B21A-0021
Using Multiple Endmember Spectral Mixture Analysis of MODIS Data for Computing the Fire Potential Index in Southern California
The Fire Potential Index (FPI) is currently the only operationally used wildfire susceptibility index in the United States that incorporates remote sensing data in addition to meteorological information. Its remote sensing component utilizes relative greenness derived from a NDVI time series as a proxy for computing the ratio of live to dead vegetation. This study investigates the potential of Multiple Endmember Spectral Mixture Analysis (MESMA) as a more direct and physically reasonable way of computing the live ratio and applying it for the computation of the FPI. A time series of 16-day reflectance composites of Moderate Resolution Imaging Spectroradiometer (MODIS) data was used to perform the analysis. Endmember selection for green vegetation (GV), non- photosynthetic vegetation (NPV) and soil was performed in two stages. First, a subset of suitable endmembers was selected from an extensive library of reference and image spectra for each class using Endmember Average Root Mean Square Error (EAR), Minimum Average Spectral Angle (MASA) and a count-based technique. Second, the most appropriate endmembers for the specific data set were selected from the subset by running a series of 2-endmember models on representative images and choosing the ones that modeled the majority of pixels. The final set of endmembers was used for running MESMA on southern California MODIS composites from 2000 to 2006. 3- and 4-endmember models were considered. The best model was chosen on a per-pixel basis according to the minimum root mean square error of the models at each level of complexity. Endmember fractions were normalized by the shade endmember to generate realistic fractions of GV and NPV. In order to validate the MESMA-derived GV fractions they were compared against live ratio estimates from RG. A significant spatial and temporal relationship between both measures was found, indicating that GV fraction has the potential to substitute RG in computing the FPI. To further test this hypothesis the live ratio estimates obtained from MESMA were used to compute daily FPI maps for southern California from 2001 to 2006. A validation with historical wildfire data from the MODIS Active Fire product was carried out over the same time period using logistic regression. Initial results show that MESMA-derived GV fraction can be used successfully for generating FPI maps of southern California.
B21A-0022
Remotely sensed vicennial changes of green phytomass, Salix cover, and leaf turnover in a sedge-shrub tundra, Arctic National Wildlife Refuge, Alaska
We obtained the relationship between spectral indices, green phytomass, Salix - non-Salix ratio, and leaf turnover in a sedge-shrub tundra, Arctic National Wildlife Refuge (ANWR), Alaska based on the field observations of spectral reflectance and phytomass, and we used Landsat TM images acquired in July of 1986, 1994, and 2006 and the time series of NOAA AVHRR (Advanced Very High Resolution Radiometer) for evaluating the vicennial changes. 51% of Beaufort coastal plain, Alaska was occupied by lowland moist sedge-shrub tundra, lowland wet sedge tundra, riverine moist sedge-shrub tundra, and riverine wet sedge tundra, where willow shrubs and sedges dominate. We set a 50-m × 50-m plot located on the floodplain of Jago River in ANWR. Shrub (Salix lanata L.) and sedge (Carex bigelowii Torr.) dominated in the plot. Ten 0.5-m × 0.5-m quadrates (Salix} quadrates) were set on the Salix cover and ten 0.5-m × 0.5-m quadrates (non-Salix quadrates) were set on the ground that was not covered with Salix lanata. Salix lanata in each of the Salix quadrates was harvested, and the leaf area index (LAI) and the oven-dried weights of the photosynthetic (leaf) and non-photosynthetic parts were measured. After harvesting Salix, other green plants were harvested and the oven-dried weights of the plants were measured. The Salix quadrates were spectrally measured with a spectroradiometer at a wavelength of 350 - 2500 nm before and after harvesting Salix and after harvesting other green plants. Non-Salix quadrates were also spectrally measured with the spectroradiometer. The coefficients of determination (R2) of the green phytomass, Salix - non-Salix ratio, and leaf turnover estimations from the spectral indices were 0.63, 0.57, and 0.79, respectively. These estimations were used for evaluating the vicennial changes using the satellite data.
B21A-0023
Evaluating Spatial Patterns of Land Use and urban Heat Island in The Fast Growing Metropolitan Shanghai, China
Abstract Remotely sensed data (Landsat TM5) were used to quantitatively characterize the patterns of land use and urban heat island (UHI) in the fast growing Metropolitan Shanghai, China. Results showed that, with dramatic change in land use and land cover driven by substantial economic growth since the 1990s, rapid expansion of the urbanized and urbanizing areas occurred at regional level during 1997 and 2004. Similarly, both the extent and magnitude of UHI in Shanghai have undergone a significant increase, though some newly emerging cooling patches were detected in the central urban area. On small and meso scales, a significant spatial patterning was present in UHI as indicated by land surface temperature (LST). Moreover, based on the satellite images, the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Normalized Difference Bareness Index (NDBaI), and Normalized Difference Build-up Index (NDBI) were produced to explore the relationship between land use and UHI effect. Although these indices were effective in characterizing the spatial and temporal patterns of UHI, there were some unexplainable factors due to the complexity in ecological process. As a whole, it can be predicted that the ongoing urban sprawl in the satellite towns will adversely cause a long term effect on regional atmospheric environment. Keywords Spatial pattern; Urban heat island (UHI); Land surface temperature (LST);urban sprawl ;Shanghai; China.
