The application of multivariate land classification to vegetation survey in the Wicklow Mountains,Ireland

Multivariate land classification and land cover mapping by aerial photographic interpretation were used to model spatial variation of land cover in the Wicklow Mountains, Ireland and to structure a stratified random sampling programme of upland blanket bog vegetation. The total area of blanket bog with gully-erosion features was estimated as 33% of the area studied. Vegetation with hand peat-cutting patterns was estimated at 5%, and there was 35% undissected (intact) vegetation. There were differences between land classes in the estimated area of land cover with gully-erosion features or hand peat-cutting patterns.Sample vegetation quadrats, stratified by land class and aerial photographic land cover type, were grouped by their plant species composition. The groups represented ombrotrophic mire, soligenous mire and shrub heath vegetation. There was significant association between vegetation group and land class, related to variation in regional landscape type, but no significant association between vegetation group and the aerial photographic land cover types, undissected (intact) and dissected (gullied and cut-over) peats. It is proposed that the similarity of vegetation between undissected and dissected blanket bog is related to vegetation regeneration. The need to consider differences in vegetation distribution, composition and dynamics in ecological management strategies is emphasised. The study demonstrated the value of stratified random field sampling for cost-efficient regional ecological assessment in upland blanket bog landscapes typified by the Wicklow mountains, Ireland.     

Digitized photographs in vegetation analysis ‐ a comparison of cover estimates

Abstract. The use of digital images in measuring plant species cover and species number in boreal forest vegetation was studied. Plant cover was estimated by manual delineation on photographs and subsequent digitized measurement of the areas. This value was regarded as the reference and compared with cover obtained by visual estimate, point frequency and automatic image analysis methods. The automatic image analysis was based on scanned photographs. Supervised image classification with ERDAS software was then used to distinguish between the covers of different plant species. When comparing the ability of the methods to detect species, the visual estimate method gave values similar to the reference. The study material was collected from three sites along a heavy‐metal pollution transect in western Finland. All four methods detected, in a similar manner, differences between the plant species abundances along the transect. Compared with the reference, the digital images underestimated the cover of the lichens Cladina spp. and Cetraria islandica, but gave similar estimates for the dwarf shrubs Empetrum nigrum and Vaccinium vitis‐idaea. The point frequency method overestimated the cover of all the species studied. The visual estimates of lichens were close to the reference, while the dwarf shrub covers were overestimated. The number of species detected using supervised image analysis and the point frequency method was lower than that with visual estimation. Visual estimation was faster, and the estimate closer to the reference cover values than the others. Digital images may be useful in detecting changes in some selected species in vegetation with a simple vertical structure but with taller, multilayered vegetation and a higher species number, the reliability of the cover estimates is lower.     

A Comparison of Remote Sensing and Ground-Based Methods for Monitoring Wetland Restoration Success

Abstract Efficient and accurate vegetation sampling techniques are essential for the assessment of wetland restoration success. Remotely acquired data, used extensively in many locations, have not been widely used to monitor restored wetlands. We compared three different vegetation sampling techniques to determine the accuracy associated with each method when used to determine species composition and cover in restored Pacific coast wetlands dominated by Salicornia virginica (perennial pickleweed). Two ground‐based techniques, using quadrat and line intercept sampling, and a remote sensing technique, using low altitude, high resolution, color and color infrared photographs, were applied to estimate cover in three small restoration sites. The remote technique provided an accurate and efficient means of sampling vegetation cover, but individual species could not be identified, precluding estimates of species density and distribution. Aerial photography was determined to be an effective tool for vegetation monitoring of simple (i.e., single‐species) habitat types or when species identities are not important (e.g., when vegetation is developing on a new restoration site). The efficiency associated with these vegetation sampling techniques was dependent on the scale of the assessment, with aerial photography more efficient than ground‐based sampling methods for assessing large areas. However, the inability of aerial photography to identify individual species, especially mixed‐species stands common in southern California salt marshes, limits its usefulness for monitoring restoration success. A combination of aerial photography and ground‐based methods may be the most effective means of monitoring the success of large wetland restoration projects.     

