October 12 - 1:00 - 5:00
October 13 - 9:00 - 6:00
Boston University
414 in the School of Management Building
595 Commonwealth Avenue
BACKGROUND Following Alan Strahler's welcome, Jeff Privette opened the meeting by reviewing milestones in EOS LAI/FPAR/NPP validation planning. He noted several formal workshops had been held (see EOS Earth Observer, Vol. 8, No 2, and 10, 3) beginning in 1996. The last workshop had been a meeting of the EOS instrument teams with the newly funded AM-1 Validation Investigators in Dec. 1997. In that meeting, Alfredo Huete and Betty Walter-Shea co-chaired a breakout session on validation of biophysical properties. This session revealed the absence of LAI/fPAR collection/validation protocols and hence the need for the present meeting.
Punctuating these workshops were two intra-EOS team Prototype Validation Exercises (PROVEs), in which ground data were collected, upscaled, and compared to satellite data/products. Privette noted that the PROVEs coupled EOS instrument teams to experts in the wider scientific community as well as in other instrument teams, a result aimed to promote more efficiency and collaboration post-launch.
Privette challenged meeting participants to work over the two days to uncover potential flaws in the evolving program, as global LAI/FPAR/NPP validation has not previously been attempted, and thus conclusions from this workshop will form a blueprint for the overall scheme. Moreover, meeting recommendations may ultimately be used in part to defend criticism. He stressed that while MODLAND is charged with developing a global validation plan, the resulting plan should apply to similar land products from other AM-1 instruments such as MISR and ASTER, as well as Landsat 7.
Thus, specific goals for the meeting of the meeting were: 1) to define criteria for "validation" of global LAI, fPAR and NPP products from EOS 2) review and assess site-level measurement and scaling techniques 3) determine acceptable protocols/priority for delivery and distribution of validation results 4) define community needs from instruments and EOS validation investigator teams
Jeff Morisette then reviewed the MODIS Validation Plan. The Plan was recently updated (Version 3, October, 1998) to include the EOS Land Validation Core Sites and MODLAND Product Sites, the potential contribution of the AM-1 Validation Investigators, the BigFoot and SAFARI 2000 activities, and other items. It can be downloaded from the redesigned MODLAND validation page at http://modarch.gsfc.nasa.gov/MODIS/LAND/VAL.
Alfredo Huete then presented a summary of the Biophysical/Vegetation validation break-out session from the December, 1997, SWAMP meeting. A variety of issues were raised without resolution in that session. For instance, would FIPAR be acceptable in place of FPAR, at what solar zenith angles should it be measured, and how can it be properly scaled? Huete suggested that a standardized sampling design be developed to help guide field personnel. He noted the data needs of validation data users vary inversely with the speed at which they must have the data. Specifically, Huete suggested prioritization be in order of: product validation, algorithm validation, interdisciplinary science, and auxiliary science. However, the in-situ data needs of end-users were not well known. Finally, Huete reminded participants that calibration issues needed attention. Some possibilities outlined at the Dec. workshop included single-point calibration labs (Thome, U. of Arizona?) and mobile calibration transfer instruments.
NPP PRODUCT ALGORITHM AND VALIDATION
Rama Nemani then presented the algorithm for the MODIS NPP product. The inputs for this product will be the land cover, LAI/fPAR products and DAO data. Ingesting the DAO data is a major processing load, and there was some discussion about using actual data for input for annual NPP estimates. Particular problems raised included partitioning aboveground vs. below ground biomass, quantifying soil respiration, and validating an NPP product from the NEE measurements of flux towers. Furthermore, the DAO outputs are not being validated, and at 2 deg. resolution, may be impossible to validate, although some comparison with measured data may be useful.
Initially the NPP product will be validated against field measurements at the four BigFoot/MODLERS sites as well as spatio-temporal patterns of NPP obtained by various studies as a part of VEMAP, GAIM/IGBP, EUROFLUX and AmeriFlux. It was noted that general product accuracy may be quite different than site product accuracy. Bill Emanuel pointed out that a model inter comparison could be developed to start a data sharing initiative among NPP modelers/users.