B21A-0024
Measurement of Eastern Siberian larch forest LAI using the Normalized Difference Water Index
In this study, we describe a new remote sensing method to measure the canopy leaf area index (LAI) over the larch forests in Eastern Siberia, and we compare the measured LAI to ground observations and to other remote sensing products. The method was established using a set of radiative transfer simulations for several scenes representative of the larch forest structure, including the typical clumped shoot structure. The results indicate that the Normalized Difference Water Index (NDWI) is more sensitive than the Normalized Difference Vegetation Index (NDVI) for higher LAI, and that the dNDWI, which is the increase in NDWI from the leaf appearance date, is a good indicator for canopy LAI estimation. Based on these simulation results, we developed a semi-empirical method to measure the canopy LAI in larch forests during the growing season: we first estimate the date of canopy leaf appearance, then the forest floor conditions, and then the LAI seasonal variations. The algorithm was then applied to the reflectance measured by an airborne sensor in the Yakutsk region. The results were compared to the time series of LAI measured in situ at four sites, showing that the timing and magnitude are of the LAI increase are correct. Then, the algorithm was applied to the SPOT VEGETATION S10 reflectance, and the measured LAI were compared to the LAI from the MODIS MOD15 collection 5 dataset and to the CYCLOPE dataset. The LAI time series from our algorithm are very close to those from CYCLOPE in term of timing and magnitude. In contrary, the MOD15 LAI magnitude is larger than the other two datasets, and it starts increasing earlier than them and than the in situ time series, indicating the MOD15 collection 5 may be unreliable over larch forests in Eastern Siberia.
B21A-0025
An integrated approach on land surface water mapping over Asia with AMSR-E and MTSAT
This research focuses on an integrated monitoring of land surface water with AMSR-E and MTSAT. The combination of optical and passive microwave sensors and innovative processing methodologies are used to measure river discharge, floods, droughts, watershed runoff, agriculture activities and wetland mapping over time. A time series of analysis on rice crop cultivation and flooding event was conducted on vegetation and inundation condition with MTSAT derived indices and those of AMSR-E. It was found that AMSR-E captures a salient points on rice cropping phenology more precisely than MTSAT because it is not affected by cloud contamination. A all-weather type monitoring with high frequency can be effective especially on inundation condition in wet season along with MTSAT data with high spatial resolution. This system is running as fully operational and can be accessed at http://webmodis.iis.u-tokyo.ac.jp/LSW/. http://webmodis.iis.u- tokyo.ac.jp/LSW/
B21A-0026
Comparison of Kriging and LOCFIT methods for interpolating gridded passive microwave brightness temperatures
Satellite passive microwave brightness temperatures (TBs) are used as the basis for measuring various land surface properties, including soil moisture and snow water equivalent. The current satellite record of brightness temperatures includes data from the SMMR, SSM/I and AMSR-E instruments, beginning in 1978 and continuing to the present day. These sensors fly on satellite platforms in sun-synchronous, polar orbits, providing near-daily global coverage of the Earth. Due to the satellite geometry, surface locations at high latitudes receive better than daily coverage, while locations at lower latitudes are observed less frequently. We have compiled a nearly 30- year record of daily, gridded passive microwave temperatures, but the lack of daily coverage from the respective sensors makes the compilation of daily derived products complicated. Techniques we have used in the past have been performed after deriving the desired geophysical parameter, and have included last-in compilations and piecewise, linear interpolation of missing snow water equivalent between days with legitimate observations. The advantages of these methods are that they are relatively simple to implement, but they suffer from not making use of any known physical spatial correlations at the brightness temperature level with neighboring locations for the data being interpolated. Our study will compare two spatial interpolation methods at the gridded brightness temperature level that take advantage of spatial and temporal correlation: kriging methods and local polynomial (LOCFIT) fitting. Our study area includes a subset of Equal-Area Scalable Earth Grid (EASE-Grid) brightness temperatures, for a region that includes portions of the Western United States and Canada, for two weeks in both January and July, for the period of record. Results of this analysis will ultimately increase our skill in filling in the gaps in microwave coverage, thus improving existing gridded brightness temperature data sets and the geophysical products derived from them.
B21A-0027
Partitioning incident radiation fluxes based on photon recollision probability in vegetation canopies
Remote sensing of vegetation and modeling of canopy microclimate requires information on the fractions of incident radiation reflected, transmitted and absorbed by a plant canopy. The photon recollision probability p allows to calculate easily the amount of radiation absorbed by a vegetation canopy and to predict the spectral behavior of canopy scattering, i.e. the sum of canopy reflectance and transmittance. However, to divide the scattered radiation into reflected and transmitted fluxes, additional models are needed. To overcome this problem, we present a simple formula based on the photon recollision probability p to estimate the fraction of radiation scattered upwards by a canopy. The new semi-empirical method is tested with Monte Carlo simulations. A comparison with the analytical solution of the two-stream equation of radiative transfer in vegetation canopies is also provided. Our results indicate that the method is accurate for low to moderate leaf area index (LAI) values, and provides a reasonable approximation even at LAI=8. Finally, we present a new method to compute p using numerical radiative transfer models.
