Yuhong He

Studying mixed grassland ecosystems I: suitable hyperspectral vegetation indices

Hyperspectral remote sensing data with a greater number of bands and narrower bandwidths can be effectively exploited for the study of ecosystem patterns and processes. Hyperspectral remote sensing of semiarid mixed grassland faces the following two challenges, however: (i) providing a good understanding of the performance of different vegetation indices (VIs) in estimating biophysical properties of grassland with a small amount of green vegetation, a large amount of dead material on the ground, and variable soil–ground conditions; and (ii) examining the spatial characterization of hyperspectral remotely sensed data to optimize sampling procedures and address scaling issues. Using ground-based hyperspectral and biophysical data, this study has compared the predictive capability of VIs for estimation of grassland leaf area index (LAI) (this paper) and examined the spatial variation of grassland LAI (the companion paper). The results in this paper indicate that the relationships between grassland LAI and VIs are significant. The performance of the renormalized difference vegetation index (RDVI), adjusted transformed soil-adjusted vegetation index (ATSAVI), and modified chlorophyll absorption ratio index 2 (MCARI2) was slightly better than that of the other VIs in the groups of ratio-based, soil-line-related, and chlorophyll-corrected VIs, respectively. By incorporating the cellulose absorption index (CAI) as a litter factor in ATSAVI, a new VI was computed (L-ATSAVI), and it improved the LAI estimation capability in our study area by about 10%.

Detecting grassland spatial variation by a wavelet approach

Insight into the spatial variation of an ecosystem can provide better understanding of ecological processes and patterns in different scales. Detecting these multiple scales of spatial variation in grassland landscapes is valuable for determining management options, designing proper sampling regimes, and selecting suitable resolutions of remote sensing products. The objective of this study is to examine how environmental factors affect spatial variation of biophysical properties in mixed grassland ecosystems. Field leaf area index (LAI), soil moisture, and topographical parameters (relative elevation, upslope length, and a wetness index) were obtained in three parallel transects of a grassland ecosystem in Saskatchewan, Canada in 2004. One 20‐m resolution SPOT 4 (HRVIR) image was acquired at the same period of the growing season but in the following year. Normalized difference vegetation index (NDVI) was calculated from the satellite image of the centre 381‐m transect and two extensive 2560‐m perpendicular transects. A wavelet approach was used to identify the scales of variations. Statistical results showed that LAI is significantly correlated to the wetness index (r2 = 0.37) and soil moisture (r2 = 0.43). The wetness index is better than relative elevation and upslope length in demonstrating the effect of topography on grassland vegetation. The variation of soil moisture is significant at two small scales of about 20 m and 40 m, and that of the wetness index is at the large scale of 120 m. The variation of grassland LAI is significant at three scales (20 m, 40 m, and 120 m), which indicates that the spatial variation of LAI might be controlled by both topography and soil moisture, though the 120 m is the dominant scale of variation in LAI. NDVI significantly correlated with grassland LAI along the centre transect. The effect of topography on grassland LAI is also proven by the significant relationships between NDVI and the wetness index. The wavelet analysis identifies the variation of two extensive transects at the scale of about 120 m, which is similar to the dominant variation scale of grassland LAI. These results confirmed that the effect of topography on spatial variation can be identified from the appropriate satellite image. This study suggests that the spatial scales of soil and topographic data aid in the selection of appropriate satellite image resolution for monitoring and managing ecosystems.

Problems in remote sensing of landscapes and habitats

Wildlife habitat mapping strongly supports applications in natural resource management, environmental conservation, impacts of anthropogenic activity, perturbed ecosystem restoration, species-at-risk recovery and species inventory. Remote sensing has long been identified as a feasible and effective technology for large-area habitat mapping. However, existing and future uncertainties in remote sensing will definitely have a significant effect on the relevant scientific research. This article attempts to identify the current challenges and opportunities in remote sensing for large-area wildlife habitat mapping, and accordingly provide possible solutions and directions for further research.

