PERS March 2015 Members - page 205

RF). The step of feature selection might be useful not only for
spectral mixture analysis of hyperspectral remote sensing, but
also for multi-spectral remote sensing classifications. The red,
NIR, and SWIR wavelength regions were found to be important
for discriminating desiccated and dead canopy spectra. Our
analysis shows again that the variations in spectral signatures
caused by stress or disturbance are not equal across the full
wavelength range, and that spectral features at specific wave-
lengths are valuable for monitoring disturbed or stressed plant
canopies (Ahern et al., 1991; Radeloff et al., 1999; Hurley et
al., 2004; Santos et al., 2010).
Further studies evaluating the impacts of the tamarisk bio-
control program could be accomplished using data acquired
from high spatial resolution sensors (i.e., WorldView-2) and
ground-based ET measurements. More detailed and accurate
mapping of tamarisk canopy classes may be possible with
high spatial resolution, multispectral data. By exploring the
relationship between spectral indices and ground-based ET
measurements, high spatial resolution data may also assist
in assessing water salvage (Nouri, 2014). Multi-temporal
analysis using either high spatial resolution multispectral or
hyperspectral data may also result in improved classification
of tamarisk defoliation and mortality, as well as identification
of mortality trends over time.
Acknowledgments
The authors would like to thank the Donald R. Currey Gradu-
ate Research Scholarship and University of Utah Rio Mesa
Center for supporting the Dolores River field work. We also
would like to thank following people for their valuable con-
tributions to the Dolores River field work: Zachary Lundeen,
Hau Truong, Kenneth Dudley, and Timothy Edgar.
References
Ahern, F.J., T. Erdle, D.A. Maclean, I.D. Kneppeck, 1991. A quantita-
tive relationship between forest growth rates and Thematic Map-
per reflectance measurements, International Journal of Remote
Sensing, 12:387–400.
Asner, G.P., and D.B. Lobell, 2000. A biogeophysical approach for au-
tomated SWIR unmixing of soils and vegetation, Remote Sensing
of Environment, 74:99–112.
Asner, G.P., C.A. Wessman, D.S. Schimel, and S. Archer,1998. Vari-
ability in leaf and litter optical properties: Implications for BRDF
model inversions using AVHRR, MODIS, and MISR, Remote
Sensing of Environment, 63:243–257.
Bian, M., A.K. Skidmore, M. Schlerf, T. Fei, Y. Liu, and T. Wang,
2010. Reflectance spectroscopy of biochemical components as
indicators of tea (Camellia sinensis) quality, Photogrammetric
Engineering & Remote Sensing, 76(12):1385–1392.
Boochs, F., G. Kupfer, K. Dockter, and W. Kuêhbauch, 1990. Shape of
the red edge as vitality indicator for plants, International Journal
of Remote Sensing, 11:1741–1753.
Breiman, L., 2001. Random forests, Machine Learning , 45:5–32.
Carruthers, R.I., C.J. DeLoach, J.C. Herr, G.L. Anderson, and A.E.
Knutson, 2008. Areawide pest management: Theory and
implementation, Salt Cedar Areawide Pest Management in the
Western USA (O. Koul, editor), CAB International, Wallingford,
UK, pp. 271–299.
Carter, G.A., 1993. Responses of leaf spectral reflectance to plant
stress, American Journal of Botany, 80:239–243.
Cochrane, M.A., 1998. Linear mixture model classification of burned
forests in the Eastern Amazon, International Journal of Remote
Sensing, 19:3433–3440.
Congalton, R.G., 1991a. A review of assessing the accuracy of clas-
sifications of remotely sensed data, Remote Sensing of Environ-
ment, 37:35–46.
Congalton, R.G., 1991b. Remote sensing and geographic information
system data integration: error sources and, Photogrammetric
Engineering & Remote Sensing, 57(5):677–687.
Congalton, R.G., and K. Green, 2009. Assessing the Accuracy of
Remotely Sensed Data: Principles and Practices, Second edition,
CRC/Taylor & Francis, Boca Raton, Florida, 150 p.
Daigo, M., A. Ono, R., Urabe, and N. Fujiwara, 2004. Pattern decom-
position method for hyper-
multi-spectral data analysis, International Journal of Remote
Sensing, 25:1153–1166.
Dale, V.H., L.A. Joyce, S. McNulty, R.P. Neilson, M.P. Ayres, M.D.
Flannigan, P.J. Hanson, L.C. Irland, A.E. Lugo, C.J. Peterson, F.J.
Swanson, B.J. Stocks, and B.M Wotton, 2001. Climate change
and forest disturbances, BioScience, 51:723–734.
Datt, B., 2000. Identification of green and dry vegetation components
with a cross-correlogram spectral matching technique, Interna-
tional Journal of Remote Sensing, 21:2133–2139.
Dennison, P.E., and D.A. Roberts, 2003. Endmember selection for
multiple endmember spectral mixture analysis using endmem-
ber average RMSE, Remote Sensing of Environment, 87:123–135.
Dennison, P.E.,P.L. Nagler, K.R. Hultine, E.P. Glenn, and J.R.
Ehleringer, 2009. Remote monitoring of tamarisk defoliation and
evapotranspiration following saltcedar leaf beetle attack, Remote
Sensing of Environment, 113:1462–1472.
Dudley, T.L., and D.W. Bean, 2012. Tamarisk biocontrol, endangered
species risk and resolution of conflict through riparian restora-
tion, BioControl, 57:331–347.
Table 4. RF Confusion Matrices for Dead, Green, Yellow, and Brown C anopies Using Feature-Selected Canopy Spectra
Dead canopies Brown canopies Green canopies Yellow canopies User’s accuracy (%)
Dead canopies 14 0 0 1 93.3
Brown canopies 0 0 0 8 0.0
Green canopies 0 0 13 4 76.5
Yellow canopies 2 4 3 18 66.7
Producer’s accuracy (%) 87.5 0.0 81.3 58.0
OOB error (%), Kappa and run time (seconds) 32.84 0.526 0.74
Table 5. RF Confusion Matrices for Dead, Green, Yellow, and Brown C anopies Using Simulated Feature-S elected WorldView-2 Spectra
Dead canopies Brown canopies Green canopies Yellow canopies User’s accuracy (%)
Dead canopies 12 0 0 3 80.0
Brown canopies 0 1 0 7 12.5
Green canopies 0 0 14 3 82.4
Yellow canopies 2 2 2 21 77.8
Producer’s accuracy (%) 85.7 33.3 87.5 61.8
OOB error (%), Kappa and run time (seconds) 28.36 0.585 0.07
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING March 2015 205