B21A-0028
Relationship Between Satellite-derived Phenology and Climatic Factors Over Northeastern Asia
Phenology means seasonal activities of vegetation, such as green-up, flowering, leaves-coloring or leaves-dropping. It is closely related to seasonal dynamics of the lower atmosphere and important elements in global models and vegetation monitoring. Time-series NDVI data derived from AVHRR or MODIS are suitable for phenological monitoring, because these sensors provide data with a high temporal frequency. In this study, we analyzed variations in green-up date over northeastern Asia from 1984 to 2004 with NOAA AVHRR data received at Institute of Industrial Science, The University of Tokyo. Firstly, daily AVHRR data were radiometrically and geometrically corrected, and data improve- ment procedures were implemented including sensor degradation, sensor change. Secondly, 10-day composite images were created and these images were converted to NDVI. However, constructed time-series NDVI data was contaminated with remained cloud or poor atmospheric condition. In order to overcome this problem, a noise reduction algorithm was applied to the data. Then, green- up date of each pixel was computed by using characteristics of annual profile of NDVI. The date was defined intersecting points between annual mean NDVI and variations in NDVI, and detected green-up date were consistent with ground observation data. In time-series analysis of green-up date, it was found that green-up in mixed forest tend to be getting earlier at a rate of -0.58 days/year, and the tendency was strong especially in northern part of study area. And as a result of sensitivity analysis between green-up date and meteorological data (temperature, precipitation, cloud cover), the area where green-up had a tendency to getting earlier showed a higher sensitivity to temperature rise.
B21A-0029
A Volume Model for Urban Heat Island Based on Remote Sensing Imagery and Its Application: A Case Study in Beijing
Along with urbanization, urban heat island (UHI) has become one of the most serious urban problems, because of its impacts on the urban microclimate, air quality, energy consuming, public health and so on. With the advent of thermal remote sensing technology, remote observations of UHIs become the focus of urban remote sensing. While some progresses have been made by scientists, remote sensing study of UHI has been slow to advance qualitative description of thermal patterns and simple correlations. As a common indicator of UHI, magnitude is often used to describe the degree of UHI occurrence. In order to develop a more quantitative and more effective indicator for UHI dynamic monitoring at regional scale, deriving its spatial feature and facilitating the comparisons between different UHIs occur at different time or different areas based on remote sensing imageries, this paper proposes a new parameter named UHI volume to integrate UHI magnitude and extent effectively. After the subtraction of the rural contribution in the land surface temperature (LST) image, the isolated UHI signature is fitted using a least-squares Gaussian surface and then the UHI volume is calculated as a double integral of the UHI signature function over its footprint. This study investigates the applicability of this volume model based on examination of thirty EOS-Terra MODIS level 1B imageries of Beijing City, China, acquired between 2004~2006. These imageries include fifteen nighttime scenes, with their corresponding daytime scenes. Firstly, the resampled digital elevation model (DEM) data, the normalized difference vegetation index (NDVI) and the modified normalized difference water index (MNDWI) are used to extract the urban areas. Secondly, a simplified method is performed to retrieve the land surface temperature from MODIS channel 31 and channel 32. Thirdly, four transects at different directions including N-S, W-E, NW-SE and NE-SW are selected to detect the UHIs to ensure the application of the volume model is valid. The detections reveal that each of the UHI in Beijing City at different time has a single core and its spatial distribution is symmetrical as a whole, despite there are some small UHIs in the outskirts, so the volume model presented is appropriate for the UHI simulations. Results of the UHI simulations demonstrate that: (1)The correlation between the modeled UHI signatures and the corresponding true values is high, especially for the nocturnal UHIs. This fact indicates that the Gaussian fit and the volume model is valid for UHI simulation. (2)There is obvious UHI effect in both daytime and nighttime in the summer. In the spring, autumn and winter, there is UHI effect in the nighttime, while no UHI in the daytime; the UHI magnitude and volume shows that the diurnal UHI is intenser than the nocturnal UHI, and the difference of the volume between the diurnal and nocturnal UHIs seems to be stable. (3)Because of different dominant factors including natural and anthropogenic factors and their different influences, the changes of the nocturnal UHIs are complicated, especially the extents of UHIs. The UHI magnitude and volume shows that the nocturnal UHI effect is intensest in the winter, while weakest in the summer.
B21A-0030
A Constrained Inverse and Forward Canopy Reflectance Modeling System for Mapping key Biophysical Properties Using Remotely Sensed Reflectance Observations
Accurate quantitative estimates of leaf chlorophyll content (Cab), which can assist in determining vegetation stress and photosynthetic productivity, and leaf area index (LAI) are important for optimizing estimates of regional scale energy and carbon exchange. Here refinements to the biophysical parameter retrieval system presented in Houborg & Boegh (2007, doi:10.1016/j.rse.2007.04.012) are discussed and the performance of the modified model is demonstrated using high-resolution aircraft and SPOT satellite data. Since a unique relationship between a single biophysical canopy variable and a spectral signature does not exist, a canopy reflectance model (CRM) was employed in inverse and forward mode to build multiple crop and site dependent formulations relating LAI and Cab to various spectral reflectance signatures. Leaf inclination angle, Markov clumping characteristics and leaf mesophyll structure were assumed spatially and temporally invariant within the field boundaries of each agricultural land cover class and estimated by iteratively inverting the CRM using multiple intra-field green (green), red and near-infrared (nir) reflectance observations. Only pixels originating from medium to high density vegetation areas were included to maximize the sensitivity of the reflectance signal to the crop specific canopy parameters. New techniques were implemented to constrain the parameter space to relatively few plausible values reducing the number of the computationally demanding inversions, and to simultaneously consider the fraction of dead leaves in the canopy. The inversion was further constrained by separating the retrieval of the soil background reflectance signal from the retrieval of the canopy parameters. Finally, a family of model generated spectral reflectance relationships, each a function of soil and canopy characteristics, was employed for a fast pixel-wise mapping of LAI and Cab. The application of LAI-NDVI, LAI-nir, and Cab-green relationships provided reliable quantitative estimates of LAI and Cab for green as well as partially senescent agricultural fields in Maryland, U.S.A. characterized by contrasting architectures and leaf biochemical constituents. The model was able to detect decreases in leaf chlorophyll content caused by stressed environmental conditions. The biophysical parameter retrieval system is completely image-based, does not require a priori ground based information, is fast enough for regional-scale applications, and can quite easily be implemented for other regions.