LAI/fPAR ALGORITHM AND VALIDATION
Ranga Myneni presented the algorithm for the MODIS/MISR LAI/fPAR products. He noted that the definition of LAI was the one sided green leaf area over a unit area of ground for all biomes except needleleaf, in which case leaf area is defined as projected leaf area. Further, he noted the 8-day LAI product as composited based on the maximum FPAR measurement.
The input for this process includes the MODIS land cover (LC, broken down into six biomes) and surface reflectance products. There are still questions about how the land cover biome will be determined since the MODIS LC classification does not map directly into the biome classification [a particular problem discussed in the subsequent LC meeting was the need for discrimination between cereal crops and broadleafed crops, and the current neglect of that division by MODIS landcover algorithm developers]. Myneni noted the biome discretization itself needs validation.
Other issues identified included mixed biome classes at the 1 km resolution of the product (the algorithm currently assumes single biome in each cell), and the current averaging of available leaf spectra to fix leaf optical properties in the model. Emanuel noted that in some biome mixture cases, a forest cell may be predominately broadleaf deciduous during summer months but evergreen during winter months, effectively a change in biome type. The product algorithm currently does not accommodate such changes. Myneni noted improvements will be possible upon validation data collection, particularly in the assigning of fixed variable values.
Myneni also noted recent improvements had allowed mapping of errors in the input data fields to a probabilistic approach in algorithm results. Thus, the "answer" is an expected value plus a domain of possible values. Theoretical accuracy varied as a function of the spectral bands available for input. For example, accuracy decreased when blue wavelengths were used since this band has more noise. Simulations suggested the overall theoretical accuracy of the product is 20% (relative). He noted that testing with SeaWiFS data showed the majority of difficulties are associated with pixels with NDVI<0.1, i.e., non-vegetated cases. He noted in these cases that if the soil reflectance is correctly retrieved, this provides good validation of the algorithm but not the LAI product itself.
Myneni gave a list of minimum requirements for validation field data, which included spectral canopy reflectances, leaf and soil properties. Ideally, field data would also include canopy structure. Exact prioritization of these variables became a topic for a breakout session.
Myneni then outlined five validation "tests" for the LAI/fPAR algorithm (descriptions available in the handout Myneni provided), including: 1 - Biome by Biome 2 - Soil Test 3 - Red-NIR Density Plots 4 - NDVI-LAI and FPAR-NDVI relationships 5 - Information conveyed about LAI by HDRF
He noted that validation of the algorithm at a point scale would help validate the product at larger scales. This important point suggests that significant natural variability at a single validation site would add more statistical significance to the overall validation of the product, despite the use of relatively few sites globally. Huete asked about the validation algorithm as it relates to the field data and the algorithm input data. Myneni described it as a triad of values to be compared: MODIS reflectance and land cover, field measured data, and the algorithm output. Privette noted that some cross-algorithm validation could occur since BRDF/albedo, LAI/fPAR and VI all use different but comparable reflectance models.
CURRENT LAI RELATED RESEARCH
Leonard Brown, of the Environmental Monitoring Section of CCRS, outlined "LAI Map of Canada", a validation procedure utilizing LAI-2000 and TRAC instruments, mid-summer TM imagery, and scaling to AVHRR to compare an LAI map with validation data. Landcover-dependent SR-LAI relationships developed through BOREAS are used to estimate maximum LAI. The canopy gap size distribution is estimated from TRAC. This leads to canopy gap estimates, from which LAI is estimated. They have found collaborators eager to participate and Brown discussed the effort involved with collecting data from the various collaborators/sources. He suggested that an agreement could develop between CCRS and EOS validation for accessing the Canadian data sets. He noted that while investigators generally were willing to collect the data on a good will basis (with coauthorships as a carrot), the coordination and troubleshooting for the network was essentially a full-time job. Instruments are also loaned out to investigators as necessary. Problem areas included finding competent users of equipment and getting enough ground data to justify purchase of high resolution satellite images. Morisette took as an action item developing a more formal collaboration with Brown's group. Brown noted that the Chen's TRAC sensor is now commercially available at roughly US$2000.
