Ruiliang Pu

Conifer species recognition: An exploratory analysis of in situ hyperspectral data

In situ hyperspectral data measured above sunlit and shaded sides of canopies using a high spectral resolution radiometer were analyzed for identification of six conifer tree species. An artificial neural network algorithm was assessed for the identification purpose. Linear discrimination analysis was compared with the neural network algorithm. The hyperspectral with the neural data were further processed to smoothed reflectance and first derivative spectra and were separately used in tree species identification. Tree species recognition with data collected front six study sites was tested in seven experiments. The average accuracy of species recognition was obtained at every site. The overall performance of the neural network algorithm was better than that of linear discriminant analysis for species recognition when the same number of training samples and test samples were used. The discriminant analysis produced better accuracy than neural network at one site where many samples (10) were taken from six individual trees. Use of the average spectra of all samples for a particular tree species in training may not result in higher accuracy than use of individual spectral samples in training. Use of sunlit samples alone resulted in an overall accuracy of greater than 91%. The effects of site background including illuminating conditions on tree species specra were large. Neural networks are sensitive to subtle spectral details and can be trained to separate samples front the same species at different sites. Our experiments indicate that the discriminating power of visible bands is stronger than that of near-infrared bands. Higher recognition accuracies can be obtained in the blue to green or the red-edge spectral region as compared with four other spectral regions. A smaller set of selected bands can generate more accurate identification than all spectral bands.

Seasonal Patterns and Remote Spectral Estimation of Canopy Chemistry Across the Oregon Transect

We examined seasonal changes in canopy chemical concentrations and content in conifer forests growing along a climate gradient in western Oregon, as part of the Oregon Transect Ecosystem Research (OTTER) study. The chemical variables were related to seasonal patterns of growth and production. Statistical comparisons of chemical variables with data collected from two different airborne remote-sensing platforms were also carried out. Total nitrogen (N) concentrations in foliage varied significantly both seasonally and among sites; when expressed as content in the forest canopy, nitrogen varied to a much greater extent and was significantly related to aboveground net primary production (r = 0.99). Chlorophyll and free amino acid concentrations varied more strongly than did total N and may have reflected changes in physiological demands for N. Large variations in starch concentrations were measured from pre- to post-budbreak in all conifer sites. Examination of remote-sensing data from two different airborne instruments suggests the potential for remote measurement of some canopy chemicals. Multivariate analysis of high-resolution spectral data in the near infrared region indicated significant correlations between spectral signals and N concentration and canopy N content; the correlation with canopy N content was stronger and was probably associated in part with water absorption features of the forest canopy. The spectral bands that were significantly correlated with lignin concentration and content were similar to bands selected in the other laboratory and airborne studies; starch concentrations were not significantly related to spectral reflectance data. Strong relationships between the spectral position of specific reflectance features in the visible region and chlorophyll were also found.

Extraction of red edge optical parameters from Hyperion data for estimation of forest leaf area index

A correlation analysis was conducted between forest leaf area index (LAI) and two red edge parameters: red edge position (REP) and red well position (RWP), extracted from reflectance image retrieved from Hyperion data. Field spectrometer data and LAI measurements were collected within two days after the Earth Observing One satellite passed over the study site in the Patagonia region of Argentina. The two red edge parameters were extracted with four approaches: four-point interpolation, polynomial fitting, Lagrangian technique, and inverted-Gaussian (IG) modeling. Experimental results indicate that the four-point approach is the most practical and suitable method for extracting the two red edge parameters from Hyperion data because only four bands and a simple interpolation computation are needed. The polynomial fitting approach is a direct method and has its practical value if hyperspectral data are available. However, it requires more computation time. The Lagrangian method is applicable only if the first derivative spectra are available; thus, it is not suitable to multispectral remote sensing. The IG approach needs further testing and refinement for Hyperion data.

Spectral absorption features as indicators of water status in coast live oak (Quercus agrifolia) leaves

A total of 139 reflectance spectra (between 350 and 2500 nm) from coast live oak ( Quercus agrifolia ) leaves were measured in the laboratory with a spectrometer FieldSpec®Pro FR. Correlation analysis was conducted between absorption features, three-band ratio indices derived from the spectra and corresponding relative water content (RWC, %) of oak leaves. The experimental results indicate that there exist linear relationships between the RWC of oak leaves and absorption feature parameters: wavelength position (WAVE), absorption feature depth (DEP), width (WID) and the multiplication of DEP and WID (AREA) at the 975 nm, 1200 nm and 1750 nm positions and two three-band ratio indices: RATIO 975 and RATIO 1200, derived at 975 nm and 1200 nm. AREA has a higher and more stable correlation with RWC compared to other features. It is worthy of noting that the two three-band ratio indices, RATIO 975 and RATIO 1200, may have potential application in assessing water status in vegetation.

