Landsat TM Satellite Research Articles

Remote sensing of grassland–shrubland vegetation water content in the shortwave domain

This study compares the ability of spectral approaches operating in the shortwave optical domain to predict absolute and relative vegetation water content (AWC and RWC, respectively) across northern prairie grassland–shrubland. We collected vegetation water content and spectral radiometer data over plots of comparable ground resolution (0.5m) at seven field sites in the Canadian mixed grass prairie in June 2004. We then aggregated observations to scale these data “up” to an observational scale consistent with that of Landsat-TM satellite imagery (30m). This allowed us to assess abilities of three spectral approaches to predict AWC and RWC at both observational scales. These approaches were: individual vegetation indices, a combination of spectral bands and a combination of spectral derivatives. Our results showed that (a) the band-combination approach provides the most accurate and precise estimates of AWC and RWC at both 0.5 and 30m sampling resolutions; (b) the combination of bands providing the greatest predictive abilities are those that emphasize the contrast in reflectance between the NIR and SWIR spectral regions; (c) the band-combination approach predicts AWC with much greater accuracy and precision than RWC and (d) the predictive ability of the band-combination approach decreases only slightly when plot-level data are aggregated to a 30m sampling resolution. These results are generally consistent with the results of other studies and with theory. While our results suggest that simple spectral methods (e.g. linear band-combinations or indices) are good predictors of AWC over grazed and ungrazed grassland–shrubland landscapes at plot- and Landsat spatial resolutions, they are less encouraging for the estimation of RWC. Despite their good predictive abilities, the temporal and geographical portabilities of the spectral approaches for estimating AWC must be further assessed before they can be considered reliable and robust predictive tools. Thus, the further testing of these techniques over larger geographical extents is required.

Assessment of land‐use impact on streamflow via a grid‐based modelling approach including paddy fields

AbstractTo investigate the hydrologic impacts of land‐use changes on streamflow for an urbanizing watershed with paddy fields, a grid‐based daily hydrologic model was adopted. The model was calibrated with two years (2000–2001) of observed streamflow data and validated using 5 months (2001) of measured soil moisture data and 1 year (2002) of observed streamflow data. After the model was tested, it was run to estimate impacts of urbanization on each hydrologic component with the land‐use data sets for 1986, 1994, and 2002 generated from Landsat TM satellite images. Total runoff increased from 41·4% (1986) to 44·9% (2002) for a 5·4% increase in urban area, implying that direct runoff increase exceeded baseflow decrease. Urbanization affected the proportions of direct runoff for each land‐use category more than the change in total runoff and the ratio of direct runoff to total runoff. The change proportions in direct runoff for urban areas, paddy fields, and forest were 14·3%, −9·8% and −6·7% respectively for a 5·4% increase, 4·6% decrease and 3·4% decrease in each land‐use area respectively. The results indicate that paddy fields play an important role in runoff regulation, and the evaluation method can assist regional policy makers in developing land management strategies that minimize hydrologic impacts on streamflow. Copyright © 2005 John Wiley & Sons, Ltd.

Land Use/Cover Changes, the Environment and Water Resources in Northeast China

Land use/cover in Northeast China went through extensive changes during the 1990s. This report explores the interaction between these changes and the environment, and the implication of these changes for rational allocation of water resources. Two maps of land use/cover produced from 1990 and 2000 Landsat TM satellite images were overlaid in Arc Info to reveal changes in land cover. Results indicate that farmland and grassland decreased by 386,195 and 140,075 ha, respectively, while water, built-up areas, and woodland increased by 238,596, 194,231, and 192,682 ha, respectively. These changes bore a mutual relationship with the environmental change. On the one hand, climate warming made some of these changes (e.g., conversion of woodland and grassland to farmland) possible. On the other hand, the changed surface cover modified the local climate. These changes, in turn, caused severe environmental degradation and increased flooding. The change between dry field and rice paddy, in particular, raised severe implications for the proper allocation of limited water resources in the Northeast. Efforts are needed to coordinate their rational allocation to reap maximum and sustainable return over the entire area, not just in some localities. Results obtained in this study should be of interest to the international audience of Environmental Management in that they highlight the interactive nature of human activities and the environment and the off-site impact of these activities on the environment.