Xubin Zeng, from Bob Dickinson's BATS EOS IDS group at U. of Ariz., discussed the climate modeling needs of the EOS LAI/fPAR data. He emphasized that three issues need to be addressed: LAI/fPAR versus fractional vegetation cover; accuracy requirement of LAI/fPAR in land modeling, and green versus total LAI. He discussed a new global 1 km fractional vegetation cover data he derived from the AVHRR NDVI data and IGBP land cover data in collaboration with several researchers. Using this data, it is possible to convert the leaf area with respect to pixel to the leaf area with respect to vegetated portion for use in climate models. Using several field experiments' data to drive a land model, he demonstrated that the same absolute accuracy of LAI would introduce much larger errors in simulating evapotranspiration over grassland (with smaller maximum LAI) than over tropical forest (with higher maximum LAI). It was suggested however that a relative error, rather than absolute, was more meaningful for that type of simulation. In addition, Zeng briefly discussed the community effort on the development of a Common Land Model (CLM) (one of the motivations for CLM is to make better use of EOS land data in land modeling). Myneni and Zeng discussed some inconsistencies in definitions between the LAI algorithm and land model input. Zeng cautioned that there appears to be a mismatch between the GCM community and product producers, and their understanding of what is needed. He suggested that two-stream radiative transfer scheme used in land models should be modified to be consistent with the derivation of satellite LAI and albedo data.
Privette briefly noted the possibility of including different modeling groups and validation networks (SAFARI 2000, LTER, SurfRad, FLUXNET, etc.) into the planning. Myneni brought up the need to know the uncertainty from the various field measurements and the frequency with which the data are collected. The discussion focused on ways to establish MODIS validation methods from the various sources. This methods could then be used to include other field work from different groups collecting LAI/fPAR measurements.
The group then broke into two discussion groups: 1) Auxiliary Measurements required for LAI/fPAR algorithm validation (Chair: Myneni) Myneni first noted that the current biome distribution of the 23 EOS Core Sites appeared sufficient as a foundation for global validation. It was decided that a subset of these sites with known LAI/fPAR measurements/investigators would be used in Year 1 to prototype a ramp-up activity. Representative biome sites included a mixed northern forest (Gower), a grassland and wheat site (Walter-Shea), a broadleaf agricultural site (Liang) and two woodland sites (Privette). Next, a prioritization of Measurements needed at these sites was outlined, including: 1. LAI and FPAR 2. nadir or bi-directional canopy (plot) spectra (or in MODIS spectral bands, particularly red, NIR, blue and green bands) 3. leaf optical properties (raw spectra preferred, although MODIS bands acceptable; only the mean value of the species in the vertical direction needed) 4. nadir soil/background spectra (or MODIS bands) 5. Fractional vegetation cover (a high IDS need was noted here) 6. Canopy allometric data (height, gap size, crown width) 7. Phenological stage (green-up, mature, senescent) and species composition (either by MODLAND structural classification or traditional taxonomy). 8. Additional parameters useful: wet/dry status (is canopy wet?), and the amount of non-photosynthesizing vegetation (litter, etc.) which could be measured at one "minimum photosythesis" stage of the year.
Myneni noted the goal was to constrain model for use in 1-D mode over a site. Not all measurements would be needed for validation, however, the more, the better. He noted also that leaf angle distribution, commonly measured, is not needed for validation. Emanuel noted that for BGC modelers could use basic meteorological information to help constrain LAI estimates.
2) Developing a Year 1 NPP product validation plan (Chair: Nemani) Nemani's
group developed a list of problem issues that needed resolved. Many of
these had already been raised in Nemani's talk, however the breakout provided
focus opportunity for BigFoot representatives to discuss their approach
with NPP algorithm developers.
_____________________________________________
DAY TWO: Defining tools and approaches at a site-level
_____________________________________________
EOS LAND VALIDATION CORE SITES
Morisette presented the EOS Land Validation Core Site list and discussed the planned EDC DAAC system that will acquire and archive image data for these sites and the ORNL DAAC "Mercury" system as a web-based access system for EOS validation field data. He noted the EDC system is in definitional stages and must be flexible to accommodate adjustments. Initially, Landsat and ASTER data ordering must be done via V0-IMS client. MODIS subsetting may be possible via the MODAPS at GSFC.