B21A-0031
Delineating Surface Water Features In The Prairie Pothole Wetlands With MODIS Data
The Prairie Pothole Region (PPR), covered with thousands of shallow ponds known as potholes, is a large wetland area in central North America. The PPR wetlands are valuable because of the ecosystem services they provide, including water balance and flood mediation, habitat for migratory birds and waterfowl production, and carbon sequestration. During the last century, impacts of intense land use (agricultural and commercial development) and climate change have caused drastic reductions of surface water area and wetland habitat in the PPR. Spatial and temporal characterizations of the surface water dynamics are important for understanding the hydrological and ecological characteristics in the PPR wetlands. In this study, we use the Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance data processed as 8-day composites at 500-meter resolution from 2000 to 2006. The modified normalized difference water index [MNDWI = (band 4 ˇ§C band 6)/(band 4 + band 6)] was computed to delineate water features. By adjusting the MNDWI threshold, we generated three land-cover classes: water, land, and water/land mixture. Additionally, we acquired 26 Landsat TM/ETM+ scenes, 14 to calibrate the MODIS MDNWI threshold and 12 to independently validate the MODIS water feature products. For the validation, the maps of percent water area derived from MODIS were compared with the maps generated from the Landsat images at a 5-kilometer grid level. The comparison illustrates a low root mean square error (0.0034 percent) and a high correlation coefficient (0.95), suggesting satisfactory accuracy of the MODIS water products. We have produced and illustrated the MODIS water feature maps for the U.S. portion of the PPR at 32-day intervals for the growing seasons from 2000 to 2006. The time series of MODIS-derived water feature maps are useful for delineating water features and monitoring water area dynamics in this area, and these products provide coarse resolution information to support conservation and management across large regions of the PPR wetlands. The development of the technique is also of value in the analysis of water dynamics for different wetland areas, such as the Yukon River Basin in Alaska.
B21A-0032
ORNL DAAC MODIS Land Product Subsets
MODIS sensor data are highly useful for field research. However, the volume of MODIS data and the complexity in data format makes MODIS data less usable in some cases. To solve this usability issue, the Oak Ridge National Laboratory Distributed Active Archive (ORNL DAAC) prepares and distributes subsets of selected MODIS (Moderate Resolution Imaging Spectroradiometer) Land Products in a scale and format useful for field researchers. MODIS Collection 5 subsets are provided for more than 1000 sites across the globe. The subsets are offered in tabular ASCII format and in GIS compatible GeoTIFF format. The ASCII subsets are 7x7 km and the GeoTIFF subsets are 25x25 km centered on the field site. ORNL DAAC also provides time series plots and grid visualizations for the MODIS land products to help characterize field sites. User-defined quality conditions can be applied to the ASCII subsets to filter data based on user's requirements. In addition to offering subsets for fixed sites, the ORNL DAAC also offers the capability to create user-defined subsets for any location worldwide. The MODIS Global subsetting tool provides subsets from a single pixel up to 201x201 km for user-defined time range. Statistics, time series plots and GIS compatible files for the customized subsets are also distributed through this tool. Users place an order for a MODIS subset online and an email is generated when the subset is created. Google maps, Google Earth and Landsat images are provided to help visualize the subset location. http://daac.ornl.gov/MODIS/modis.html
B21A-0033
Comparison of Normalized Burn Ratio, Normalized Difference Vegetation Index, and Enhanced Vegetation Index in Areas Burned by the Jasper Wildfire of Black Hills South Dakota
The Jasper wildfire of August and September 2000 was the largest fire to occur in the Black Hills in at least a century. The disturbance on ecosystem characteristics will be widespread and long-term. Monitoring postfire vegetation changes using remote sensing data can provide unique and timely information about ecosystem dynamics. In this study, the Normalized Burn Ratio (NBR), Normalized Difference Vegetation Index (NDVI), and Enhanced Vegetation Index (EVI) data were derived from Landsat imagery and compared before and after the Jasper fire. Landsat 5 images acquired on June 2, 2000 (preburn), and June 5, 2001 (10 months postburn), were analyzed. In addition, a Landsat 7 image acquired on May 31, 2002 (22 months postburn), was used in the study. Landsat data were converted to at-sensor reflectance, and NBR, NDVI, and EVI values were calculated for low, moderate, and high burn severity areas defined by using the difference of NBR between 2001 and 2000. NBR values in areas characterized as low burn severity changed very little between 2001 and 2002. Meanwhile, areas characterized as moderate or high severity showed substantial increases in NBR values between 2001 and 2002, implying some ecosystem recovery occurring for these areas over a relatively short time. EVI and NDVI show similar patterns of change, but it was found that EVI and NBR indices are more sensitive than is NDVI for capturing vegetation cover changes during the early postfire years. Further research is planned to use Landsat and MODIS imagery to assess spectral trends as a function of time in areas affected by fire.