Comparison of different methods for measuring leaf area index in a mixed grassland

Available LAI instruments have greatly increased our ability to estimate leaf area index (LAI) non-destructively. However, it is difficult to infer from existing studies which instrument has the advantages in measuring LAI over other instruments for grassland ecosystems. The objective of our study was to compare the LAI estimates by two instruments (AccuPAR, and LAI2000), and correlate the LAI measurements to remote sensing data for a mixed grassland. Leaf area index of four grass communities was measured by both the destructive method and instruments. Ground canopy reflectance was measured and further calculated to be LAI-related vegetation indices. Statistical analysis showed that destructively sampled LAI ranged from 0.61 to 5.7 in the study area. Both instruments underestimated LAI in comparison with the destructive method. However, the LAI2000 is better than AccuPAR for estimating LAI. Comparison of four grass communities indicated that the lower the grass LAI, the greater the underestimated percenta...

Studying mixed grassland ecosystems II: optimum pixel size

It was determined in a companion paper that the litter-corrected adjusted transformed soil-adjusted vegetation index (L-ATSAVI) was the best leaf area index (LAI) indicator in a mixed grassland ecosystem. To optimize the sampling procedures and address the scaling issues for the mixed grass ecosystem, this study examined the dominant scale of spatial variation in both LAI and L-ATSAVI using two methods, namely Mexican hat wavelet analysis and semivariogram analysis. The results showed that both methods can identify grassland spatial variation, and the cyclicity (the nature repetition in a dataset) of grassland LAI was about 140 m along the central transect of five parallel transects within the study area. The advantage of wavelet analysis over semivariogram analysis for spatial pattern interpretation was that it could identify the exact location of the transition. The wavelet analysis demonstrated that the cyclicity of L-ATSAVI also corresponded well with features of grassland LAI along the transect. Therefore, following the sampling theorems, a pixel size of less than 35 m will retain most of the spatial variation of grassland LAI in our study area. In terms of this optimum pixel size, the scale of ground-based hyperspectral data and LAI along the transect was simulated using a low-pass filtering procedure with a 30 m moving window. Statistical analysis indicated that scale-simulated L-ATSAVI could significantly explain more grassland LAI (r2 up to 89%) than the original 3 m resolution. This conclusion can be further applied to select the optimal pixel size of remote sensing images and detect the hierarchical characteristics in a grassland landscape.

Scaling up Semi-Arid Grassland Biochemical Content from the Leaf to the Canopy Level: Challenges and Opportunities

Remote sensing imagery is being used intensively to estimate the biochemical content of vegetation (e.g., chlorophyll, nitrogen, and lignin) at the leaf level. As a result of our need for vegetation biochemical information and our increasing ability to obtain canopy spectral data, a few techniques have been explored to scale leaf-level biochemical content to the canopy level for forests and crops. However, due to the contribution of non-green materials (i.e., standing dead litter, rock, and bare soil) from canopy spectra in semi-arid grasslands, it is difficult to obtain information about grassland biochemical content from remote sensing data at the canopy level. This paper summarizes available methods used to scale biochemical information from the leaf level to the canopy level and groups these methods into three categories: direct extrapolation, canopy-integrated approach, and inversion of physical models. As for semi-arid heterogeneous grasslands, we conclude that all methods are useful, but none are ideal. It is recommended that future research should explore a systematic upscaling framework which combines spatial pattern analysis, canopy-integrated approach, and modeling methods to retrieve vegetation biochemical content at the canopy level.

An Automated Method to Parameterize Segmentation Scale by Enhancing Intrasegment Homogeneity and Intersegment Heterogeneity