Liang inquired about cost-sharing of ETM+ imagery over core sites with MODLAND. Privette noted MODLAND's goal of purchasing at least four images/year/site. Emanuel suggests actively mining past LAI/fPAR data from Core Site scientists to develop expected temporal profiles and "reasonable ranges" for EOS products, and further to make recommendations to NASA HQ on the in-situ needs from site investigators. Morisette will take this as an action item. Liang questioned whether additional satellite data sets could be archived with the Core Site archives. Privette noted that EDC has thus far been fairly reluctant to take on Core Site archiving needs, however Strahler, a member of the EDC working group, suggested that EDC is taking a "wait and see" attitude for years 1 and 2, and if need is there, they will provide additional capacity and automation.
MQUALS
Huete presented the MODIS Quick Airborne Looks (MQUALS) package, which is near completion. He noted the difference in heterogeneity as a function of scale, and discussed the need for high resolution imagery. MQUALS is designed to provide: - direct radiometric validation - establish site-level biophysical-radiometric relationships - document surface conditions with 3-, 4-, band digital imagery at 0.1 - 1.0 meter spatial resolution.
The instrument package will feature an albedometer (full shortwave), a 3 band digital camera system and a 4 band radiometer (profiler). Huete recently acquired the digital camera system and had done initial testing on the ground. The system produced reasonable-looking NDVI imagery. He emphasized the need for irradiance data with which reflectances could be generated from the profiler (a ground calibration panel and second profiler will be used simultaneously with flight). He noted the profiler is to be used for radiometric integrity, while the integrity of the imager will be tested upon flight. He suggested the radiometry and ground data sets could provide the biophysical relationships needed to scale point data over larger areas. Resolution from the camera will be 10 cm to 1 m. He suggested that initial flights include 2 altitudes to test scaling, and be over MODIS calibration sites. Off-nadir pointing profilers could be added to the system.
Huete outlined the initial protocol for MQUALS: 1) conduct preliminary stratification for a core site, 2) optimally interpolate to fill in gaps in the biophysical field data (e.g, LAI), 3 ) provide radiometric data over area (corresponding to each LC or soil stratification type), 4) overlay radiometric data on digital imagery and develop relationships, and 5) extend to VIs, LAI, FPAR. This procedure is to be completed in 7 days after flight. Liang suggested that 3-4 MQUALS occur per year, and asked who would cover costs. Huete noted expected costs per deployment were $2500-$3000, but responsible party needs to be discussed.
BigFoot
Warren Cohen and Dave Turner presented the strategy for BigFoot (the successor to MODLERS), a program designed to measure and scale surface data for input to NPP models. BigFoot will model four FLUXNET sites in year one (BOREAS NSA, Harvard Forest, ARM/CART and Bondville, IL), and gear its activities to support MODIS LAI/fPAR and NPP validation. They presented direct and indirect approaches to collecting LAI sample and techniques for scaling up to the MODIS, 1km resolution. The field measurements will be scaled up to 1km for a 25 square km area over the flux tower. He noted the project is mostly centered on NPP estimation, though LAI and landcover are included to get to NPP. Together with Nemani, they will study how NPP and NEE covary, and hope to translate between these using modeling. Climatological and ecological controls on NPP will also be studied, as will be scaling errors on global NPP characterization.
BigFoot's basic approach will be to measure LAI, LC, FPAR and NPP on a 5x5 km area, extrapolate these to high resolution grids, characterize the errors, compare to MODIS products, isolate the effects of LC generalization, image grain size, and ecological modeling parameters, and examining spatial autocorrelation and other effects. A minimum seasonal repeat cycle will be used for direct measurements, possibly more for indirect. Cohen noted that details of the scaling methods will be worked out in a February, 99 meeting. Emanuel cautioned the BigFoot group about validating a BIOME-BGC-based MODLAND product with results from BIOME-BGC (circularity). All agreed that MQUALS initial science flights should focus on BigFoot sites.