B21A-0034
MODIS Collection 5 Land Cover Type and Land Cover Dynamics: Algorithm Refinements and Early Assessment
The Moderate Resolution Imaging Spectroradiometer (MODIS) onboard NASA's Terra and Aqua spacecraft provides land surface data at global scales that are useful for a wide array of scientific applications related to land surface properties and processes. In the past year, land products from collection 5 MODIS reprocessing have become available. In this paper, we describe algorithm refinements and recent results from collection 5 reprocessing of the MODIS land cover type and land cover dynamics products (MOD12). Specifically, we describe the algorithms and data sets that are being used to characterize the geographic distribution and phenology of vegetation and land cover types at global scales. In collection 5, the MOD12 product is being produced at 500- meter spatial resolution using 8-day inputs from the MODIS Nadir-BRDF-Adjusted Reflectance (NBAR) product. The increased spatial and temporal resolution of the input data used to produce the MOD12 product represent significant steps forward and result in substantial improvements relative to the MOD12 collection 4 products. This paper will describe specific changes to algorithms and input data that are being used in collection 5, and will provide preliminary assessments regarding changes in product quality.
B21A-0035
Object-Based Land Use Classification using Airborne LiDAR
Better information on roughness of various types of vegetation is needed for use in resistance equations and eventually in flood modelling. These types include woody riparian species with different structural characteristics. Remote Sensing information such as 3D point cloud data from LiDAR can be used as a tool for extracting simple roughness information relevant for the condition of below canopy flow, as well as roughness relevant for more complex tree morphology that affects the flow when it enters the canopy levels. A strategy for extracting roughness parameters from remote sensing techniques is to use a data fusion object classification model. This means that multiple datasets such as LiDAR, digital aerial photography, ground data and satellite data can be combined to produce roughness parameters estimated for different vegetative patches, which can subsequently be mapped spatially using a classification methodology. Airborne LiDAR is used in this study in order to classify forest and ground types quickly and efficiently without the need for manipulating multispectral image files. LiDAR has the advantage of being able to create elevation surfaces that are in 3D, while also having information on LiDAR intensity values, thus it is a spatial and spectral segmentation tool. This classification method also uses point distribution frequency criteria to differentiate between land cover types. The classification of three meanders of the Garonne and Allier rivers in France has demonstrated overall classification accuracies of 95%. Five types of riparian forest were classified with accuracies between 66-98%. These forest types included planted and natural forest stands of different ages. Classifications of short vegetation and bare earth also produced high accuracies averaging above 90%.
B21A-0036
Application of MODIS Normalized Differential Vegetation Index for Local Land Use Indicators of Impervious Surface Areas
Data derived from satellite measurements offer tremendous potential to contribute to environmental indicators broadly, and land cover/use indicators specifically, given satellite data's consistent, repetitive nature with broad spatial and temporal coverage. This study focuses on the translation of satellite data into meaningful measures that fit within the frame of environmental indicators used by policymakers, resources managers, and the general public. The study area encompasses the Greater Cincinnati Area (6,898 km2), a mid-sized city seeking to address sustainability in the context of urban change and growth. At the center of the study area is downtown Cincinnati in Hamilton County, with suburban areas extending into parts of the seven surrounding counties. Hamilton County staff are particularly interested in tracking impervious surface areas (ISAs) and forest cover as both have been shown to impact nearby water quality. The purpose of this study is to provide Hamilton County and other planning organizations with near real-time information on ISAs and forest cover through a simple, inexpensive methodology that leverages publicly available satellite data products. We obtained 250m resolution Normalized Differential Vegetation Indices (NDVI) data files derived from NASA MODIS (MOD13Q1) for 16-day periods in June/July of 2001-2006. NDVI is calculated based on transformations of the red (620-670 nm), near- infrared (841-876 nm), and blue (459-479 nm) bands designed to enhance the vegetation signal and allow for comparison in terrestrial photosynthetic activity. We examined multiple thresholds of NDVI to act as a surrogate for ISAs (low NDVI) and forest cover (high NDVI). We also calculated changes in NDVI throughout the study period and correlated large decreases in NDVI to known large developments. While this method has potential, further study is needed to ground truth the results, a process that is currently underway. In addition, calculation of NDVI with higher-resolution LANDSAT or ASTER data may improve the results.
B21A-0037
Investigation of uncertainty in surface reflectance measures derived from AVHRR caused by orbital drift and variations in viewing geometry.
Data derived from the AVHRR (Advanced Very High Resolution Radiometer) series of instruments has been extensively used for monitoring global and regional vegetation dynamics for last 25+ years. Past studies have shown that the influence of systematic change in satellite overpass time during the course of satellite's life due to orbital drift remains, even after sensor calibration, owing to change in sun-sensor geometry. In order to assess the impacts of orbital drift on time-series analysis of derived metrics like vegetation indices and phenology. We investigated simulated AVHRR time series data created with actual AVHRR viewing and illumination geometry. To examine orbital drift alone interannual variations in phenology and atmospheric composition was not used in the simulations. A three dimensional canopy radiative transfer model (FLIGHT) was coupled with an atmospheric radiative transfer model (6s) to create simulated time series data for three AERONET sites to represent boreal forest, tropical forest, and semi-arid landcover. This was further used to compare various methods used to deal with impacts of orbital drift like EMD (Empirical Mode Decomposition) and semi empirical BRDF correction.
