Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: USING REMOTE SENSING & MODELING FOR EVALUATING HYDROLOGIC FLUXES, STATES, & CONSTITUENT TRANSPORT PROCESSES WITHIN AGRICULTURAL LANDSCAPES Title: Estimating leaf area index from Landsat using MODIS LAI products and field measurements as reference

Authors
item GAO, FENG
item ANDERSON, MARTHA
item Houborg, Rasmus -
item KUSTAS, WILLIAM

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: March 23, 2012
Publication Date: July 27, 2012
Citation: Gao, F.N., Anderson, M.C., Houborg, R., Kustas, W.P. 2012. Estimating leaf area index from Landsat using MODIS LAI products and field measurements as reference[abstract]. IEEE Geoscience and Remote Sensing Symposium. 2012 CDROM.

Technical Abstract: Leaf area index (LAI) is a key biophysical parameter used in most land surface models. Operationally, LAI products currently used typically come from coarse resolution sensors such as the Moderate Resolution Imaging Spectroradiometer (MODIS). While coarse resolution data at the kilometer scale are often sufficient for global, continental and regional scale applications, field scale applications require finer pixel resolution data such as Landsat (30m). Standard LAI retrieval approaches (either empirical or physical) for Landsat data will likely produce LAI fields that are inconsistent with MODIS LAI product. To make a MODIS-consistent LAI product from Landsat imagery, which provides field scale estimates of LAI, a simple reference-based regression tree approach was recently developed. Previous study shows that the reference-based approach can produce accurate estimates of LAI from Landsat as evaluated using field measurements. However, there is some underestimation (about 0.4) for the high LAI (>3) due to several reasons. One major reason is that the reference-based approach limits MODIS LAI samples to being both pure and of high quality, LAI reference information may be lost for certain surface types or phenological stages. Pixels with high LAI are easily saturated in surface reflectance and thus LAI retrieval quality may be low. This may affect the results when MODIS LAI is used as the sole reference in retrieving Landsat LAI, especially when low or high LAI values are missing from the MODIS samples resulting in the regression models extrapolating LAI outside the training data range. In such cases, LAI measurements acquired on the ground may help to fill the gap, providing LAI information over a full data range of LAI expected at the Landsat scale. This paper tested this hypothesis by combining MODIS LAI samples with field measurements. Results show that the combined samples provide better estimates, especially for fields with high LAI. All difference statistics are improved in comparison to using only MODIS samples. The strategy to combine samples from MODIS LAI product and field measurements provided better field scale estimates yet still consistent with MODIS LAI products. Results will be useful for models that require consistent LAI inputs at different spatial resolutions.

Last Modified: 9/29/2014
Footer Content Back to Top of Page