EXAMPLE OF FIELD MEASUREMENTS
Brown discussed the various measurement used for validation of the Canada products and reviewed the TRAC instrument. He noted TRAC compensated for the limitations imposed by the LAI-2000, primarily related to clumping which occurs extensively in boreal forests. The result is underestimation of LAI (up to 100%). The TRAC measures sun flecks over transects to give gaps size distribution, which is translated to gap fraction.. He recommends transects be at least 50 m long and up to 200 m in more heterogeneous areas. A button on TRAC can be used to record fixed intervals on the transect in the data stream. Software takes solar angle into account. It samples at 32 Hz. Canadian researchers measure with LAI-2000 simultaneously. Wood-to-total area ratio and needle-to-shoot data from previous destructive measurements are used to estimate leaf area. LAI-2000 data is used for multiple-angle gap fraction data. The Canadians have measured 8-10 different sites, with 10-70 plots per site. The network is 1-2 years old and the data are ready for distribution.
Bev Law, working in a mixed-age ponderosa pine canopy (45 and 250 years old), young ponderosa pine (15 years old) and a juniper/sagebrush ecosystem discussed the Decagon Ceptometer (AccuPAR) and LAI-2000. These areas are very open with clumped canopies. She recently compared a variety of methods to estimate LAI at the mixed-age ponderosa pine forest. She compared estimates from ceptometer data (transmittance inversion; Campbell 1991), sapwood, litterfall and LAI-2000 data (corrected for needle clumping within shoot using shoot samples, and clumping at scales larger than shoot using tram data; Chen 1991). She has a complete stem map for the 100m x 100 m study area, and her group inverted a 3-D radiative transfer model to estimate LAI. Based on her comparisons, she recommends the LAI-2000 when corrections for clumping are used and uniform overcast sky conditions exist. She recommended leaf litterfall data only for hardwood deciduous forests, not conifers. She emphasized that green leaf area, not total leaf area, data should be reported, and that transmittance data must compensate for green-to-total area, because the instruments measure only white light, which is susceptible to woody material. Also, one must measure above and below the herbaceous understory. For moss backgrounds, only reflectance data or mass data needed. She suggests that the midsummer maximum LAI, and some spring and fall values are the most that a person can handle due to time.
Jiaquo Qi presented measurement techniques used for arid and semi-arid regions (San Pedro Basin, SALSA, HAPEX-Niger) using the LAI-2000, and various sampling strategies and scatter plots showing the relationship between different biophysical parameters. He estimates total LAI and green/total biomass over a 1.6 km x 1.8 km area. He noted difficulty in getting 0.5 LAI accuracy in grassland, as destructive is best but slow. He believes using gapping information outlined by Law may improve his LAI-2000 based estimates. He notes that leaves are too small in relation to the aperture of the leaf integrating sphere to get leaf spectra very well.
Betty Walter-Shea presented the work being done on the Osage, OK (ARM/CART area) wheat and tallgrass sites. This involves canopy reflectances, FPAR, soil reflectance, leaf optical properties, LAI, biomass, and others. She suggested that the group needs to develop a cost/effort table to PIs and NASA HQ on what it takes to acquire requisite data sets. It was noted that Walter-Shea has developed an efficient WWW-based way to provide up-to-date descriptions of her site and measurements to the greater community. See: http://enso.unl.edu/agmet/websummary.htm
The presentations and the discussion that followed indicated that different biomes do require unique considerations and that, even within a biome, different measurement techniques and methods of estimation can give different results. At the very least, the uncertainty in these measurements should be considered in their use as validation data or modeling efforts.
REVIEW OF RELATED MODIS PRODUCTS
Privette, substituting for Eric Vermote, reviewed the MODIS surface reflectance product and its validation. He noted validation is largely based on a relatively immature CIMEL sunphotometer/BRDF network that needs further development. He has submitted to GSFC a proposal to fund development of a 14-band next generation sunphotometer that could solve many of the problems in the current configuration. Status of that proposal should be known by Dec. 1. [Note the current BRDF instrument is currently collecting data at Howland, Maine; USDA Beltsville Farms, MD; and Oak Ridge, TN]. In addition, aircraft data from MAS and MASTER will be used for validation.