Computerized classification of Mediterranean vegetation using panchromatic aerial photographs

Abstract. Historical aerial photographs are an important source for data on medium- to long-term (10 - 50 yr) vegetation changes. Older photographs are panchromatic, and manual interpretation has traditionally been used to derive vegetation data from such photographs. We present a method for computerized analysis of panchromatic aerial photographs, which enables one to create high resolution, accurate vegetation maps. Our approach is exemplified using two aerial photographs (from 1964 and 1992) of a test area on Mt. Meron, Israel. Spatial resolution (pixel size) of the geo-rectified photos was 0.30 m and spatial accuracy (RMS error) ca. 1 m. An illumination adjustment prior to classification was found to be essential in reducing misclassification error rates. Two classification approaches were employed: a standard maximum-likelihood supervised classifier, and a modification of a supervised classification, which takes into account spectral properties of individual pixels as well as their neighbourhood characteristics. Accuracy of the maximum likelihood classification was 81 % in the 1992 image and 54 % in the 1964 image. The neighbour classifier increased accuracy to 89 % and 82 % respectively. The overall results suggest that computerized analysis of sequences of panchromatic aerial photographs may serve as a valuable tool for the quantification of medium-term vegetation changes.     

Change in Vegetation Patterns Over a Large Forested Landscape Based on Historical and Contemporary Aerial Photography

Changes to vegetation structure and composition in forests adapted to frequent fire have been well documented. However, little is known about changes to the spatial characteristics of vegetation in these forests. Specifically, patch sizes and detailed information linking vegetation type to specific locations and growing conditions on the landscape are lacking. We used historical and recent aerial imagery to characterize historical vegetation patterns and assess contemporary change from those patterns. We created an orthorectified mosaic of aerial photographs from 1941 covering approximately 100,000 ha in the northern Sierra Nevada. The historical imagery, along with contemporary aerial imagery from 2005, was segmented into homogenous vegetation patches and classified into four relative cover classes using random forests analysis. A generalized linear mixed model was used to compare topographic associations of dense forest cover on the historical and contemporary landscapes. The amount of dense forest cover increased from 30 to 43% from 1941 to 2005, replacing moderate forest cover as the most dominant class. Concurrent with the increase in extent, the area-weighted mean patch size of dense forest cover increased tenfold, indicating greater continuity of dense forest cover and more homogenous vegetation patterns across the contemporary landscape. Historically, dense forest cover was rare on southwesterly aspects, but in the contemporary forest, it was common across a broad range of aspects. Despite the challenges of processing historical air photographs, the unique information they provide on landscape vegetation patterns makes them a valuable source of reference information for forests impacted by past management practices.     

Effects of grazing and topography on long-term vegetation changes in a Mediterranean ecosystem in Israel

The dynamics of Mediterranean vegetation over 28 years was studied in the Northern Galilee Mountains, Israel, in order to identify and quantify the major factors affecting it at the landscape scale. Image analysis of historical and current aerial photographs was used to produce high resolution digital vegetation maps (pixel size = 30 cm) for an area of 4 km² in the Galilee Mountains, northern Israel. GIS tools were used to produce corresponding maps of grazing regime, topographic indices and other relevant environmental factors. The effects of those factors were quantified using a multiple regression analyses. Major changes in the vegetation occurred during the period studied (1964–1992); tree cover increased from 2% in 1964 to 41% in 1992, while herbaceous vegetation cover decreased from 56% in 1964 to 24% in 1992. Grazing, topography and initial vegetation cover were found to significantly affect present vegetation patterns. Both cattle grazing and goat grazing reduced the rate of increase in tree cover, yet even intensive grazing did not halt the process. Grazing affected also the woody-herbaceous vegetation dynamics, reducing the expansion of woody vegetation. Slope, aspect, and the interaction term between these two factors, significantly affected vegetation pattern. Altogether, 56% and 72% of the variability in herbaceous and tree cover, respectively, was explained by the regression models. This study indicates that spatially explicit Mediterranean vegetation dynamics can be predicted with fair accuracy using few biologically important environmental variables.     

