Extended Research Plans for EO-1 Validation and Evaluation for Agricultural Monitoring

Final report "EO-1 validation and evaluation for agricultural monitoring" (March, 2000 – Feb., 2003) (Draft)

S. Liang (PI), F. Hummerich, C. Walthall, C. Daughtry and J. Morisette

1 Originally proposed objectives

In our original three-year proposal, there were five objectives, including

· 1) evaluating the ALI reflectance using ground measurements, airborne and other high-resolution satellite observations (i.e., ETM+ and ASTER) over Beltsville, Maryland;

· 2) evaluating the improvements to atmospheric correction of ALI and ETM+ through use of data from the LAC and Hyperion. A new algorithm for estimating atmospheric water vapor content from both Hyperion and LAC will be tested;

· 3) evaluating the improvements of ALI and Hyperion to the inversion of geophysical (shortwave radiation budget components) and biophysical parameters (LAI and FPAR) over TM and ETM+ by using the same inversion algorithm;

· 4) estimating fAPAR and crop photosynthetic efficiency from Hyperion data;

· 5) estimating litter cover from Hyperion data.

Because the proposal was funded for two years (later with a half-year extension) and particularly the launch was delayed, we therefore revised our working plans accordingly.

The emphasis was then on evaluating how land surface biophysical and geophysical variables can be retrieved more effectively from ALI imagery than ETM+ imagery. Most goals have been achieved, and many activities were far beyond what we proposed. Several journal papers have been published or are in the publication process. In particular, the PI finished a book (Liang, 2003) during the period of this project. Several graduate students were also financially supported by this grant.

The major achievements are briefly summarized in the following section.

2 Major achievements

2.1 Development of new atmospheric correction algorithms

To retrieve surface reflectance, atmospheric correction is a necessary step. If atmospheric parameters, particularly aerosol optical depth, are known, the retrieval is relative easy. However, there is no simultaneous measurement of aerosol properties in many cases. Moreover, aerosol horizontal distribution often varies dramatically. In the original proposal, we proposed to apply the “dark-object” method that we developed earlier (Liang, 1997). We later found that this method is quite limited mainly because it requires the existence of large homogeneous dense vegetation canopies around the scene. Therefore, we developed a new method suitable for all land surfaces. Because of the delay of the EO1 launch, we first tested the method using ETM+ imagery (Liang, et al., 2001) and then applied to ALI (Liang, et al., 2003b) and other imagery (MODIS and SeaWiFS) (Liang, et al., 2002). The basic idea of this algorithm is its use of spectral signatures, specifically correcting visible reflectance using near-IR bands. This algorithm relies on determination of hazy regions using shortwave bands and clusters using near-IR bands. Since ALI has one more blue band and two more near-IR bands than ETM+, this apoorach performs better using ALI imagery. Details are available in the published articles.

We have also further extended this algorithm specifically for correcting hyperspectral imagery (e.g., Hyperion imagery). Several examples were presented in the last science team meeting. More testing and validation are needed before it can be published.

2.2 Retrieval of surface broadband albedo

Land surface broadband albedo is a critical variable affecting the earth's climate. It has been well recognized that surface albedo is among the main radiative uncertainties in current climate modeling. There are usually three basic steps in calculating surface broadband albedo: 1) atmospheric correction that converts top-of-atmosphere (TOA) radiance to surface directional reflectance, 2) BRDF modeling that converts directional reflectance to spectral band albedos; and 3) narrowband to broadband conversion that converts band albedos to broadband albedos. Atmospheric correction was discussed in the previous section. For a single nadir-viewing imagery, such as ETM+ and ALI, a Lambertian surface has been assumed at this stage. Our emphasis was on the third step (i.e., narrowband to broadband conversion).

We developed a new approach for generating conversion formulae based on extensive radiative transfer simulations. To develop a representative formula, lots of surface reflectance spectra must be used. We developed a new procedure to reduce the computational burden and generated the conversion formulae for a variety of sensors, and ground measurements verified that these formulae are very accurate (Liang, 2000; Liang et al., 2003a,b; Van Niel et al., 2003). The details are omitted here since they are available from the published papers.

