To classify trees along a spectrum of isohydric to anisohydric behavior is a promising new framework for identifying tree species' sensitivities to drought stress, directly related to the vulnerability of carbon uptake of terrestrial ecosystems with increased hydroclimate variability. Trees with isohydric strategies regulate stomatal conductance to maintain stationary leaf water potential, while trees with anisohydric strategies allow leaf water potential to fall, which in the absence of significant hydraulic cavitation will facilitate greater rates of carbon uptake. Despite the recognition of the gas exchange consequences of isohydric and anisohydric strategies for individual tree species, there have been few studies regarding whether isohydric trees produces distinct spectral signatures under drought stress that can be remotely sensed. Here, we examined the capability of four vegetation indices (PRI, NDVI, NDVI705, and EVI) to capture the differences in spectral responses between isohydric and anisohydric trees within a deciduous forest in central Indiana, USA. Both leaf-level spectral measurements and canopy-scale satellite observations were used to compare peak growing-season spectral signatures between a drought and a non-drought year. At the leaf scale, two vegetation indices (NDVI and NDVI705) failed to capture the drought signal or the divergent isohydric/anisohydric behavior. EVI successfully captured the drought signal at both leaf and canopy scales, but failed to capture the divergent behavior between isohydric and anisohydric tree species during the drought. PRI captured both drought signals and divergent isohydric/anisohydric behavior at both leaf and canopy scales once normalized between sunlit (backward direction images) and shaded (forward direction images) portions of canopy, which indicates drought stress and subsequent photosynthetic downregulation are greater in the sunlit portion of canopy. This study presents a significant step forward in our ability to directly mapping emergent isohydricity at different scales based on divergent spectral signatures between sunlit and shaded canopies.

Predicted decreases in water availability across the temperate forest biome have the potential to offset gains in carbon (C) uptake from phenology trends, rising atmospheric CO2, and nitrogen deposition. While it is well established that severe droughts reduce the C sink of forests by inducing tree mortality, the impacts of mild but chronic water stress on forest phenology and physiology are largely unknown. We quantified the C consequences of chronic water stress using a 13‐year record of tree growth (n = 200 trees), soil moisture, and ecosystem C balance at the Morgan–Monroe State Forest (MMSF) in Indiana, and a regional 11‐year record of tree growth (n > 300 000 trees) and water availability for the 20 most dominant deciduous broadleaf tree species across the eastern and midwestern USA. We show that despite ~26 more days of C assimilation by trees at the MMSF, increasing water stress decreased the number of days of wood production by ~42 days over the same period, reducing the annual accrual of C in woody biomass by 41%. Across the deciduous forest region, water stress induced similar declines in tree growth, particularly for water‐demanding ‘mesophytic’ tree species. Given the current replacement of water‐stress adapted ‘xerophytic’ tree species by mesophytic tree species, we estimate that chronic water stress has the potential to decrease the C sink of deciduous forests by up to 17% (0.04 Pg C yr−1) in the coming decades. This reduction in the C sink due to mesophication and chronic water stress is equivalent to an additional 1–3 days of global C emissions from fossil fuel burning each year. Collectively, our results indicate that regional declines in water availability may offset the growth‐enhancing effects of other global changes and reduce the extent to which forests ameliorate climate warming.

The bi-directional reflectance distribution function (BRDF) has been widely studied across different vegetation types. However, these studies generally report values for only one point in time. We were interested in the potential for seasonal and inter-annual variation in BRDF parameters. NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on board the EOS satellites has now been collecting data for 10 years. Since BRDF parameters are reported for the individual spectral bands, these data can be used to examine intra-annual variation. However, MODIS BRDF parameters are not calculated for the various vegetation indices which are derived from the spectral bands. Our objective in this study was to use the 10 years of MODIS data now available to examine seasonal and inter-annual variation in the view angle sensitivity of three vegetation indices; the normalized difference vegetation index (NDVI), the enhanced vegetation index (EVI), and the photochemical reflectance index (PRI) at 3 flux tower sites (Harvard Forest, Howland Forest and Morgan Monroe State Forest). For these 3 sites, only EVI was significantly affected by view angle. There was also a substantial variation in the view angle sensitivity of EVI across seasons and this variation was different for backscatter vs. forward scatter data. It is possible that differences in the scattering of radiation between the spring and the fall are responsible for the seasonal difference in view angle sensitivity. There were also complimentary variations in forward and backscatter view angle sensitivity of EVI across years. The greater view angle sensitivity of EVI, as opposed to NDVI, suggests that greater care must be taken to correct for BRDF effects when using this vegetation index.