Remote Sensing Assessment of Paspalum quadrifarium Grasslands in the Flooding Pampa, Argentina

The knowledge of the distribution, area, and current conservation status of relict natural grasslands dominated by the tall-tussock grass Paspalum quadrifarium Lam. (“pajonal”) in the Flooding Pampa (Argentina) is relevant for the identification of conservation sites and sustainable management and land-use planning. Since European settlement, vast areas of pajonal were converted to croplands and short-grass prairies. The only available vegetation map of these grasslands was made in the mid-20th century. We evaluated 2 methods of land-cover classification (supervised and unsupervised) using a Landsat TM satellite image over an area of 2 258.21 km2 in Ayacucho county, where pajonal still persists as an important ecosystem. At the paddock scale, this grassland community presents a complex structure in which the pajonal is not a pure category but a mosaic of tall and short grasses. Six categories of land cover were adopted (crops, sown pastures, short grasses, pajonal, wetlands, and urban areas). A very good overall accuracy was obtained for both classifications (86.9% and 87.9% for supervised and unsupervised classifications, respectively). However, both producer's and user's accuracies for the pajonal and short grasses were better for the unsupervised classification than for the supervised classification. The pajonal class occupied only 20% of the study area with patch size ranging between 0.09 and 1 653 ha. This work suggests an important replacement of tall-tussock grass by short-grass matrix, which represents noticeable structural and functional changes. The unsupervised classification of Landsat image seems a particularly suitable method for mapping complex vegetation units like the highly fragmented pajonal of the Flooding Pampa and should be an important tool for management and tracking future changes.

Estimation of Potential Soil Erosion for River Perkerra Catchment in Kenya

River Perkerra catchment with an area of 1207 km2 is drained by River Perkerra, which is one of the rivers flowing into Lake Baringo whose drainage area is 6820 km2. The lake is in a semi-arid area of Kenya. Its depth has reduced from 8 m in 1972 to 2.5 m in 2003 due to siltation resulting from high erosion rates in the catchment. The entire catchment is characterised by very steep slopes on the hillsides and gentle slopes in the middle and lower reaches where the surface is bare with very little undergrowth. Interventions to control soil erosion in this fragile ecosystem have been limited partly because of lack of data on erosion and its spatial distribution. In the present study, Universal Soil Loss Equation (USLE) was used in conjunction with GIS Arc/Info and Integrated Land and Water Information Systems (ILWIS) to estimate potential soil loss from River Perkerra catchment. Various physical parameters of the equation were derived by analysing spatial data and processing Landsat TM satellite imagery of the catchment. The estimated potential soil erosion from the catchment was 1.73 million tonnes/year while the sediment yield at the catchment outlet was found to be 1.47 million tonnes/year. The sediment delivery ratio derived using an empirical equation was 0.83. This figure indicates that a higher proportion of sediments generated in the catchment is delivered at the outlet. The use of GIS enabled the results of erosion potential to be mapped back onto the catchment. This is useful in identifying priority areas that require urgent management interventions in controlling soil erosion.

The Evolution of a WILDLAND Forest FIRE FRONT

The rate of the spread and shape of a forest fire front is a problem that has not been thoughtfully studied from a computer graphics perspective. Here, using physically based computer graphics modeling, we propose a model for the simulation of wildland fires over 3D complex terrain. The model is based on conservation laws of energy and species, which includes radiation convection, reaction and natural convection, and takes into account the endothermic and exothermic phases of this kind of phenomenon. As an application, a simulation of a wildland fire in the Ebro basin of Spain is presented. The results are visualized on synthetic imagery, obtained by using the digital model of the studied terrain plus its corresponding images acquired by the Spot 4 and LandSat TM satellites.

Incorporating spatial analysis of habitat into spiny lobster (Panulirus argus) stock assessment at Alacranes reef, Yucatan, México

In this paper, spatial analysis of submerged habitats was incorporated into spiny lobster (Panulirus argus) stock assessment at Alacranes reef, Yucatan, México. Two sources of information were used: a thematic map of submerged habitats obtained using a Landsat TM satellite image and “geo-referenced” field data for lobster catches obtained during July 1998 and February 1999. Geographical information systems (GIS) tools were used as the major basis for displaying, analysing and relating lobster catches data and the thematic map information. Estimations of lobster abundance, density and biomass per habitat class and for the whole reef, at the beginning and end of the fishing season, were obtained. Utilizing Monte Carlo modelling, uncertainty was incorporated to estimations of transects initial and final position, appreciation of area and probability to detect lobsters. Results were compared against commercial catches reported for the previous year. The distribution of fishing effort is also described.

