

where α_bs and α_ws are BSA and WSA; θ is the solar zenith angle (here 0 was set for all 61 sites because it has been reported that errors in MODIS algorithm of albedo estimation increase as the solar zenith angle increases; Liu et al., 2009); b is the spectral region; and f_iso, f_vol and f_geo are three weighting parameters of isotropic, volumetric and geometric-optical surface scattering, respectively. The others are constant polynomial coefficients given by Lucht et al. (2000).

The three weighting parameters were obtained from MCD43A1 BRDF/Albedo Model Parameters Product (Collection 5; NASA LP DAAC, 2017), and they rely on multi-date, cloud-cleared, atmospherically corrected surface reflectance data over a 16-day period from both MODIS and MISR instruments on Terra and Aqua Satellites. These data are provided for both MODIS narrow and broad (0.3-0.7μm, 0.7-5.0μm, and 0.3-5.0μm) bands every eight days since early 2000 with a resolution of 500m (Gao et al., 2005; Schaaf et al., 2002).

Considering the different assumptions involved with black-sky and white-sky albedos and the different reflectance of vegetation/soil in visible and near-infrared spectral regions, we calculated both black-sky and white-sky albedos for three different spectral domains. These albedos were: shortwave black-sky (SHO_BSA), shortwave white-sky (SHO_WSA), visible black-sky (VIS_BSA), visible white-sky (VIS_WSA), near-infrared black-sky (NIR_BSA) and near-infrared white-sky (NIR_WSA). Finally, to avoid snow cover and to make sure that vegetation leaves were green, we calculated the average albedo value (all albedo types) from May to September for sites located in the Northern hemisphere, and from November to March for those from the Southern hemisphere. These values were calculated for five years (2006-2010, according to the time of field data collection) by using MODIS Land Subsets (2010). The primary statistical information of the six albedo indices calculated is shown in Supplementary Table S2.

2.4. Vegetation indices

One of the VIs most widely used in remote sensing studies is NDVI (Verbesselt et al., 2016; Huang et al., 2017). The Enhanced Vegetation Index (EVI) was proposed to overcome some of the limitations of NDVI by incorporating both background adjustment and atmospheric resistance concepts into its calculations (Liu and Huete, 1995). EVI was adopted by the MODIS Land Discipline Group as the second global-based vegetation index (Huete, 1999), and has gained increased attention since then (Matsushita et al., 2007; Dubovyk et al., 2014; Seddon et al., 2016). Both NDVI and EVI have been successfully employed to predict ecosystem functions in drylands (García-Gómez and Maestre, 2011; Gaitán et al. 2013).

Both NDVI and EVI were obtained from the MOD13Q1 product (Collection 5; NASA LP DAAC, 2017b). This product has a spatial resolution of 250 m, and provides images every 16 days (16-day maximum value composites were calculated to reduce residual clouds and atmospheric effects). Data quality/reliability is indicated by flags of –1 (raw data or absent for different reasons), 0 (good quality data), 1 (useful data), 2 (snow/ice), 3 (cloudy). The VIs with flags of -1, 2 or 3 were replaced by the mean VIs value of the two closest dates having a pixel reliability 0 or 1. We then summed, respectively, the NDVI and EVI values obtained from May to September (sites in the Northern hemisphere) or from November to March (sites in the Southern hemisphere), hereafter referred as SumNDVI and SumEVI, respectively. Finally, the 5-year (2006-2010) average of SumNDVI values and that of SumEVI were used in further analyses.

2.5. Statistical analyses

We evaluated the relationships between the six types of albedo and the four multifunctionality indices, and that between VIs (SumNDVI and SumEVI) and these indices, using Spearman correlation analysis. Furthermore, and to check whether the direction and magnitude of the correlation would change depending on the sites selected, we calculated the correlation between albedos and multifunctionality indices, and that between VIs (SumNDVI and SumEVI) and these indices, based on a randomly selected subset of 40 sites from the 61 sites used in this study, and repeated this process 100 times. The number of positive (+) and negative (-) correlations were counted, and their magnitude was quantified by averaging the positive and negative correlation coefficients, respectively. All the statistical analyses were performed with SPSS for Windows, version 20.0 (SPSS Inc., Chicago, IL, USA).

3. Results

Of the 24 correlation coefficients obtained between albedos and multifunctionality indices, 37.5% were significant at the 0.05 level (Table 1). We found a negative correlation between land surface albedo and multifunctionality (Table 1; see Supplementary Figs. S1-S8 for scatterplots). Black-sky albedo had a stronger correlation with multifunctionality than white-sky albedo. The multifunctionality index and similar indices obtained with carbon and nitrogen cycling had a stronger correlation with visible black-sky albedo (VIS_BSA) than with near-infrared and shortwave albedos (NIR_BSA and SHO_BSA), while PCY showed a stronger correlation with NIR_BSA and SHO_BSA than with VIS_BSA.