Spectral Network (SpecNet) began as a Working Group in 2003 with the goals of integrating remote sensing with biosphere-atmosphere carbon flux measurements and standardizing field optical sampling methods. SpecNet has evolved into an international network of collaborating sites and investigators, with a particular focus on matching optical sampling tools to the temporal and spatial scale of flux measurements and ecological sampling. Current emphasis within the SpecNet community is on greater automation of field optical sampling using simple cost-effective technologies, improving the light-use-efficiency (LUE) model of carbon dioxide flux, consideration of view and illumination angle to improve physiological retrievals, and incorporation of informatics and cyberinfrastructure solutions that address the increasing data dimensionality of cross-site and multiscale sampling. In this review, we summarize recent findings and current directions within the SpecNet community and provide recommendations for the larger remote sensing and flux communities. These recommendations include comparing the LUE model to other flux models driven by remote sensing, considering a wider array of biogenic trace gases in addition to carbon dioxide, adoption of standardized and automated field sensors and sampling protocols where possible, continued development of cyberinfrastructure tools to facilitate data comparison and integration, expanding the network itself so that a greater range of sites are covered by combined optical and flux measurements, and encouraging a broader communication between the flux and remote sensing communities.

Many current models of ecosystem carbon exchange based on remote sensing, such as the MODIS product termed MOD17, still require considerable input from ground based meteorological measurements and look up tables based on vegetation type. Since these data are often not available at the same spatial scale as the remote sensing imagery, they can introduce substantial errors into the carbon exchange estimates. Here we present further development of a gross primary production (GPP) model based entirely on remote sensing data. In contrast to an earlier model based only on the enhanced vegetation index (EVI), this model, termed the Temperature and Greenness (TG) model, also includes the land surface temperature (LST) product from MODIS. In addition to its obvious relationship to vegetation temperature, LST was correlated with vapor pressure deficit and photosynthetically active radiation. Combination of EVI and LST in the model substantially improved the correlation between predicted and measured GPP at 11 eddy correlation flux towers in a wide range of vegetation types across North America. In many cases, the TG model provided substantially better predictions of GPP than did the MODIS GPP product. However, both models resulted in poor predictions for sparse shrub habitats where solar angle effects on remote sensing indices were large. Although it may be possible to improve the MODIS GPP product through improved parameterization, our results suggest that simpler models based entirely on remote sensing can provide equally good predictions of GPP.

The goal of this study was to explore the utility of the 970 nm water band index (WBI) in estimating evapotranspiration and vegetation water status for a semiarid shrubland ecosystem. Between 2001 and 2003, spectral reflectance coupled with CO2 and water flux data were collected at Sky Oaks Biological Field Station, a chaparral-dominated ecosystem in southern California, and one of the sites within the SpecNet network. The reflectance data were collected either by walking along a 100 m transect or by using a semi-automated tram system installed later at the site along the same 100 m transect. CO2 and water flux data were gathered with an eddy covariance flux tower adjacent to the tram system. The 970 nm WBI and normalized difference vegetation index (NDVI) were derived from the spectral reflectance. The two indices were expressed both as points approximately a meter apart along the transect and as whole-transect averages, where all of the reflectance values along the transect were averaged together, simulating a large pixel. This study encompassed a wet year with normal precipitation (2001), a 100-year record drought (2002), and a recovery year (2003), allowing for comparison over time and between precipitation regimes. Species-specific responses to wet and dry periods were evident in the reflectance spectra, providing a basis for separating species based on their optical properties. The WBI was significantly correlated with the NDVI revealing a strong link between canopy water content and green canopy structure; however this relationship varied with species and water status, providing evidence for the independence of these two optical indices. The WBI was also strongly linked to surface-atmosphere fluxes, explaining 49% of the variance in the water vapor flux, and 24% of the carbon dioxide fluxes. These results suggest that WBI or other similar water status indices may be useful variables in modeling CO2 and water fluxes when combined with other physiological, environmental, and atmospheric factors. (c) 2006 Published by Elsevier Inc.