B21A-0038
Classification of forested wetlands using ordination of multitemporal Landsat reflectance
There are important limitations in interpreting satellite imagery in dynamic environments. In forested wetlands, where flooding and aquatic vegetation vary temporally, a multitemporal approach is needed for extracting stable patterns. We used field measurements of forest composition and structure and seven cloud-free Landsat images from a time span of six years to classify forest vigor in 95,000 ha of cypress-tupelo forested wetlands surrounding Lake Verret in southern Louisiana. A principal component (PCA) ordination of the reflectance in Landsat bands 3/4/5 for each image was the basis of the classification. No single Landsat band or image dominated the first few PCs, so that the multitemporal and multispectral aspect of the data were fully expressed in the ordination. Each Landsat pixel was classified as to forest vigor according to its scores in the first two PCs by comparing pixel scores to those associated with field plots. The reflectance PCA, and thus the classification, was directly interpretable in terms of ecosystem structure because the scored in the first two PCs in field-plot pixels were correlated to field measurements of forest structure, such as leaf area index, stand density, and understory composition, and because the field plots occupied interpretable regions of ordination space.
B21A-0039
Derivation of Clumping Index via Using BRDF Models and MISR and MODIS Data
Clumping index quantified the level of foliage grouping within distinct canopy structures relative to a random distribution. Vegetation foliage clumping significantly alters its radiation environment and therefore affects vegetation growth as well as water and carbon cycles. The clumping index is useful in ecological and meteorological models because it provides new structural information in addition to the effective Leaf Area Index retrieved from mono-angle remote sensing. Multi-angle sensor, for example Multi-angle Imaging SpectroRadiometer (MISR) and Moderate Resolution Imaging Spectroradiometer (MODIS), observations provide a means to characterize the anisotropy of surface reflectance, which has been shown to contain information on the structure of vegetated surfaces. This study shows that available multi-angle data products, MISR RPV Bi-directional Reflectance Distribution Function (BRDF) model parameters and MOIDS kernel-driven BRDF model parameters, can be used to computing the clumping index. First of all, the hotspot and darkspot reflectance were calculated by using MISR RPV and MOIDS kernel-driven BRDF models. Then the clumping index was obtained by applying existed relationship between the clumping index and an index derived hotspot and darkspot. The preliminary findings on estimating the clumping index are: 1) red band is better than near infrared band. 2) The hotspot and darkspot from smaller solar zenith angle is better than ones from larger solar zenith angle. 2) The kernel-driven model and MODIS BRDF parameters product is superior that the RPV model and MISR BRDF parameters product. In order to improve existed approaches of deriving the clumping from multi-angle measurements, the underlying mechanism of foliage structure determining the clumping index have been explored using a geometrical optical and radiative transfer model named GORT. This presentation will display the new methods of computing the clumping index from multi-angle measurements, and evaluation of performance on estimating the clumping index the RPV model and the kernel-driven model.
B21A-0040
Seasonally Varying Leaf Area for Climate and Carbon Models from Assimilation of Satellite Reflectance data into a Dynamical Leaf Model
Leaf area index is an important land surface parameter and is a necessary input for climate and carbon models. The widely-used leaf area products derived from satellite observed surface reflectances contain substantial erratic fluctuations in time due to incomplete atmospheric corrections and observational and retrieval uncertainties, which are inconsistent with the seasonal dynamics of leaf area that are generally gradual in nature. We propose a data assimilation approach to combine the satellite observations with a dynamical leaf model so that the seasonal cycle of the directly retrieved leaf areas can be also constrained by the dynamical model simulations. The data assimilation allows automatic adjustment of the dynamical model parameters, such that the optimal compromise between the estimated surface reflectances based on the modeled leaf area and that of satellite observations can be reached. Testing results at three United State deciduous forests with relatively homogenous landscapes have shown that the data assimilation significantly smoothens the seasonal cycle of the estimated leaf areas compared to that without the dynamical leaf model constraining. Meanwhile, it does not deteriorate the agreement between the modeled and satellite observed surface reflectances.
B21A-0041
Trends of Vegetation Greenness in the Arctic from 1982-2005
The Arctic region has experienced a continuous trend of warming during the past 30 years. Meanwhile, many areas of the Arctic are undergoing large-scale industrial development, e.g. oil and gas exploration, at a rapid pace, indicating an increasing human pressure and land use changes even in this frontier wilderness. Major questions face arctic terrestrial ecologists are what will happen to the tundra ecosystems as the global climate warms and what will happen to the indigenous people way of life as land cover changes proceed? Here, we combine multi-scale sub-pixel analysis and remote sensing time-series analysis to investigate recent decadal changes in vegetation photosynthetic activity along spatial gradients of summer temperature and vegetation in the Arctic. The datasets used here are NASA Gimms data at 8 km pixel resolution and MODIS land cover data. Fractional vegetation cover was analyzed in order to select homogenously vegetated areas of tundra and autoregression analysis was performed on time series of those homogenous pixels. Only pixels below 70 degree north were included for 2004-2005 due to calibration errors occurred beyond 70 degree north for those years. Linear trends in Arctic tundra vegetation greenness over period 1982-2005 were positive. However, there were different magnitudes between Eurasia and North America. The rate of change was +0.64%/yr over North American Arctic compared to +0.44%/yr over Eurasian Arctic. Vegetation productivities increase from north to south along bioclimatic gradient, therefore, greenness is much higher in areas below 70 degree north compared to entire tundra biome. Higher rates of greening in High Arctic contributed to a stronger positive trend in the longer time series. The rate of greening detected here was higher than that reported in previous studies. This is likely due to two reasons: 1) we restricted our study area in tundra biome only with a phenological tundra-taiga boundary identification approach, therefore, have less chance to mix information of boreal forest in the south; 2) we applied homogenous vegetation approach to avoid noise from lakes and bare ground and more likely detect initial changes over tundra vegetation.