Table 1.

Spearman correlations of six albedo types and two VIs with four multifunctionality indices. CCY: carbon cycling index; NCY: nitrogen cycling index; PCY: phosphorus cycling index; MUL: multifunctionality index; SHO_BSA, VIS_BSA, NIR_BSA are black-sky albedos in shortwave, visible and near-infrared regions respectively; SHO_WSA, VIS_WSA, NIR_WSA are white-sky albedos in shortwave, visible and near-infrared regions respectively. SumNDVI and SumEVI are the sum of NDVI/EVI during the growing season. Significant (P < 0.05) correlations are in bold. Correlation results of back-sky albedo types with multifunctionality indices are in grey.

	MUL		CCY		NCY		PCY
	P	P	ρ	P	ρ	P	ρ	P
SHO_BSA	-0.211	0.102	-0.058	0.658	-0.295	0.021	-0.325	0.011
SHO_WSA	-0.156	0.230	-0.026	0.842	-0.239	0.064	-0.252	0.050
VIS_BSA	-0.314	0.014	-0.216	0.094	-0.410	0.001	-0.233	0.071
VIS_WSA	-0.298	0.020	-0.236	0.067	-0.382	0.002	-0.177	0.172
NIR_BSA	-0.162	0.212	-0.024	0.853	-0.218	0.092	-0.339	0.008
NIR_WSA	-0.098	0.452	0.021	0.873	-0.158	0.225	-0.260	0.043
SumNDVI	0.425	0.001	0.360	0.004	0.496	< 0.001	0.176	0.176
SumEVI	0.365	0.004	0.334	0.009	0.408	0.001	0.109	0.403

The two VIs both showed significant positive correlations with MUL, CCY, and NCY, and the magnitudes of the correlation coefficients obtained were higher than that between VIS_BSA and the three multifunctionality indices. It was interesting to note that the two VIs were not correlated with PCY, which had significant correlations with both shortwave and near-infrared albedo.

The results of correlation analysis based on randomly selected sites showed that most of the correlations observed were negative, with the exceptions of the correlations of NIR_WSA and NIR_BSA with CCY (Table 2). We did not find any clear directionality in these correlations, which agrees with the very low correlation coefficients of both NIR_BSW and NIR_WSA with CCY obtained when using the 61 sites (Table 1). The coefficients obtained for black-sky albedos had a stronger correlation with ecosystem multifunctionality than those from white-sky albedos. Among the black-sky albedos (gray background in Table 3), MUL, CCY, and NCY all had a stronger correlation with VIS_BSA than with NIR_BSA and SHO_BSA, while PCY showed a stronger correlation with NIR_BSA and SHO_BSA than with VIS_BSA.

Table 2.

Average Spearman correlations of six albedo types and two VIs with four multifunctionality indices based on data from 40 sites randomly selected. D represents the direction of the correlation of albedo with multifunctionality indices, ‘+’ represents positive and ‘-’ represents negative; N represents the number of times where a given direction (‘+’/‘-’) of the correlations is found; Negative correlations of back-sky albedo types with multifunctionality indices are in grey. The rest of abbreviations are same to those reported in Table 1.

	MUL			CCY			NCY			PCY
	D	N	ρ	D	N	ρ	D	N	ρ	D	N	ρ
SHO_BSA	+	1	0.009	+	38	0.054	+	0		+	0
	-	99	-0.216	-	62	-0.123	-	100	-0.259	-	100	-0.326
SHO_WSA	+	5	0.036	+	43	0.077	+	1	0.056	+	0
	-	95	-0.169	-	57	-0.100	-	99	-0.124	-	100	-0.254
VIS_BSA	+	0		+	1	0.053	+	0		+	0
	-	100	-0.322	-	99	-0.226	-	100	-0.416	-	100	-0.242
VIS_WSA	+	0		+	1	0.026	+	0		+	3	0.078
	-	100	-0.305	-	99	-0.243	-	100	-0.387	-	97	-0.194
NIR_BSA	+	5	0.027	+	47	0.072	+	1	0.047	+	0
	-	95	-0.176	-	53	-0.107	-	99	-0.221	-	100	-0.338
NIR_BSA	+	18	0.053	+	57	0.099	+	8	0.041	+	0
	-	82	-0.134	-	43	-0.074	-	92	-0.175	-	100	-0.259
SumNDVI	+	100	0.420	+	100	0.361	+	100	0.488	+	97	0.180
	-			-			-			-	3	-0.027
SumEVI	+	100	0.360	+	100	0.334	+	100	0.400	+	86	0.133
	-			-			-			-	14	-0.053

4. Discussion and conclusions

As hypothesized, we found that land surface albedo had a negative relationship with ecosystem multifunctionality. This is in agreement with the fact that the amount of light reflected from the land in the range 0.28 to 6.0μm tends to increase due to a reduction in vegetation (Tripathy et al., 1996). Our results also agree with studies showing positive relationships between vegetation cover and ecosystem functions linked to soil nutrient cycling and storage in drylands (Maestre and Escudero, 2009; Maestre et al., 2016).