Automatic detection of woody vegetation in repeat landscape photographs using a convolutional neural network

Repeat photography is an efficient method for documenting long-term landscape changes. So far, the usage of repeat photographs for quantitative analyses is limited to approaches based on manual classification. In this paper, we demonstrate the application of a convolutional neural network (CNN) for the automatic detection and classification of woody regrowth vegetation in repeat landscape photographs. We also tested if the classification results based on the automatic approach can be used for quantifying changes in woody vegetation cover between image pairs. The CNN was trained with 50 × 50 pixel tiles of woody vegetation and non-woody vegetation. We then tested the classifier on 17 pairs of repeat photographs to assess the model performance on unseen data. Results show that the CNN performed well in differentiating woody vegetation from non-woody vegetation (accuracy = 87.7%), but accuracy varied strongly between individual images. The very similar appearance of woody vegetation and herbaceous species in photographs made this a much more challenging task compared to the classification of vegetation as a single class (accuracy = 95.2%). In this regard, image quality was identified as one important factor influencing classification accuracy. Although the automatic classification provided good individual results on most of the 34 test photographs, change statistics based on the automatic approach deviated from actual changes. Nevertheless, the automatic approach was capable of identifying clear trends in increasing or decreasing woody vegetation in repeat photographs. Generally, the use of repeat photography in landscape monitoring represents a significant added value to other quantitative data retrieved from remote sensing and field measurements. Moreover, these photographs are able to raise awareness on landscape change among policy makers and public as well as they provide clear feedback on the effects of land management.     

Estimating boreal forest ground cover vegetation composition from nadir photographs using deep convolutional neural networks

Ground cover and surface vegetation information are key inputs to wildfire propagation models and are important indicators of ecosystem health. Often these variables are approximated using visual estimation by trained professionals but the results are prone to bias and error. This study analyzed the viability of using nadir or downward photos from smartphones (iPhone 7) to provide quantitative ground cover and biomass loading estimates. Good correlations were found between field measured values and pixel counts from manually segmented photos delineating a pre-defined set of 10 discrete cover types. Although promising, segmenting photos manually was labor intensive and therefore costly. We explored the viability of using a trained deep convolutional neural network (DCNN) to perform image segmentation automatically. The DCNN was able to segment nadir images with 95% accuracy when compared with manually delineated photos. To validate the flexibility and robustness of the automated image segmentation algorithm, we applied it to an independent dataset of nadir photographs captured at a different study site with similar surface vegetation characteristics to the training site with promising results.     

Floodplain vegetation phenology in the Southeast USA: Optimizing the timing of aerial imagery acquisition

Phenological patterns of wetland vegetation were studied on the Savannah River Site, Aiken, South Carolina, USA, to aid in delineation of wetland vegetation community types and plant species from aerial imagery. Relative phenology patterns were recorded for 16 dominant wetland plant species from May 1984 to June 1985 on the floodplains of three streams. These patterns were identified and compared with aerial photographs acquired in the same season as ground observations. Leafout and leaf senescence varied between and within floodplains. In spite of this variability, vegetation classes and individual species may be distinguished best during the late April leafout period.     

Vegetation change in Southeast Greenland? Tasiilaq revisited after 40 years

Questions: Are there changes in species composition of the oceanic, Low‐Arctic tundra vegetation after 40 years? Can possible changes be attributed to climate change? Location: Ammassalik Island near Tasiilaq, Southeast Greenland. Methods: Species composition and cover of 11 key vegetation types were recorded in 110 vegetation survey plots in 1968–1969 and in 11 permanent plots in 1981. Recording was repeated in 2007. Temporal changes in species composition and cover between the surveys were tested using permutation tests linked with constrained ordinations for vegetation types, and Mann–Whitney tests for individual species. Changes in vegetation were related to climate change. Results: Although climate became warmer over the studied period, most of the vegetation types showed minor changes. The changes were most conspicuous in mire and snowbed vegetation, such as the Carex rariflora mire and Hylocomium splendens snowbed. In the C. rariflora mire, species number and cover of vascular plants and cover of bryophytes increased, whereas in the H. splendens snowbed species numbers of vascular plants, bryophytes, and also lichens increased. Lichen richness increased in the Carex bigelowii snowbed and cover of bryophytes in the Salix herbacea snowbed. No such changes occurred in the Alchemilla glomerulans meadow, Alchemilla alpina snowbed and Phyllodoce coerulea heath. There was no change of species composition within the Salix glauca scrub, A. alpina snowbed, lichen grassland and the Empetrum nigrum and Phyllodoce coerulea heaths. Most changes resulted from increasing frequency or cover of some species; there were very few decreasing species. Most of the increasing species indicate drier substrate conditions. Conclusions: Only minor changes in species composition and cover were detected in the vegetation types studied. These changes were probably caused by milder winters and warmer summers during the years before the 2007 sampling. Climate warming may have reduced the duration of snow cover and soil moisture, particularly in snowbed and mire habitats, where species composition change was most pronounced. However, its magnitude was insufficient to cause a major change in species composition. Thus, on the level of plant community types, tundra vegetation near Tasiilaq was rather stable over the last 40 years.     