SHAPING THE LANDSCAPE: FIRE–GRAZER INTERACTIONS IN AN AFRICAN SAVANNA

The effects of fire–grazing interactions on grass communities are difficult to identify because fire and grazing influence each other on a landscape scale. Persistent heavy grazing can prevent the spread of fire by breaking up the grass layer. In contrast, frequent burning might inhibit the persistence of grazed patches by attracting grazers to the post-burn green flush. We explored the effect of burning on grazing activity, and the persistence of grazed patches, in a landscape-level experiment in a South African savanna. We created 17 grazed patches by mowing grass in a 20 m diameter plot, with an adjacent un-mown control. We used dung counts as a measure of grazer visitation, and grass height as a measure of grazing intensity, at each of the sites over a year. Nearly all mowing treatments resulted in a rapid increase in grazing activity relative to controls (on average, 4–6 times more dung was found on mown sites). Subsequent fates of the grazed patches depended on their location with respect to fire. Burned areas drew animals off nearby unburned grazed patches, which then recovered lost biomass. Patches >1.5 km from burns remained grazed short. Frequent large fires might prevent areas of heavy grazing from persisting in the landscape, and thus limit the spread of grazing-adapted grasses. Spatial information on fire frequencies in the park was used to explore the influence that the “magnet effect” of fire can have on grass communities. We mapped the distribution of tall, bunch grasslands and grazing-lawn grasslands using a 1999 Landsat TM satellite image. The extent of grazing lawns was directly related to fire return interval. Areas with a fire return of <4 years had less lawn grass than would be expected from the proportions of lawn grass in the park. A logistic regression analysis, which used various environmental variables known to influence grazing, showed fire history to be an important predictor of grazing-lawn distributions. This work shows that, by influencing where, when, and for how long animals graze a patch, fire can influence the competitive balance between grazing-tolerant, and grazing-intolerant grass species and affect their distributions in the landscape. We discuss the implications of this research for the management of natural grazing systems and rangelands.

Remote Sensing Assessment of Paspalum quadrifarium Grasslands in the Flooding Pampa, Argentina

The knowledge of the distribution, area, and current conservation status of relict natural grasslands dominated by the tall-tussock grass Paspalum quadrifarium Lam. (‘‘pajonal’’) in the Flooding Pampa (Argentina) is relevant for the identification of conservation sites and sustainable management and land-use planning. Since European settlement, vast areas of pajonal were converted to croplands and short-grass prairies. The only available vegetation map of these grasslands was made in the mid-20th century. We evaluated 2 methods of land-cover classification (supervised and unsupervised) using a Landsat TM satellite image over an area of 2 258.21 km2 in Ayacucho county, where pajonal still persists as an important ecosystem. At the paddock scale, this grassland community presents a complex structure in which the pajonal is not a pure category but a mosaic of tall and short grasses. Six categories of land cover were adopted (crops, sown pastures, short grasses, pajonal, wetlands, and urban areas). A very good overall accuracy was obtained for both classifications (86.9% and 87.9% for supervised and unsupervised classifications, respectively). However, both producer’s and user’s accuracies for the pajonal and short grasses were better for the unsupervised classification than for the supervised classification. The pajonal class occupied only 20% of the study area with patch size ranging between 0.09 and 1 653 ha. This work suggests an important replacement of tall-tussock grass by short-grass matrix, which represents noticeable structural and functional changes. The unsupervised classification of Landsat image seems a particularly suitable method for mapping complex vegetation units like the highly fragmented pajonal of the Flooding Pampa and should be an important tool for management and tracking future changes.

Using satellite imagery to assess breeding habitat availability of the endangered loggerhead shrike in Qu�bec