Object-based urban detailed land cover classification with high spatial resolution IKONOS imagery

Improvement in remote sensing techniques in spatial/spectral resolution strengthens their applicability for urban environmental study. Unfortunately, high spatial resolution imagery also increases internal variability in land cover units and can cause a ‘salt-and-pepper’ effect, resulting in decreased accuracy using pixel-based classification results. Region-based classification techniques, using an image object (IO) rather than a pixel as a classification unit, appear to hold promise as a method for overcoming this problem. Using IKONOS high spatial resolution imagery, we examined whether the IO technique could significantly improve classification accuracy compared to the pixel-based method when applied to urban land cover mapping in Tampa Bay, FL, USA. We further compared the performance of an artificial neural network (ANN) and a minimum distance classifier (MDC) in urban detailed land cover classification and evaluated whether the classification accuracy was affected by the number of extracted IO features. Our analysis methods included IKONOS image data calibration, data fusion with the pansharpening (PS) process, Hue–Intensity–Saturation (HIS) transferred indices and textural feature extraction, and feature selection using a stepwise discriminant analysis (SDA). The classification results were evaluated with visually interpreted data from high-resolution (0.3 m) digital aerial photographs. Our results indicate a statistically significant difference in classification accuracy between pixel- and object-based techniques; ANN outperforms MDC as an object-based classifier; and the use of more features (27 vs. 9 features) increases the IO classification accuracy, although the increase is statistically significant for the MDC but not for the ANN.

Penalized discriminant analysis of in situ hyperspectral data for conifer species recognition

Using in situ hyperspectral measurements collected in the Sierra Nevada Mountains in California, the authors discriminate six species of conifer trees using a recent, nonparametric statistics technique known as penalized discriminant analysis (PDA). A classification accuracy of 76% is obtained. Their emphasis is on providing an intuitive, geometric description of PDA that makes the advantages of penalization clear. PDA is a penalized version of Fishers linear discriminant analysis (LDA) and can greatly improve upon LDA when there are a large number of highly correlated variables.

Broadleaf species recognition with in situ hyperspectral data

Timely and accurate identification of tree species by spectral methods is crucial for forest and urban ecological management. In this study, a total of 394 reflectance spectra (between 350 and 2500 nm) from foliage branches or canopy of 11 important urban forest broadleaf species were measured in the City of Tampa, Florida, USA with a spectrometer. The 11 species include American elm (Ulmus americana), bluejack oak (Quercus incana), crape myrtle (Lagerstroemia indica), laurel oak (Q. laurifolia), live oak (Q. virginiana), southern magnolia (Magnolia grandiflora), persimmon (Diospyros virginiana), red maple (Acer rubrum), sand live oak (Q. geminata), American sycamore (Platanus occidentalis), and turkey oak (Q. laevis). A total of 46 spectral variables, including normalized spectra, derivative spectra, spectral vegetation indices, spectral position variables, and spectral absorption features were extracted and analysed from the in situ hyperspectral measurements. Two classification algorithms were used to identify the 11 broadleaf species: a nonlinear artificial neural network (ANN) and a linear discriminant analysis (LDA). An analysis of variance (ANOVA) indicates that the 30 selected spectral variables are effective to differentiate the 11 species. The 30 selected spectral variables account for water absorption features at 970, 1200, and 1750 nm and reflect characteristics of pigments and other biochemicals in tree leaves, especially variability of chlorophyll content in leaves. The experimental results indicate that both classification algorithms (ANN and LDA) have produced acceptable accuracies (overall accuracy from 86.3% to 87.8%, kappa from 0.83 to 0.87) and have a similar performance for classifying the 11 broadleaf species with input of the 30 selected spectral variables. The preliminary results of identifying the 11 species with the in situ hyperspectral data imply that with current remote sensing techniques, including high spatial and spectral resolution data, it is still difficult but possible to identify similar species to such 11 broadleaf species with an acceptable accuracy.Timely and accurate identification of tree species by spectral methods is crucial for forest and urban ecological management. In this study, a total of 394 reflectance spectra (between 350 and 2500...