Huete discussed the validation of the VI products. He noted that MODLAND reflectances are composited, not just the VIs, such that development of other indices is possible. The MODLAND VI product in general measures chlorophyll-associated change, regardless of cause. He would prefer to validate the model, but will initially seek only to validate the product. He also pointed out that validation in this context is more than just validating the input reflectance data -- it is also important to validate that the index responds to actual vegetation changes. For example, what is the lower threshold of vegetation change? What is the upper range of saturation? What is the sensitivity in different biomes? Huete will seek to express the VI and its uncertainties in terms of biophysical units like LAI and FPAR. He noted that there is room to develop VI vs. LAI/fPAR model validation.
Wolfgang Lucht and J.P. Muller discussed validation of the BRDF/Albedo products. They will seek to relate isotropic scattering to overall land reflectance, and volumetric scattering to LAI, and geometric-optical scattering to land surface structure/roughness. These are three components of albedo that might be checked, at least for consistency. They will also develop relationships between BRFs and albedo based on in-situ measurements, and develop continental albedo with these relationships and AVHRR. Lucht also will attempt to relate NDVI, shadowing and kernel values that are retrieved from MODIS. They have a declining interest in the BSRN and SurfRad global networks, which appear to have too limited footprints. They are putting heavy emphasis on ARM/CART, MQUALS, and some dependence on episodic data and one-time images.
Doug Muchoney reviewed the IGBP land cover validation workshop. The presentation discussed both the uncertainly of the products and its effect on the LAI/FPAR/NPP products and their validation. The surface reflectance and land cover directly effect the LAI/fPAR product because they are input (biome type is the direct input, which is derived from land cover). The VI product uncertain will effect the LAI/fPAR validation through the VI/LAI and VI/FPAR relationship (as described as "test 4" by Myneni in day one). The BRDF/Albedo validation addressed methods for assessing the field measurement scaling issue.
DISCUSSION
The group then discussed how the presentation and discussions from the past two days could help meet the objectives of the meeting and further develop a validation plan for MODIS biophysical products. Two general topics were 1) data protocols and 2) scaling/mixtures.
The group decided to focus on 11 Sites in the first year after launch.
The sites were chosen primarily to cover all MODLAND biome types and to have funded data collection and access, and included:
1 Uardry, Australia (grass; Simon Hook, AM-1 Val.) 2 Osage, Oklahoma (grass; Shashi Verma) 3 East Anglia, England (grass; J.P. Muller, MODIS) 4 Skukuza, South Africa (woodland; Privette, AM-1 Val.) 5 Mongu, Zambia (woodland; Privette, AM-1 Val.) 6 San Pedro Basin/SALSA, Ariz. (shrubland; Huete/Qi, MODIS) 7 USDA BARC, Maryland (broadleaf crop; Liang, AM-1 Val.) 8 Bondville, Illinois (broadleaf crop; BigFoot) 9 Harvard Forest, Mass. (broadleaf forest; BigFoot) 10 Park Falls, Wisc. (broadleaf forest; Gower, AM-1 Val.) 11 BOREAS NSA, Canada (needle leaf forest; BigFoot)
Several independently-funded investigators have also volunteered for possible participation in year one validation, including:
1 Cascades, Or. (needleleaf forest; Law) 2 Canada LAI Network (forest;
Brown) 3 Krasnoyarsk, Russia (needleleaf forest; Slava Kharuk) 4 New Zealand
(forest; Brown) 5 Gainesville, Florida (broadleaf crop; John Craig) 6 EMATREF,
France (needleleaf forest; Jean-Louis Roujean)
This is a pilot activity, and participants stressed that the number of sites will ramp up significantly with time.