Current remote sensing based models of vegetation carbon exchange are complex and require a large number of input parameters that can not be determined directly from remote sensing. Models based entirely on remote sensing data would simplify the estimation of vegetation carbon fluxes and potentially reduce errors associated with estimation of ground based parameters. We used hyperspectral image data from the AVIRIS sensor and vegetation carbon exchange data from eddy covariance towers in the Ameriflux network to explore relationships between spectral reflectance indices and either gross carbon flux or the light use efficiency (LUE) of photosynthesis. Indices related to the greenness of the vegetation proved to be the best predictors of both gross flux and LUE. The enhanced vegetation index (EVI), which has greater sensitivity to variation in greenness of dense canopies than does NDVI, was most strongly correlated with both gross flux and LUE. The photochemical reflectance index (PRI), which has shown promise as a predictor of LUE in other studies, did not show a consistent relationship across all the vegetation types. The relationships between spectral indices and either gross flux or LUE were just as good when the flux data were averaged over one week as they were for flux data specific to the time of the spectral data collection. Consequently, these relationships appear to be useful for estimation of 8-day composites, as is done with MODIS. Work is continuing using MODIS satellite data to test these relationships across a wider range of sites.

Using field data, Airborne Visible Infrared Imaging Spectrometer (AVIRIS) imagery, and Moderate-Resolution Imaging Spectroradiometer (MODIS) data, a multi-scale analysis of ecosystem optical properties was performed for Sky Oaks, a Southern California chaparral ecosystem in the spectral network (SpecNet) and FLUXNET networks. The study covered a 4-year period (2000–2004), which included a severe drought in 2002 and a subsequent wildfire in July 2003, leading to extreme perturbation in ecosystem productivity and optical properties. Two vegetation greenness indices (Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI)), and a measure of the fraction of photosynthetically active radiation absorbed by vegetation (fPAR), were compared across sampling platforms, which ranged in pixel size from 1 m (tram system in the field) to 1000 m (MODIS satellite sensor). The three MODIS products closely followed the same seasonal trends as the tram and AVIRIS data, but tended to be higher than the tram and AVIRIS values, particularly for fPAR and NDVI. Following a wildfire that removed all green vegetation, the overestimation in MODIS fPAR values was particularly clear. The MODIS fPAR algorithm (version 4 vs. v.4.1) had a significant effect on the degree of overestimation, with v. 4.1 improving the agreement with the other sensors (AVIRIS and tram) for vegetated conditions, but not for low, post-fire values. The differences between MODIS products and the products from the other platform sensors could not be entirely attributed to differences in sensor spectral responses or sampling scale. These results are consistent with several other recently published studies that indicate that MODIS overestimates fPAR and thus net primary production (NPP) for many terrestrial ecosystems, and demonstrates the need for proper validation of MODIS terrestrial biospheric products by direct comparison against optical signals at other spatial scales, as is now possible at several SpecNet sites. The study also demonstrates the utility of in-situ field sampling (e.g. tram systems) and hyperspectral aircraft imagery for proper interpretation of satellite data taken at coarse spatial scales.