B21A-0042
A Practical Method for Retrieving Land Surface Temperature from AMSR-E over the Amazon Forest
Remote sensing of land surface temperature (LST) using infrared sensors, such as the Moderate Resolution Imaging Spectroradiometer (MODIS), is only capable of retrieval under clear-sky conditions. Such LST observations over tropical forests are very limited due to clouds and rainfall, especially during the wet season, and high atmospheric water vapor content. In comparison, low frequency microwave radiances are minimally influenced by meteorological conditions. Exploring this advantage, we have developed an algorithm to retrieve LST over the Amazonian forest. The algorithm uses multi-frequency polarized microwave brightness temperatures from the Advanced Microwave Scanning Radiometer (AMSR-E) on NASA's Earth Observing System. Relationships between polarization ratio and surface emissivity are established for forested and non-forested areas, such that LST can be calculated solely from microwave radiance. Results are presented over three time scales: at each orbit, daily, and monthly. Results are evaluated by comparing with available air temperature records on daily and monthly intervals. Our findings indicate that the AMSR-E derived LST agrees well with in situ measurements. Results during the wet season over the tropical forest suggest that AMSR-E LST is robust under all-weather conditions and shows higher correlation to meteorological data (r=0.70) than infrared based LST approaches (r=0.42).
B21A-0043
Multiangular and Hyperspectral Forest Reflectance Modeling in the Hemiboreal Zone: A Case Study With CHRIS PROBA Data
The hemiboreal and boreal forests of the northern hemisphere form the largest unbroken forest zone in the world and play a key role in global climate studies. Obtaining information on the state of forests over large regions at a weekly or monthly schedule is possible only by remote sensing. A prerequisite for the development of image interpretation methods for such large scale remote sensing applications is an understanding of the physical principles behind the spectral signal measured by satellite or airborne instruments. Efficient use of forest canopy reflectance models which adopt the physical principles of remote sensing of forests has been hindered by the lack of hyperspectral and multiangular data. Recently, several missions, e.g. CHRIS PROBA, providing such imagery have been launched. In this study, we examine and identify links between canopy structure and multiangular, hyperspectral reflectance properties of typical deciduous and coniferous hemiboreal forests using CHRIS PROBA data and a recently improved forest reflectance model PARAS. The PARAS model is based on the concept of photon recollision probability, and thus provides a powerful, yet simple, method for modeling the spectral behavior of forest canopy scattering. Empirical ground reference data for the study were collected from the Jarvselja Training and Experimental Forestry District in southeastern Estonia, and model simulation results were validated using three simultaneous CHRIS PROBA mode 3 scenes over the study site.
B21A-0044
Derivation of Time Series Vegetation Map from MODIS over Taiwan
The monitoring of land-cover and its greenness is very important for modern environmental issue. Satellite remote sensing can serve as a very suitable tool for gathering such relevant temporal information for various analyses. However, there are many factors, such as atmospheric (cloud, aerosols etc) effects, view and sun angle's effects and geometric distortions, should be considered during the task. This study is to carry out the derivation of the time series vegetation index from satellite data with considering above mentioned effects. In the study, the characteristic of the different wavelength solar radiation interacted with the different surface are analyzed, and find out the suitable cloud masking threshold based on the visible and near infrared brightness temperature/reflectance for Taiwan region. Also the viewing and sun angle effects are analyzed for each 3x3 pixel window and try to find out their correlation relationship between vegetation index and angles. Finally the vegetation maps for a given time period can be derived with MODIS.
B21A-0045
Deforestation and pasture establishment in watersheds: Implications for stream biogeochemistry in the Amazon Basin
The ecological response of landscapes to land cover change depends on the rate and extent of conversion. Rapid clearing can cause transient pulses of nutrients to streams, while slow but extensive change can result in chronic changes. The measurement of the rate and extent of change also need to be made in ecologically relevant landscape units. A time-series of Landsat TM imagery are combined with a large population of watersheds (N=4788) to quantify the rate and extent of deforestation in a region of the Amazon Basin. The time series shows that clearing was slow (5-6 percent per year during active clearing) but proceeded to more than 75 percent of the watershed area for more than 25 percent of all watersheds draining less than 10 km2. Geostatistical methods suggest that the extent of deforestation is highly autocorrelated, while the rate is a random variable. The autocorrelation causes even large watersheds to be heavily deforested. A small fraction of the cleared area reverted to forest, suggesting that clearing was both extensive and permanent. Clearing of forest followed by pasture establishment in a wide range of watershed sizes is likely to lead to chronic changes in stream biogeochemistry, rather than transient pulses from cutting and burning of forest biomass.
B21A-0046
An Automated Multi-temporal Image Co-registration Technique for Lake-rich Environments Based on Pseudo Invariant Features
The increasing availability of remotely sensed data has facilitated an efficient and in-depth study of lake dynamics in the context of global change. Regional-scale lake monitoring requires effective change detection using multi- temporal and multi-sensor satellite imagery. Change detection involves a pixel-by-pixel comparison of multi- temporal images, necessitating precise image co-registration. The performance of widely-used co-registration techniques that rely upon manual selection of tie points is limited by the accuracy of individual tie points identified from the image pair to be co-registered. When the number of image pairs to be co-registered is large, these interactive methods however become prohibitive due to the significant volume of data. Therefore, the use of automated image co-registration techniques is required. The performance of area-based co-registration techniques is limited by factors such as atmospheric degradations, illumination effects, and sensor sensitivity variations in multi-temporal images, whereas feature- based techniques are robust to the above limiting factors. Feature based co-registration techniques establish correspondence between the features derived from the respective images, eliminating the need for radiometric normalization. However, the performance of such techniques is limited in lake-rich areas such as Arctic, where lakes are the dominant surface features and change over time. To overcome this limitation, we introduce the concept of Pseudo Invariant Features (PIFs) based on feature shape criteria to identify stable lakes - lakes that have not undergone significant change in shape over the time, and propose an approach that employs their center points in developing the co-registration model. Performance of the proposed approach is evaluated quantitatively, and sub-pixel co-registration accuracy is achieved.