Spectroscopic determination of wheat water status using 1650-1850 nm spectral absorption features

Wheat leaves were measured radiometrically in order to spectrally characterize the water deficiency symptoms. In this study, a FieldSpec-FR was used for measuring wheat leaf spectra. After the spectral analysis using a spectral normalizing technique, the spectral absorption feature parameters: wavelength position (nm), depth and area (relative value) were extracted from each wheat leaf spectrum. The relative water content (RWC) was measured for each wheat leaf sample. A linear regression analysis was conducted between the spectral absorption feature parameters and corresponding RWCs. The experimental results from 110 samples indicated that reflectance spectra of wheat leaves in the 1650-1850 nm region were dominated by water content. With a decrease in wheat leaf RWC, the 1650-1850 nm spectral absorption features gradually become obvious. The relative errors of predicted RWCs and the absolute error of predicted wavelength positions were calculated from 12 validation samples by established regression equations. The relative errors of predicted RWCs and the absolute error of predicted wavelength position (nm) were both low (<6% for RWCs by the depth and area and <12 nm for the wavelength position, respectively). Furthermore, we discuss the potential and limitations of spectroscopic determination of wheat RWC by using remote sensing technology.

Using classification and NDVI differencing methods for monitoring sparse vegetation coverage: a case study of saltcedar in Nevada, USA

A change detection experiment for an invasive species, saltcedar, near Lovelock, Nevada, was conducted with multidate Compact Airborne Spectrographic Imager (CASI) hyperspectral datasets. Classification and NDVI differencing change detection methods were tested. In the classification strategy, a principal component analysis (PCA) was performed on single‐date CASI imagery separately in the visible bands and NIR bands. Then the first five PCs from the visible bands and the first five PCs from the NIR bands were used to classify six to eight cover types with a maximum likelihood classifier. A complete matrix of change information and change/no‐change maps were produced by overlaying two single‐date classification maps. In the NDVI differencing strategy, a linear regression model was developed between two Normalized Difference Vegetation Index (NDVI) images to normalize the index differences caused by factors not related to land cover change. Then the actual time 2 NDVI image was subtracted by the predicted time 2 NDVI image to obtain the differencing image. The NDVI differencing image was further processed with a new threshold method into change/no‐change of saltcedar. By testing the single‐date classification results and validating the change/no‐change results, both change detection results indicated that CASI hyperspectral data have the potential to map and monitor the change of saltcedar. However, the accuracy assessment and change/no‐change validation results (overall accuracy 91.56% and kappa value 0.618 for the classification method against corresponding values of 93.04% and 0.684 for the NDVI differencing method) indicate that the NDVI differencing method outperformed the classification method in this particular study. In addition, use of the new method of determining thresholds for differentiating change pixels from no‐change pixels from the NDVI differencing image improved the change detection accuracy compared to a traditional method (kappa value increased from 0.813 to 0.888 from a test sample). Therefore, according to the criteria of higher accuracy of change/no‐change maps and fewer spectral bands, the NDVI differencing method is recommended for use if a suitable spectral normalization between multi‐temporal images can be carried out before performing image differencing.

Crown Closure Estimation of Oak Savannah in a Dry Season with Landsat TM imagery: Comparison of Various Indices Through Correlation Analysis

In this paper, we assess the capability of Landsat Thematic Mapper (TM) for oakwood crown closure estimation in Tulare County, California. Measurements made from orthorectified aerial photographs for the same area were used as a reference. The linear relationship between crown closure and digital values of each band of the TM image was examined. TM Band 3 had the highest correlation ( @ = m 0.828; R 2 = 0.687) with crown closure measurements. The simple ratio (SR) and the normalized difference vegetation index (NDVI) were generated for correlation analysis and only NDVI showed better correlation ( A = 0.836; R 2 = 0.699) than use of single bands. An additional index (NIR N - R N )/(NIR N + R N ), called NDVIN, was experimented, NDVISQ ( N = 2) and NDVICUB ( N = 3) showed some improvements over SR and NDVI ( A = 0.855; R 2 = 0.732 for N = 3). Through multiple regression with all six bands, we found that there was a considerable amount of improvement in variability explanation over any individual band or index tested ( R 2 = 0.803). NIR, red and blue bands were able to adequately model crown closure as using all the six TM bands ( R 2 = 0.802). Principal component analysis (PCA) and Kauth-Thomas (K-T) transform were applied to reduce multi-collinearity among bands. The third principal component and greenness in K-T transform showed similar effects to those of NDVI. Transformation of digital numbers (DNs) to radiances kept the results of single band and multiple band estimation the same, and did not improve the index estimation very much. A simple radiometric correction of the TM image improved results for the NDVI ( A = 0.840; R 2 = 0.705) and NDVISQ estimation ( A = 0.861; R 2 = 0.741), but worsened estimation results of single band and multiple bands.