Since ALI has three more bands than ETM+, it is able to produce more accurate conversions of narrowband to broadband albedo using our technique.

2.3 Retrieve of Leaf Area Index

Leaf area index (LAI) has been widely used in ecology and almost all land surface process models. Its radiative representation is fraction of photosynthetically active radiation absorbed by vegetation canopies (FPAR). In this study, we have mainly focused on LAI since we did not have instruments for measuring FPAR in the field.

We have refined and developed the hybrid algorithm proposed in our original proposal. The new hybrid algorithm combines extensive canopy radiative transfer simulations (physical) with nonparametric regression (i.e. neural network) (statistical). The results indicate that this technique is very effective. We tested this idea on ETM+ before ALI was available (Fang and Liang, 2003) and then applied to ALI (Liang, et al., 2003b). It was also compared with a genetic algorithm developed during this project (Fang et al.,, 2003). It is found that the additional two near-IR bands are very useful to estimate LAI using this technique (Liang, et al., 2003b).

This hybrid algorithm is being extended to hyperion data. Unfortunately, we have not been able to produce journal papers at this point.

2.4 Estimating Litter Cover from Hyperspectral Data

Because of the fact that we got Hyperion data very late and few acquisitions are available over BARC, our work have been mainly based on AVIRIS data acquired over Beltsville, Maryland through this project. There have been several presentations in the JPL AVIRIS workshops. More details will be provided later.

2.5 Field campaigns and data collection/analysis

Many field campaigns were took place in Beltsville, Maryland, and Colleambally, Australia with EO1 and other aircrafts overpass. Extensive ground measurements

were made. These data sets have been used for validating a series of new algorithms mentioned in the previous sections. They will also be extremely valuable for other projects in the future.

3 References

There are several conference papers, but only peer reviewed papers are listed here.

Fang, H. and S. Liang, (2003), "Retrieve LAI from Landsat 7 ETM+ Data with a Neural Network Method: Simulation and Validation Study," IEEE Transactions on Geoscience and Remote Sensing, In press.

Fang, H., S. Liang and A, Kuusk, (2003), Retrieving Leaf Area Index (LAI) Using a Genetic Algorithm with a Canopy Radiative Transfer Model, Remote Sensing of Environment, in press.

Liang, S., Quantitative Remote Sensing of Land Surfaces, John Wiley and Sons, Inc., approximately 550 pages, in press.

Liang, S., Narrowband to Broadband Conversion of Land Surface Albedo: I. Algorithms, Remote Sensing of Environment, 76:213-238, 2001.

Liang, S., H. Fang, M. Chen, Atmospheric Correction of Landsat ETM+ Land Surface Imagery: I. Methods, IEEE Transactions on Geosciences and Remote Sensing, 39:2490-2498, 2001.

Liang, S., H. Fang, J. Morisette, M. Chen, C. Walthall, C. Daughtry, and C. Shuey, (2002), Atmospheric Correction of Landsat ETM+ Land Surface Imagery: II. Validation and Applications, IEEE Transactions on Geosciences and Remote Sensing, 40(12):2736-2746.

Liang, S., C. Shuey, A. Russ, H. Fang, M. Chen, C. Walthall, C. Daughtry, (2003a), Narrowband to Broadband Conversions of Land Surface Albedo: II. Validation, Remote Sensing of Environment, 84(1):25-41.

Liang, S., H. Fang, M. Kaul, T. Van Niel, T. McVicar, J. Pearlman, C. Walthall, C. Daughtry, F. Huemmerich, (2003b), Estimation of land surface broadband albedos and leaf area index from EO-1 ALI data and validation, EO-1 special issue of IEEE Transactions on Geoscience and Remote Sensing, in press.

Van Niel, T., T. McVicar, H. Fang, S. Liang, (2003), Calculating environmental moisture for pre-field discrimination of rice crops, International Journal of Remote Sensing, in press.