Chris Justice discussed the need to better involve the Japanese GLI and EOS MISR validation groups/sites/activities, and underscored the need to develop real site contacts at all core sites. Emanuel suggested that existing networks be contacted about historical data sets that could provide realistic ranges. The ORNL DAAC mercury system will facilitate the distribution and access of this data. Beyond these sites, there was discussion about other locations collecting biophysical reference data and how this might be used. The discussion turned to incentives for encouraging collaboration and the idea of data sharing (both EOS image data and field data between sites) seemed to be a good solution. A suggestion for a CD-ROM with the initial Core Site data sets, in addition to the WWW pages, received strong interest. An EOS Earth Observer article upon AM-1 launch might help as well. Discussion turned to how these "other data sources" could be utilized for validation and it was proposed that the 11 initial sites will provide the lessons/tools needed to incorporate other data sets.
The initial sites will produce validation results within a year after launch and the lessons learned and techniques developed from these will be used to ramp-up validation analysis to include the wider array of sites (and networks of sites) that could be available during the second year after launch. It was noted that in some cases site personnel will only collect the data, but not compare it to MODIS products. Myneni discussed his plan to increase his group's attention to validation issues beginning in the New Year, and expected they would be able to handle all year 1 in situ data comparisons.
What number of in-situ sample sites should we be driving towards in year 2? 30
2. At what frequency (per season) do LAI/fPAR in-situ measurements need to be taken over:
green-up mature senescencing dormant temperate broadleaf forest 1 1 1 1 temperate coniferous forest 0 1 0 1 broadleaf cropland 2 2 1 0 cereal cropland 2 2 1 0 desert 0 1 0 1 shrubland/woodland. 1 1 1 1
3. Using high spatial resolution remote sensing imagery (TM/ASTER/MQUALS) to extrapolate in-situ values, how many LAI/fPAR assessments are needed per year over each of the sites?
Estimates: temperate broadleaf forest 6 temperate coniferous forest 4 broadleaf cropland 8 cereal cropland 8 desert 4 shrubland/woodland. 6
4. What is the minimum area per site over which in-situ measurements need to be taken?
Sufficient, representative coverage to extrapolate to 5 by 5 km. Estimated: 2x2 km
5. What is the minimum area over which aerial extrapolation, using high resolution imagery should occur?
5 x 5 km area
6. In two-storied systems, are both under- and overstory LAI values expected or a single, total value?
Under and overstory LAI values are needed
7.What is the maximum uncertainty tolerable in site LAI values? +/-0.6 What uncertainty is realistically desired? +/-0.2- to +/-0.4
8. How should areas containing land cover mixtures in every 1km^2 be handled in year 1?
Data provider measures individual LAI/fPAR per Land Cover-type, and spatial Land Cover map. Mixed LC sites should not be emphasized in year 1 sampling, emphasize more "mixed" sites for years 2-4.
In such mixed areas, what sampling frequency should be used? That of fastest changing LC-type
9. What additional data is most useful but otherwise would not be collected for validation?
See meetings minutes for full description.
10. What fraction of international (vs. domestic) validation sites are needed for LAI/fPAR in year 1?
GOAL: 25% International sites
11. How many sites will already be sampled under existing Validation funding at the resolutions noted above? AM-1 Funding = 6 : Privette (2), Meyer (1) Walter-Shea (2), Liang (1), Gower (1) (Site funded by BigFoot program = 4, Possibility of other NASA sites ~2, or independent activity ~10)
12. What is the approximate cost of sampling each of the biome types one time (in-situ)? Person*days*salary/person=1*2*40k/365=$800
13. FLUXNET/GTOS sites offer a way to expand the EOS LAI/fPAR validation. What would constitute an affordable pilot activity to demonstrate how EOS validation could interact with GTOS ö which sites, what can they collect, do we have protocols for in-kind provisions?
(Part of ongoing work of Baldocchi and Olson's EOS FLUXNET program.)
14. Is there some kind of binding authority/contract that can be set up to ensure that a non-funded group delivers the required in-situ data?
Leverage on professional good will and emphasis from NASA Research program for non-EOS projects. Exchanges for MQUALS and TM data. 15. If we suggest prioritized archive and distribution of AM-1 data, or L7 data sets, can EDC meet this?
Five sites for year one pilot, possibly all core site if a need is shown.
If not, what is needed and where will it come from? MODAPS, processing ESIPs, and ORNL are possible distribution points
16. How to best integrate MQUALS into validation protocol?
Huete has developed a protocol, this will put on-line. The sites for 1999 MQUALS will be worked out at the December 1998 MODIS Land Science team meeting.