Automatic segmentation of cell nuclei in Feulgen-stained histological sections of prostate cancer and quantitative evaluation of segmentation results

Digital image analysis of cell nuclei is useful to obtain quantitative information for the diagnosis and prognosis of cancer. However, the lack of a reliable automatic nuclear segmentation is a limiting factor for high-throughput nuclear image analysis. We have developed a method for automatic segmentation of nuclei in Feulgen-stained histological sections of prostate cancer. A local adaptive thresholding with an object perimeter gradient verification step detected the nuclei and was combined with an active contour model that featured an optimized initialization and worked within a restricted region to improve convergence of the segmentation of each nucleus. The method was tested on 30 randomly selected image frames from three cases, comparing the results from the automatic algorithm to a manual delineation of 924 nuclei. The automatic method segmented a few more nuclei compared to the manual method, and about 73% of the manually segmented nuclei were also segmented by the automatic method. For each nucleus segmented both manually and automatically, the accuracy (i.e., agreement with manual delineation) was estimated. The mean segmentation sensitivity/specificity were 95%/96%. The results from the automatic method were not significantly different from the ground truth provided by manual segmentation. This opens the possibility for large-scale nuclear analysis based on automatic segmentation of nuclei in Feulgen-stained histological sections.     

Spatial analysis of oblique photo‐point images for quantifying spatio‐temporal changes in plant communities

Question: Can spatial analytical techniques be used to extract quantitative measurements of vegetation communities from ground‐based permanent photo‐point images? Location: Mount Aspiring National Park, south‐western South Island, New Zealand. Methods: Sets of ground‐based photographs representing two contrasting vegetation types were selected to test two spatial analytical techniques. In the grid technique, a grid was superimposed onto the photographs and the frequency of species presence in each grid‐square was calculated to estimate species abundance/cover over the defined area. In the object‐oriented technique, the photographs were segmented into meaningful objects, based on the colour of the pixels and the textural patterns of the images, and the area occupied by an object in the image was used to derive species abundance/cover over the area. Results: Both techniques allow quick and easy classification of digital elements into ecologically relevant categories of vegetation components. The grid technique appeared more robust, being quick and efficient, accommodating all image types and providing presence/absence matrices for multivariate analysis. Fewer classes were identified using the object‐oriented technique, in particular for the forest interior site and for small individual plants such as Astelia spp. Conclusions: Both techniques showed potential for the objective quantitative analysis of long‐term vegetation monitoring of cover and changes of several component species, using repeat ground‐based photographs more specifically for grassland habitats. However, both rely to various degrees on manual classification. Corrective factors and strict protocols for taking the photographs are necessary to account for variation in view angles and to compute values more representative of absolute species abundance.     

Numeric classification as an aid to spectral mapping of vegetation communities

This study demonstrates a vegetation mapping methodology that relates the reflectance information contained in multispectral imagery to traditionally accepted ecological classifications. Key elements of the approach used are (a) the use of cover rather than density or presence/absence to quantify the vegetation, (b) the inclusion of physical components as well as vegetation cover to describe and classify field sites, (c) development of an objective land cover classification from this quantitative data, (d) use of the field sample sites as training areas for the spectral classification, and (e) the use of a discriminant function to effectively tie the two classifications together. Land cover over 39000 ha of Australian chenopod shrubland was classified into nine groups using agglomerative hierarchical clustering, a discriminant function developed to relate cover and spectral classes, and the vegetation mapped using a maximum likelihood classification of multi-date Landsat TM imagery. The accuracy of the mapping was assessed with an independent set of field samples and by comparison with a map of land systems previously interpreted from aerial photography. Overall agreement between the digital classification and the land system map was good. The units that have been mapped are those derived from numeric vegetation classification, demonstrating that accepted ecological methods and sound image analysis can be successfully combined.     