Image segmentation is a key step in geographic object-based image analysis. Numerous segmentation techniques, e.g., watershed segmentation, mean-shift segmentation, and fractal net evolution algorithm, have been proposed and applied for various types of image analysis tasks. The majority of the segmentation algorithms require a user-defined parameter, namely, the scale parameter, to control the sizes of segments, yet the automation of the scale parameter remains a great challenge. Over the past few years, several automated parameterization methods, such as the estimation of scale parameters (ESP) tool, have been developed. However, few of the existing methods are able to enhance both intrasegment homogeneity and intersegment heterogeneity. In this letter, we proposed an energy function method that aimed at enhancing the characteristics of intrasegment homogeneity and intersegment heterogeneity, simultaneously, to identify the optimal segmentation scale for image segmentation. The intersegment heterogeneity was calculated as the weighted gradient from a segment to its neighbors by spectral angle, whereas the intrasegment homogeneity was quantified by the mean spectral angle within a segment. The performance of the proposed method was evaluated by applying it to a WorldView-2 multispectral image of Toronto, Canada, and comparing it with the local-peak-based method, which considered only the intrasegment homogeneity of an image. The scale parameter identified by the proposed method can better characterize the reference geo-objects over the entire image. The accuracy assessment result shows that the proposed method outperformed the existing technique by reducing the discrepancy by 17.9%.

Reflectance measures of grassland biophysical structure

The goal of this study is to develop an efficient method to retrieve vegetation biophysical properties based on ground LAI measurements and satellite data, and thus avoid the labour‐intensive and time‐consuming process for collecting biomass and canopy height in the future. The field data was conducted in Grasslands National Park (GNP), Saskatchewan, Canada. The two vegetation indices, ATSAVI and RDVI, were derived from SPOT 4 HRV images to estimate LAI and to prepare LAI and biophysical maps for the GNP. The results demonstrated strong relationships between LAI and selected vegetation indices. However, a detailed accuracy assessment indicated that ATSAVI was likely to be better in estimating and mapping LAI than the RDVI. The accuracy of the LAI map was calculated to be 66.7%. The significant relationship between measured LAI and the biophysical data solves the difficulty for mapping biophysical information due to insufficient sampling coverage for GNP.

Automated mapping of impervious surfaces in urban and suburban areas: Linear spectral unmixing of high spatial resolution imagery

Abstract Quantifying impervious surfaces in urban and suburban areas is a key step toward a sustainable urban planning and management strategy. With the availability of fine-scale remote sensing imagery, automated mapping of impervious surfaces has attracted growing attention. However, the vast majority of existing studies have selected pixel-based and object-based methods for impervious surface mapping, with few adopting sub-pixel analysis of high spatial resolution imagery. This research makes use of a vegetation-bright impervious-dark impervious linear spectral mixture model to characterize urban and suburban surface components. A WorldView-3 image acquired on May 9th, 2015 is analyzed for its potential in automated unmixing of meaningful surface materials for two urban subsets and one suburban subset in Toronto, ON, Canada. Given the wide distribution of shadows in urban areas, the linear spectral unmixing is implemented in non-shadowed and shadowed areas separately for the two urban subsets. The results indicate that the accuracy of impervious surface mapping in suburban areas reaches up to 86.99%, much higher than the accuracies in urban areas (80.03% and 79.67%). Despite its merits in mapping accuracy and automation, the application of our proposed vegetation-bright impervious-dark impervious model to map impervious surfaces is limited due to the absence of soil component. To further extend the operational transferability of our proposed method, especially for the areas where plenty of bare soils exist during urbanization or reclamation, it is still of great necessity to mask out bare soils by automated classification prior to the implementation of linear spectral unmixing.

A multi-band watershed segmentation method for individual tree crown delineation from high resolution multispectral aerial image

A comprehensive forest resource inventory needs more detailed species information at individual tree level. Although conventional ground-based measurement fails to achieve this target in an efficient way, the emergence of high resolution remote sensing images has made it possible in the past decade. Individual tree crown delineation is one of the most critical steps for tree species classification from remote sensing images. However, it is still challenging to delineate individual tree crowns in deciduous forests because of the continuous canopy. In this study, a multi-band watershed segmentation method is proposed to delineate deciduous tree crowns by constructing a spectral angle space. The proposed algorithm is further examined by a high resolution multispectral aerial image of a deciduous forested area in Haliburton Forest, Ontario, Canada. Results demonstrate that, the proposed multi-band watershed segmentation method outperforms the existing valley-following based ITC map, in terms of visual interpretation and quantitative evaluation.