Though white-sky and black-sky albedos are both associated with the intrinsic optical properties of the land surface (Strahler et al., 1999), assumptions involved in the calculation of white-sky albedo with the algorithms used by MODIS (diffuse skylight is an isotropic distribution, and multiple interactions between the ground and atmosphere are ignored) may explain their low correlations observed with our multifunctionality indices. It has been reported that errors resulting from the assumptions involved in the albedo calculations can be as high as 10%, exceeding the accuracy required by many climate applications (Román et al., 2010; Pinty et al. 2005).

Our results showed that visible black-sky albedo had the strongest correlation with MUL, CCY, and NCY, while near-infrared black-sky albedo had the strongest correlation with PCY. Both NDVI and EVI showed stronger correlations with MUL, CCY and NCY than with PCY. The different mechanisms involved in carbon, nitrogen and phosphorus cycling can account for these results. Carbon and nitrogen cycling are primarily linked to biological processes such as photosynthesis, atmospheric nitrogen fixation and subsequent microbial mineralization (Schlesinger, 1996; Vitousek, 2004). Plants absorb visible spectral regions during photosynthesis, particularly the blue and red light, and carbon fixation by this process largely affects the stocks and cycling of carbon and nitrogen in the soil (Finzi et al., 2011). Thus, is not surprising to find that VIs were poorly correlated with PCY, and that visible albedos showed stronger correlations with CCY and NCY than near-infrared albedo. The phosphorus cycle is mainly linked to rock weathering (Cross and Schlesinger, 2001; Lajtha and Schlesinger, 1988) and the release of available phosphorus is affected by soil moisture (Tate and Salcedo, 1988), a key determinant of albedo in drylands (He et al., 2014; Dorigo et al., 2012). Decreasing moisture would result in an increased albedo due to the spectral absorption of soil moisture, and the maximum absorptions occur at 1.4, 1.9 and 2.2μm, which are all in the infrared region. This could explain the correlations between near-infrared black-sky albedo and functions related to phosphorus cycling.

Previous studies have highlighted the importance of climatic variables as drivers of changes in ecosystem functioning in drylands at regional and global scales (e.g., Sala et al., 2012; Gaitan et al., 2014; Maestre et al., 2016). Albedo has been long recognized as a primary controlling factor for the surface energy budget, and the classic albedo-based-biogeophysical feedback (Charney et al., 1975) has been used to study the dynamics of drought in drylands (Courel et al., 1984; Green et al., 2017; Yu et al., 2017). The finding of a global albedo-multifunctionality relationship across the variability in soil features and vegetation characteristics accounted by our survey is noteworthy, and set the stage for further explorations of the links between ecosystem functioning and abiotic features using remotely sensed albedo proxies. In this regard, it is worth noting that both the shortwave and near-infrared albedo had significant correlations with the phosphorus cycling index (PCY), which was not correlated with VIs. These results suggest that the albedo indices used can capture some aspects of ecosystem multifunctionality that are not fully captured by commonly used vegetation indices such as NDVI and EVI.

Overall, our results indicate that the visible black-sky albedo is correlated mainly with functions related to carbon and nitrogen cycling, while the near-infrared black-sky albedo shows significant correlations with functions related to phosphorus cycling. Though the magnitudes of correlations were not high, something that which may result from the wide variety of soil and vegetation types found in our dataset; our findings suggest that albedo indices derived from the MODIS MCD43A1 product can be linked to ecosystem multifunctionality in drylands from all over the world. However, it should be noted that more research is needed before albedo can be reliably employed as a proxy of ecosystem function when comparing different sites because of the spatial heterogeneity of vegetation (e.g., types, structures) and soil (e.g., white lichen, CaCO₃ content) properties, which can influence albedo values. Despite this, and given the strong relationships between albedo estimated with remote sensing tools and vegetation cover (Chopping et al., 2012; Brovkin et al., 2013), and between cover and ecosystem functioning in dryland ecosystems (e.g. Maestre & Escudero, 2009; Maestre et al., 2016; Berdugo et al., 2017), our results suggest that albedo indices have the potential to be incorporated into the suite of ground- and remote-sensing based tools currently available to monitor changes in ecosystem functioning and degradation processes in drylands.

Supplementary Material

Supplementary data related to this article can be found at https://doi.org/10.1016/j.jaridenv.2018.05.010.