Some form of the light use efficiency (LUE) model is used in most models of ecosystem carbon exchange based on remote sensing. The strong relationship between the normalized difference vegetation index (NDVI) and light absorbed by green vegetation make models based on LUE attractive in the remote sensing context. However, estimation of LUE has proven problematic since it varies with vegetation type and environmental conditions. Here we propose that LUE may in fact be correlated with vegetation greenness (measured either as NDVI at constant solar elevation angle, or a red edge chlorophyll index), making separate estimates of LUE unnecessary, at least for some vegetation types. To test this, we installed an automated tram system for measurement of spectral reflectance in the footprint of an eddy covariance flux system in the Southern California chaparral. This allowed us to match the spatial and temporal scales of the reflectance and flux measurements and thus to make direct comparisons over time scales ranging from minutes to years. The 3-year period of this study included both “normal” precipitation years and an extreme drought in 2002. In this sparse chaparral vegetation, diurnal and seasonal changes in solar angle resulted in large variation in NDVI independent of the actual quantity of green vegetation. In fact, one would come to entirely different conclusions about seasonal changes in vegetation greenness depending on whether NDVI at noon or NDVI at constant solar elevation angle were used. Although chaparral vegetation is generally considered “evergreen”, we found that the majority of the shrubs were actually semi-deciduous, leading to large seasonal changes in NDVI at constant solar elevation angle. LUE was correlated with both greenness indices at the seasonal timescale across all years. In contrast, the relationship between LUE and PRI was inconsistent. PRI was well correlated with LUE during the “normal” years but this relationship changed dramatically during the extreme drought. Contrary to expectations, none of the spectral reflectance indices showed consistent relationships with CO2 flux or LUE over the diurnal time-course, possibly because of confounding effects of sun angle and stand structure on reflectance. These results suggest that greenness indices can be used to directly estimate CO2 exchange at weekly timescales in this chaparral ecosystem, even in the face of changes in LUE. Greenness indices are unlikely to be as good predictors of CO2 exchange in dense evergreen vegetation as they were in the sparse, semi-deciduous chaparral. However, since relatively few ecosystems are entirely evergreen at large spatial scales or over long time spans due to disturbance, these relationships need to be examined across a wider range of vegetation types.

Using simple models derived from spectral reflectance, we mapped the patterns of ecosystem CO2 and water fluxes in a semi-arid site in southern California during a period of extreme disturbance, marked by drought and fire. Employing a combination of low (similar to 2 km) and high (similar to 16 km) altitude images from the hyperspectral Airborne Visible Infrared Imaging Spectrometer (AVIRIS), acquired between April 2002 and September 2003, and ground data collected from an automated tram system, several vegetation indices were calculated for Sky Oaks field station, a FLUX-NET and SpecNet site located in northern San Diego County (CA, USA). Based on the relationships observed between the fluxes measured by the eddy covariance tower and the vegetation indices, net CO2 and water vapor flux maps were derived for the region around the flux tower. Despite differences in the scale of the images (from similar to 2 in to 16 m pixel size) as well as marked differences in environmental conditions (drought in 2002, recovery in early 2003, and fire in mid 2003), net CO2 and water flux modeled from AVIRIS-derived reflectance indices (NDVI, PRI and WBI) effectively tracked changes in tower fluxes across both drought and fire, and readily revealed spatial variation in fluxes within this landscape. After an initial period of net carbon uptake, drought and fire caused the ecosystem to lose carbon to the atmosphere during most of the study period. Our study shows the power of integrating optical and flux data in LUE models to better understand factors driving surface-atmosphere carbon and water vapor flux cycles, one of the main goals of SpecNet. (c) 2006 Published by Elsevier Inc.

Reliable and repeatable field sampling methods are needed for monitoring ecosystem optical properties linked to carbon flux. Here we describe a tram system, consisting of a dual-detector spectrometer mounted on a robotic cart for mobile sampling of ecosystem spectral reflectance. To illustrate the application of this system for monitoring dynamic ecosystem activity, we illustrate how the tram can be used for exploring the multiple factors influencing the Normalized Difference Vegetation Index (NDVI), a measure of vegetation greenness and a key optical indicator of vegetation carbon dioxide assimilation. With this system, we collected five years of NDVI data for a chaparral ecosystem in Southern California subject to extreme disturbance. Key factors affecting NDVI at this site included snow cover, sky conditions (clear vs. cloudy), time of day, season, species composition, and environmental perturbations such as rainfall, drought and fire. Applications of this tram system include ecosystem monitoring, satellite validation, and developing surface-atmosphere flux models from remote sensing.