The loggerhead shrike (Lanius ludovicianus) is a grassland bird species whose preferred nesting habitat in eastern Canada is pastureland. This species has been extirpated from much of its historical range in this region, and breeding habitat loss is suspected to be an important cause of this decline. We evaluated the availability of suitable breeding habitats in Quebec using satellite imagery. Because this species no longer breeds in Quebec, we established habitat selection criteria from known nesting sites in the adjacent province of Ontario, from analysis of a Landsat-TM satellite image, and applied these criteria to Landsat-TM images covering southern Quebec. We developed regional landscape criteria in 100 km2 plots and patch indices criteria at the pasture level. Spatial analyses were conducted to characterize plots and pastures on the basis of pasture availability and spatial distribution. Pastures suitable for nesting loggerhead shrikes were those fulfilling patch criteria at the pasture level and located in plots fulfilling regional landscape criteria. Overall, 310 out of 1700 plots located in the historical breeding range of the loggerhead shrike in Quebec fulfilled landscape criteria, supporting 3988 pastures that fulfilled patch criteria. More than 500 of these pastures were visited to validate their current status. The Outaouais region would be the most suitable region for nesting loggerhead shrikes in southern Quebec, where suitable breeding habitat still remains because more than two-thirds of visited sites were still pastureland, hawthorns were well-distributed in the region, and pastureland fragmentation was lowest. We conclude that the availability of breeding habitat does not limit the establishment of a breeding population of loggerhead shrike in southern Quebec, as we estimated that thousands of hectares of suitable habitat still remain in that province.

The Remote Sensing Method of Forest Fire Danger Rating Categorization

Abstract The article presents the use of satellite images in the determination of forest fire danger rating categories. The assessment was carried out based on images from the LANDSAT TM, IKONOS and NOAA satellites, with the finding that the LANDSAT TM images are the most useful. A new solution proposed is to make forest fire danger rating categories refer to forest ranger sub-districts, what gives the forest service greater control over forest fire prevention activities. Forest fire danger assessment was done taking into account remote sensing indices such as the NDVI, TNDVI, and IHT, as well by the analysis of the spatial distribution and the number of fires in the previous six years. In accordance with the Polish State Forest Classification System, three classes were specified: 1 – high fire danger, 2 – moderate fire danger, 3 – low fire danger.

Landslide susceptibility mapping using GIS and the weight-of-evidence model

The weights-of-evidence model (a Bayesian probability model) was applied to the task of evaluating landslide susceptibility using GIS. Using landslide location and a spatial database containing information such as topography, soil, forest, geology, land cover and lineament, the weights-of-evidence model was applied to calculate each relevant factor's rating for the Boun area in Korea, which had suffered substantial landslide damage following heavy rain in 1998. In the topographic database, the factors were slope, aspect and curvature; in the soil database, they were soil texture, soil material, soil drainage, soil effective thickness and topographic type; in the forest map, they were forest type, timber diameter, timber age and forest density; lithology was derived from the geological database; land-use information came from Landsat TM satellite imagery; and lineament data from IRS satellite imagery. Tests of conditional independence were performed for the selection of factors, allowing 43 combinations of factors to be analysed. For the analysis of mapping landslide susceptibility, the contrast values, W + and W -, of each factor's rating were overlaid spatially. The results of the analysis were validated using the previous landslide locations. The combination of slope, curvature, topography, timber diameter, geology and lineament showed the best results. The results can be used for hazard prevention and land-use planning.

An integration of GIS and remote sensing in groundwater investigations: A case study in Burdur, Turkey

The importance of groundwater is growing based on an increasing need and decreasing spring discharges in the Burdur area. Remote Sensing and the Geographic Information System (GIS) have been used for investigation of springs, which are an important groundwater source. The chemical composition of groundwater is not of drinking water quality in Burdur city and water in the Burdur residential area is being obtained from the Cine plain. The purpose of this study was to investigate new water sources by using remote sensing and GIS methods. Geology, lineament and land use maps of the research area were prepared using the Landsat TM satellite image composed of different analyses on the TM 7–4-1 band. In addition, contours, creeks, roads and springs were digitized using a topographic map of 1/100,000 scale to produce a drainage density map. A groundwater potential map was produced which integrated thematic maps, such as annual rainfall, geology, lineament density, land use, topography, slope and drainage density. According to this investigation, the surrounding villages of Askeriye, Bugduz, Gelincik, Taskapi and Kayaalti were determined to be important from the point of view of groundwater potential in the research area.

The Impact of Surrounding Land Use and Vegetation on Single-Family House Prices

The aim of this paper is to assess the marginal effect of land-use locational externalities on the sale price of single-family houses, considering various spatial scales—in accordance with perception theories—and trade-off with accessibility to the city centre. From land-use and vegetation data derived from aerial photographs and Landsat TM satellite images, two sets of hedonic models, using OLS regression, are built from two samples of single-family properties sold in Quebec City. A standard model integrates property-specific factors, census factors, accessibility, and location attributes. In a second model, land-use and vegetation variables are considered on various spatial scales; a third step introduces the interaction effect of the surrounding land use with location, with car-time distance to the main activity centres being used as the main indicator. This allows for an analysis of the spatial variation of the environmental impact throughout the city considering relative proximity to the centre. The successful integration of environmental variables concerning location enhances our understanding of the local land-use and vegetation effects. It also improves the overall performance of the model while virtually removing spatial autocorrelation among residuals. Such models could be used in order to assess the fiscal impacts of various land zoning by law policies, thereby providing planning administrations with a useful decisionmaking tool.

