Remote Sensing of Forest Canopy Structure Using Landsat TM and Laser Altimetry Data

Johnston, Andrew K.(1) Lefsky, Michael A. (2) Parker, Geoffrey G. (3)

(1) Center for Earth and Planetary Studies, National Air and Space Museum, Smithsonian Institution
(2) Pacific NW Research Station, USDA Forest Service
(3) Smithsonian Environmental Research Center, Smithsonian Institution


Forests cover about a third of the EarthÕs land surface and form the central component of the stocks and acquisition of carbon in the biosphere. Measuring the growth and distribution of forests using the Landsat Thematic Mapper (TM) and similar remote sensing instruments is a central issue in global change research. In this study, the effect of forest canopy structure on reflectance measured by Landsat TM was investigated using data from an airborne laser altimetry system known as SLICER (Scanning Lidar Imager of Canopies by Echo Recovery). SLICER provides high resolution three dimensional measurements that are sensitive to the vertical organization of forest canopy structure (Lefsky et al. 1999, Harding et al. 1999). While the relationship between canopy structure and reflectance has been the focus of a great deal of research (e.g. Rosema et al. 1992, Hall et al. 1995), these studies have not produced accurate estimates of important variables such as biomass. Unlike a lidar instrument such as SLICER, the organization of the forest canopy is not directly visible to Landsat TM and similar sensors.

The purpose of this work is to perform a simple comparison of Landsat TM and SLICER measurements to determine the effect of canopy height on reflectance. The ground site chosen for the analysis consists of mixed deciduous forests with variable stand ages. The results demonstrate relationships between Landsat reflectance and SLICER-measured canopy height. In the future, knowledge of this relationship may allow canopy-reflectance relationships to be applied across much larger areas.

Figure 1.
Map showing the location of the study area.

Figure 2.
Landsat TM image of the Smithsonian Environmental Research Center.
Data and Methods

The remote sensing and field data used in this study covers the Smithsonian Environmental Research Center (SERC), which is located on the Chesapeake Bay east of Washington, DC (figure 1). A Landsat TM image of SERC (figure 2) shows the mixture of landcover types, mostly deciduous forest, present in the area. Extensive measurements have been made of the forests at SERC (Parker et al. 1989), including measurements of vertical canopy structure (Brown and Parker 1994).

The design of the SLICER instrument is based on laser altimetry technology, in which the round-trip travel time of a pulse of laser light is measured to estimate the distance to a reflecting surface (Harding et al. 1994). The entire laser return signal is digitized, resulting in a waveform that records the reflection of light from multiple canopy elements over a 10m footprint with 11cm vertical resolution. This type of data, recording the vertical distribution of canopy surface area, can be used to predict stand attributes such as above ground biomass and basal area (Lefsky 1997, Harding et al. 1999). SLICER footprints were geolocated using GPS data and avionic records. SLICER transects shown on a grayscale Landsat image (figure 3) indicate the coverage of airborne lidar data.

Both the SLICER waveform and ground data can be transformed to yield the vertical distribution of reflective surfaces, called the Canopy Height Profile (CHP). A mathematical description of canopy height, introduced by Lefsky et al. (1999), can be calculated from this:

where QMCH is the quadratic mean canopy height and CHP[i] is fraction of canopy surfaces at height i. The QMCH is highly correlated to mean and median canopy height within 10m SLICER footprints (Lefsky et al. 1999). SLICER measurements of QMCH (see figure 4) have been shown to be sensitive to field data (Lefsky et al. 1999, Harding et al. 1999). Validation of the ability of SLICER to measure canopy structure is being reported elsewhere (Harding et al. 1999).

In this study, lidar data were spatially referenced to the Landsat data by aggregating the SLICER footprints within each TM pixel. SLICER footprints are circular with a 10m diameter, and Landsat TM pixels are 30m squares. Each TM pixel, therefore, can contain nine SLICER footprints. In this analysis, the values for SLICER footprints within each TM pixel were averaged. Only TM pixels that contained at least seven SLICER footprints were used.

The Landsat TM image corresponding to the date of the SLICER data (September 1995) was used in this analysis. Ground reflectance was calculated by correcting for solar zenith angle and assuming a standard mid-latitude atmosphere. Sites containing satellite-measured reflectance and SLICER-measured canopy height (n=1269) were extracted and compared. The relationship between forest canopy height and reflectance was defined for TM bands 3 (.63-.69mm), band 4 (.76-.9mm), and band 7 (2.08-2.35mm). This consists of visible, near-IR, and mid-IR wavelengths.

Figure 3.
SLICER tracks shown on grayscale Landsat TM image.