Can digital image classification be used as a standardised method for surveying peatland vegetation cover?

The ability to carry out systematic, accurate and repeatable vegetation surveys is an essential part of long-term scientific studies into ecosystem biodiversity and functioning. However, current widely used traditional survey techniques such as destructive harvests, pin frame quadrats and visual cover estimates can be very time consuming and are prone to subjective variations. We investigated the use of digital image techniques as an alternative way of recording vegetation cover to plant functional type level on a peatland ecosystem. Using an established plant manipulation experimental site at Moor House NNR (an Environmental Change Network site), we compared visual cover estimates of peatland vegetation with cover estimates using digital image classification methods, from 0.5 m × 0.5 m field plots. Our results show that digital image classification of photographs taken with a standard digital camera can be used successfully to estimate dwarf-shrub and graminoid vegetation cover at a comparable level to field visual cover estimates, although the methods were less effective for lower plants such as mosses and lichens. Our study illustrates the novel application of digital image techniques to provide a new way of measuring and monitoring peatland vegetation to the plant functional group level, which is less vulnerable to surveyor bias than are visual field surveys. Furthermore, as such digital techniques are highly repeatable, we suggest that they have potential for use in long-term monitoring studies, at both plot and landscape scales.     

Estimation of carbon storage based on individual tree detection in <Emphasis Type="Italic">Pinus densiflora</Emphasis> stands using a fusion of aerial photography and LiDAR data

The objective of this study was to estimate the carbon storage capacity of Pinus densiflora stands using remotely sensed data by combining digital aerial photography with light detection and ranging (LiDAR) data. A digital canopy model (DCM), generated from the LiDAR data, was combined with aerial photography for segmenting crowns of individual trees. To eliminate errors in over and under-segmentation, the combined image was smoothed using a Gaussian filtering method. The processed image was then segmented into individual trees using a marker-controlled watershed segmentation method. After measuring the crown area from the segmented individual trees, the individual tree diameter at breast height (DBH) was estimated using a regression function developed from the relationship observed between the field-measured DBH and crown area. The above ground biomass of individual trees could be calculated by an image-derived DBH using a regression function developed by the Korea Forest Research Institute. The carbon storage, based on individual trees, was estimated by simple multiplication using the carbon conversion index (0.5), as suggested in guidelines from the Intergovernmental Panel on Climate Change. The mean carbon storage per individual tree was estimated and then compared with the field-measured value. This study suggested that the biomass and carbon storage in a large forest area can be effectively estimated using aerial photographs and LiDAR data.     

Random Forest characterization of upland vegetation and management burning from aerial imagery

Aim The upland moorlands of Great Britain form distinctive landscapes of international conservation importance, comprising mosaics of heathland, acid grassland, blanket bog and bracken. Much of this landscape is managed by rotational burning to create gamebird habitat and there is concern over whether this is driving long‐term changes in upland vegetation communities. However, the inaccessibility and scale of uplands means that monitoring changes in vegetation and burning practices is difficult. We aim to overcome this problem by developing methods to classify aerial imagery into high‐resolution maps of dominant vegetation cover, including the distribution of burns on managed grouse moors. Location  Peak District National Park, England, UK. Methods Colour and infrared aerial photographs were classified into seven dominant land‐cover classes using the Random Forest ensemble machine learning algorithm. In addition, heather (Calluna vulgaris) was further differentiated into growth phases, including sites that were newly burnt. We then analysed the distributions of the vegetation classes and managed burning using detrended correspondence analysis. Results Classification accuracy was c. 95% and produced a 5‐m resolution map for 514 km² of moorland. Cover classes were highly aggregated and strong nonlinear effects of elevation and slope and weaker effects of aspect and bedrock type were evident in structuring moorland vegetation communities. The classification revealed the spatial distribution of managed burning and suggested that relatively steep areas may be disproportionately burnt. Main conclusions Random Forest classification of aerial imagery is an efficient method for producing high‐resolution maps of upland vegetation. These may be used to monitor long‐term changes in vegetation and management burning and infer species–environment relationships and can therefore provide an important tool for effective conservation at the landscape scale.     