17. What fraction of LAI/fPAR collectors are simply providing the raw data to MODLAND, and what fraction plan to compare these data to MODLAND products and make evaluations themselves?
Encourage field personnel to pick up task, otherwise BU can accommodate
year 1 load.
Monday, Oct. 12 TITLE: Defining Broad Requirements for Global LAI/FPAR/NPP Products1:00 - 1:15 Charge for meeting and day (Justice/Privette)
1:15 - 1:30 MODIS validation overview (Morisette)
1:30 - 1:45 Summary of December, 1997, SWAMP Validation Meeting (Huete)
1:45 - 2:00 Production and Validation of NPP (Nemani)
2:00 - 2:30 Production and Validation of LAI/fPAR (Myneni)
2:30 - 3:15 current players and anticipated roles (Privette)
Zen -- ids application of LAI
Brown -- Canadian val. network
Privette -- other possible val. networks
(and Airborne sensors)3:15 - 3:30 Break
3:30 - 5:00 Group discussion on LAI/FPAR/NPP validation (outline - Privette)
__________________________________________
Tuesday: TITLE: Defining tools and approach at a site-level
9:00-9:10 Charge (Justice/Privette)
9:10 - 9:20 TM/ASTER orders/availability (Morisette)
9:15 - 9:40 MQUALS (Huete)
9:40 - 10:00 Validation protocol within BigFoot (Running, Cohen))
10:00 - 12:30 Biome-specific protocols - focus on sampling design and scaling12:30 - 1:30 Lunch
1:30 - 2:00 Surface Reflectance Relationship (Vermote)
2:00 - 2:30 VI relationship (Huete)
2:30 - 3:00 BRDF/Albedo relationship (Strahler/Muller)3:00 - 3:15 Break
3:15 - 4:00 Discussion on protocols (group)
4:00 - 5:30 Develop implementation plan (Morisette/Privette - group)
5:45 - 6:00 Conclusion and responsibilities (Morisette - group)
August 10th and 11th meeting was postponed as of July 10th, 1998
There is parking underneath at $12/day and on the street. Since Monday is a legal holiday, the garage is closed but the meters are free. However, I've been told that if the city wants to enforce the time limits for a space, they can, but probably won't due to reduced staffing.
Tunnel Warning:
If you are driving to BU from Logan Airport on the Columbus Day Holiday, you will find you have two tunnels to choose from. You want the SUMNER TUNNEL, NOT the Ted Williams Tunnel.
Explanation: The new Ted Williams Tunnel connects the airport and South Boston. Eventually it will link up with the Mass Turnpike and I-91, but now it only debauches you in the middle of a major construction zone for our new underground central artery. Normally it is only open to commercial traffic, but on weekends and holidays it is open to general traffic as well.
BY PUBLIC TRANSPORTATION: To School of Management
By Taxi: Take a taxi from the airport. Tell the driver you want the Boston University School of Management located just past the Howard Johnson's in KENMORE SQUARE on COMM AVE. The fare should be around $15. Tip 10-15% on the meter. There will be an extra charge for airport service.
By Subway (The "T"):
1. Take the airport shuttle bus (free) to the T stop, buy a token (85
cents). Then take the BLUE LINE inbound.
2. At the GOVERNMENT CENTER stop, change to the GREEN LINE westbound,
B, C, or D car. All of these will stop at KENMORE.
3. Get off at KENMORE. As you exit from Kenmore at the turnstiles,
turn right. You will emerge on the north side of Commonwealth Avenue. (If
you can see the Barnes and Noble Bookstore across the street, you are on
the south side of Comm Ave. Go back down and take the tunnel to the north
side.)
4. Walk west about 2 blocks to the School of Management, on the right.
Related Sites:
http://enso.unl.edu/agmet/websummary.htm : Betty Walter-Shea's sites
http://cdiac.esd.ornl.gov/programs/NIGEC/fluxnet/
: FLUXNET
Responsible NASA official: Dr. Jeffrey L. Privette, Code 923, Goddard Space Flight Center