Figure 4.
SLICER-measured canopy height vs. ground data (figure from Lefsky et al. 1999).

Results and Conclusions

Comparisons of Landsat reflectance and SLICER-measured canopy height (figure 5) demonstrate broad relationships. The horizontal axis in figures 5a-c shows canopy height in meters, and the vertical axis shows Landsat TM reflectance. Taller forest canopies are associated with higher reflectance in the near-IR and lower reflectance in visible and mid-IR wavelengths. Figures 5a-c show relationships that are similar to those seen in other projects (e.g. Nilson and Peterson 1994). Mid-IR and visible wavelength reflectance is negatively correlated to canopy height due to chlorophyll absorption and structural shadowing effects, while near-IR reflectance is positively correlated due to the light scattering properties of leaf internal structure.

These relationships saturate at low reflectance values for visible and mid-IR bands, perhaps caused by smaller canopies being shadowed by adjacent tall trees. This heterogeneity at 10m deservers attention in future work. The lack of a close relationship is also caused by the limited type of ground data (solely from canopy height), although a variety of factors determine reflectance. This variable was chosen because it facilitated simple comparisons to satellite data. Future comparisons can incorporate more extensive SLICER-derived parameters of the vertical canopy structure.

Figure 5d shows the relationship between SLICER-measured canopy height and a spectral vegetation index, the NDVI. While vegetation indices have been found to be broadly related to vegetation amount and vigor (e.g. Goward et al. 1991) they have not been found sensitive to canopy structure. The results in figure 5d also show that this relationship is not close, especially at high NDVI values. This saturation is likely caused by pockets of smaller but rapidly growing vegetation that absorbs high levels of photosynthetically active radiation.

The results indicate a wide range of scatter in the reflectance/canopy height relationship. Although this is similar to results seen in other work, canopy height cannot be predicted with accuracy from these relationships. However, the use of laser altimetry in this study provided two important advantages. The lidar data created a unique opportunity to incorporate data at a scale similar to the orbital data. This contrasts with allometric measurements of forest canopies, which usually cover areas much smaller than satellite image pixels. The use of SLICER data also allowed these comparisons to be made for a far greater number of points than would have been possible using ground data. These advantages can be explored in future work by comparing reflectance to both lidar and ground data in a variety of locations, indicating where the use of lidar may provide measurements of forest canopy structure that are more complementary to satellite remote sensing.

Figure 5. Landsat TM reflectance plotted against SLICER-measured canopy height.


Brown, M.J., and Parker, G.G., (1994) Canopy light transmittance in a chronosequence of mixed-species deciduous forests. Can. Journal Forest Research 24:1694-1703.
Goward, S.N., Markham, B., Dye, D.G., Dulaney, W., Yang, J. (1991) Normalized difference vegetation index measurements from the Advanced Very High Resolution Radiometer, Remote Sensing of Environment 35:257-277
Hall, F.G., Shimabukuro, Y.E., Huemmrich, K.E. (1995) Remote Sensing of forest biophysical structure using mixture decomposition and geometric reflectance models, Ecological Applications 5:993-1013
Harding, D.J., Lefsky, M.A., Parker, G.G., and Blair, J.B. (1999) Lidar alitmeter measurements of canopy structure: methods and validation for closed-canopy, broadleaf forests. Remote Sensing of Environment (in review).
Harding, D.J., Blair, J.B., Garvin, J.G., Lawrence, W.T. (1994) Laser altimeter waveform measurement of vegetation canopy structure. Proceedings of IGARSS 1994.
Lefsky, M.A. , Harding, D., Cohen, W.B., Parker, G.G., and Shugart, H.H. (1999), Surface Lidar Remote Sensing of Basel Area and Biomass in Deciduous Forests of Eastern Maryland, USA, Remote Sensing of Environment 67, 83-98.
Lefsky, M.A. (1997) Lidar remote sensing of canopy height profiles: application to spatial and temporal trends in canopy stucture. Ph.D. Thesis. University of Virginia.
MacArthur, R.H., and Horn, H.S. (1969) Foliage profile by vertical measurements. Ecology 50: 802-804.
Nilson, T., Peterson, U. (1994) Age dependence of forest reflectance: analysis of main driving factors, Remote Sening of Environment 48:319-331
Parker, G.G., OÕNeill, and Higman, D., (1989) Vertical profile and canopy organization in a mixed deciduous forest. Vegetatio 89: 1-12.
Rosema, A., Verhoef, W., Noorbergen, H., Borgesius, J.J. (1992) A new forest light interaction model in support of forest monitoring, Remote Sensing of Environment 42:23-41.