Vegetal cover estimated by digital photographs related to biomass in a grassland site in northern Mexico

A regression model was used to determine the relationship between aerial herbaceous biomass and vegetation coverage estimated by digital images. Four samplings (n=36 each date) of vegetation cover and herbaceous biomass were performed during the growing season in 2011 in a grassland dominated by Bouteloua gracilis in La Cieneguilla, Municipality of Villa Hidalgo, Durango. Average production of dry biomass was 37.36 ± 9.66 g/m², and mean vegetation cover 30.02%. Dry biomass data were tested for normality using the test of Kolmogorov Smirnov, finding a lack of fit. The data were subjected to a logarithmic transformation and the model Ln(y) = 1.637926 + 0.08501X - 0.000586X² with an adjusted R² = 0.89 was found. In order to validate this model, another five samplings were carried out in 2013 at the same site during summer and autumn, using the same sampling size for each date as in 2011. Data collected in 2013 were analyzed with the model Ln (y) = β₀ + β₁X + β₂X². A comparison of regression coefficients was carried out between the 2011 and 2013 models with t (180+144-9-11-2=302, p<0.05) = 1.967. The results indicated that it is possible to use the 2011 regression model to estimate herbaceous aerial biomass from vegetation cover measurements with aerial photographs in La Cieneguilla site during summer and fall.     

Mapping mountain vegetation using species distribution modeling,image‐based texture analysis,and object‐based classification

Objective: The objective of this study was to map vegetation composition across a 24 000 ha watershed. Location: The study was conducted on the western slope of the Sierra Nevada mountain range of California, USA. Methods: Automated image segmentation was used to delineate image objects representing vegetation patches of similar physiognomy and structure. Image objects were classified using a decision tree and data sources extracted from individual species distribution models, Landsat spectral data, and life form cover estimates derived from aerial image‐based texture variables. Results: A total of 12 plant communities were mapped with an overall accuracy of 75% and a χ‐value of 0.69. Species distribution model inputs improved map accuracy by approximately 15% over maps derived solely from image data. Automated mapping of existing vegetation distributions, based solely on predictive distribution model results, proved to be more accurate than mapping based on Landsat data, and equivalent in accuracy to mapping based on all image data sources. Conclusions: Results highlight the importance of terrain, edaphic, and bioclimatic variables when mapping vegetation communities in complex terrain. Mapping errors stemmed from the lack of spectral discernability between vegetation classes, and the inability to account for the confounding effects of land use history and disturbance within a static distribution modeling framework.     

Agricultural Legacies in the Great Basin Alter Vegetation Cover, Composition, and Response to Precipitation

The land-use history of an ecosystem influences current structure and possibly response to modern disturbances and stresses. In semiarid systems the nature of land-use legacies is poorly understood, confounding efforts to establish sustainable management approaches. We compare previously cultivated and non-cultivated lands in Owens Valley, California, where cultivation once extended to 34% of the valley floor but was largely discontinued by 1940, to measure the influence of past disturbance on modern vegetation. We combined historic maps of cultivated and non-cultivated land with an extensive vegetation survey, historic aerial photographs, and satellite measurements of vegetation response to precipitation variability to examine the importance of land-use history in determining the sensitivity of vegetation to annual variations in precipitation. Remote sensing analysis showed that total plant cover on previously cultivated lands was lower and fluctuations in cover were marginally more dependent on precipitation compared with plant cover on non-cultivated lands. We then compared modern plant assemblages within cultivated and non-cultivated land to determine if compositional differences could explain the current patterns of vegetation cover. We found lower species richness on previously cultivated parcels, and higher frequency and cover of perennial grasses on non-cultivated lands. Therefore, we showed differences in land-cover patterns, isolated a mechanism that could account for the differences (species differences), and developed a method for remotely analyzing land regions that have experienced historic anthropogenic disturbance.