B21A-0047
A New Approach to Testing the Fossorial Rodent Hypothesis of Mima Mound Formation Using Airborne-Based LIDAR and a Diffusive Sediment Transport Model
Mima mounds are nearly circular soil mounds, found in grassland landscapes. In California, Mima mounds are often associated with vernal pools, seasonal wetlands that harbor rare and endemic plants and animals. The processes that form and maintain the mound-pool complexes have not yet been conclusively identified, even though such information is necessary to understand the effects that land use and climate change may have on the resilience and longevity of these landscapes. One hypothesis for the origin and persistence of Mima mound- vernal pool systems (termed the Fossorial Rodent Hypothesis) proposes that burrowing organisms such as pocket gophers (Rodentia: Geomyidae) maintain and possibly create the mounds by preferentially translocating soils towards mound centers as an adaptive response to high water tables. In order to investigate this hypothesis, the topographic characteristics and aboveground gopher activity of one of the largest remaining Mima mound-vernal pool systems in California were studied. Detailed topographic information for the mound-pool systems was obtained via an airborne-based LIDAR (Light Detection and Ranging) survey of a 25km2 region near Merced, CA. An object-oriented classification scheme, which combined different scale, shape, and spectral parameters, was employed in order to characterize the mounds. Based on the initial classification results, roughly 275,000 mounds were identified, indicating a mound density of 11,000km-2. Within the larger study area, gopher sediment transport was monitored on a 3507m2 site by conducting periodic surveys of sediment mounds created by gopher activity using a Global Positioning System and mass measurements. Downslope erosion rates (off Mima mounds) were estimated using a mass balance model which incorporates a diffusive sediment transport law. The median calculated net downslope erosion rate was 15 cm of soil per 1000 years, while the measured rate of aboveground gopher sediment movement was approximately 57cm of soil per 1000 years. Assuming that some portion of the gopher sediment movement is in an upslope direction, these results suggest that gopher soil transport may be large enough to compensate for erosion, and this activity may play a dominant role in maintaining Mima mound-vernal pool systems. The results of this study were used to guide the development of a quantitative model of gopher-driven sediment transport on Mima mound-vernal pool landscapes. This model will be used to infer the origin of the landscapes and the rates of processes critical to their continued functioning and will also help to determine, quantitatively, the role of burrowing animals as a "keystone" species on these landscapes.
B21A-0048
Interpreting the Spatial Heterogeneity of Lake Drying in the Minchumina Basin, Central Alaska
The Minchumina Basin is a large wetland complex located in the northwestern corner of Denali National Park and Preserve, AK. The wetland is underlain by a mosaic of discontinuous and continuous permafrost and is the site of the Central Alaska Network Shallow Lake Monitoring program. The objective of the program is to understand the hydrologic dynamics of this diverse system. Understanding the physical structure and temporal variation in lake level provide critical information to interpreting the spatial heterogeneity in lake drying found throughout the boreal forest. Lake surface area estimates were made using RADARSAT imagery collected monthly during the 2006 growing season. We also use RADARSTAT to identify wetlands within the park that exhibit varying degrees of connectivity to streams or rivers. Using ground monitoring techniques we try to relate bathymetry, water quality, soil type and distribution of discontinuous and continuous permafrost to patterns of connectivity and lake drying. On the ground field observations and ground-truthing demonstrate a variety of limitations in the interpretation of remote sensed data used for estimating lake area. Particularly noteworthy is the measurable decline in water level on what appear to be stable lake ecosystems.
B21A-0049
Multi-temporal MODIS-Landsat Data Fusion for Relative Radiometric Normalization and gap Filling of Landsat Data
The primary limitation to the utility of Landsat data, other than data cost, is the availability of cloud-free surface observations. As currently only degraded Landsat 5 and Landsat ETM+ systems are acquiring data, and only a single successor Landsat Data Continuity Mission sensor is scheduled, a potential alternative to provide more surface observations is the fusion of Landsat data with data from other remote sensing systems. A semi-physical fusion approach that uses the MODIS BRDF/Albedo surface anisotropy characterization product and Landsat ETM+ data to predict 30m Landsat reflectance on any date is presented. The methodology may be used for ETM+ cloud and SLC-off gap filling and for relative radiometric normalization. It does not require any tuning parameters and so may be automated, it is applied on a per-pixel basis and is unaffected by the presence of missing or contaminated neighboring Landsat pixels. The methodology and spatially explicit results are presented for two Landsat acquisitions at three Landsat scenes, one in Africa and two in the U.S., selected to encompass a range of land cover land use types, and temporal variations in solar illumination, land cover, and phenology. Summary statistics of the difference between the predicted and observed ETM+ reflectance (prediction residual) are compared with the difference between the ETM+ reflectance observed on the two dates (temporal residual) and with respect to the MODIS BRDF model parameter quality. For all three Landsat scenes, and for all bands, except one short wavelength band, the mean prediction residual is smaller than the mean temporal residual, typically by a factor two. This fusion methodology may be applied to any high spatial resolution satellite data with similar spectral bands as MODIS and where the sensor viewing and solar illumination geometry can be accurately derived.