Application of Likelihood Ratio and Logistic Regression Models to Landslide Susceptibility Mapping Using GIS

For landslide susceptibility mapping, this study applied and verified a Bayesian probability model, a likelihood ratio and statistical model, and logistic regression to Janghung, Korea, using a Geographic Information System (GIS). Landslide locations were identified in the study area from interpretation of IRS satellite imagery and field surveys; and a spatial database was constructed from topographic maps, soil type, forest cover, geology and land cover. The factors that influence landslide occurrence, such as slope gradient, slope aspect, and curvature of topography, were calculated from the topographic database. Soil texture, material, drainage, and effective depth were extracted from the soil database, while forest type, diameter, and density were extracted from the forest database. Land cover was classified from Landsat TM satellite imagery using unsupervised classification. The likelihood ratio and logistic regression coefficient were overlaid to determine each factor's rating for landslide susceptibility mapping. Then the landslide susceptibility map was verified and compared with known landslide locations. The logistic regression model had higher prediction accuracy than the likelihood ratio model. The method can be used to reduce hazards associated with landslides and to land cover planning.

하이브리드 분류기법을 이용한 위성영상의 분류에 관한 연구

본 논문에서는 위성영상의 분류에 대한 성능 개선을 위하여 ISODATA 클러스터링, 퍼지 C-Means 알고리즘, 베이시안 최대우도 분류기법을 통합한 하이브리드 분류기법을 제안하였다. 본 연구에서는 분석자에 의하여 분류항목별 학습 데이터를 선정한 후 이를 ISODATA 클러스터링을 이용하여 각각의 분류항목별로 분광특징에 따라 학습 데이터를 세분화하여 새로운 학습 데이터를 선정하였다. 새롭게 선정된 학습 데이터를 이용하여 퍼지 C-Means 알고리즘을 이용하여 분류를 수행하고 그 결과를 베이시안 최대우도 분류기의 사전확률로 적용하여 분류를 수행하였다. 그 결과 분석자가 선정한 분류항목별 훈련데이터의 분광적인 특징에 관계없이 분류를 수행할 수 있었으며 위성영상의 분류의 성능을 개선할 수 있었다. 제안된 기법은 Landsat TM 위성영상을 이용하여 그 적용성을 시험하였다. This paper presents hybrid classification method to improve the performance of satellite images classification by combining Bayesian maximum likelihood classifier, ISODATA clustering and fuzzy C-Means algorithm. In this paper, the training data of each class were generated by separating the spectral signature using ISODATA clustering. We can classify according to pixel's membership grade followed by cluster center of fuzzy C-Means algorithm as the mean value of training data for each class. Bayesian maximum likelihood classifier is performed with prior probability by result of fuzzy C-Means classification. The results shows that proposed method could improve performance of classification method and also perform classification with no concern about spectral signature of the training data. The proposed method Is applied to a Landsat TM satellite image for the verifying test.

Carbon balance in East European tundra

We studied the carbon dioxide and methane fluxes from early June to mid‐September 2001 in the Russian tundra of northeast Europe. Gas fluxes were measured with chamber techniques to determine the seasonal (100 days) carbon gas balance for terrestrial ecosystems representing various vegetation types. Also, the gas balance for aquatic ecosystems in the region was measured. The 2001 fluxes were compared to colder and wetter season fluxes from 1999. The Sphagnum sp. dominated peat plateau fen and Carex sp. and Sphagnum sp. dominated intermediate flarks were carbon sinks of 106 and 110 g C m−2, respectively. In addition, methane emissions were highest from these sites. Other terrestrial surfaces lost carbon to the atmosphere (28–118 g C m−2). The thermokarst lake and the river had seasonal carbon losses of 15 and 34 g C m−2, respectively. For areal integration, the distributions of the various functional surfaces were classified based on Landsat TM satellite image and on‐site validation. This data was used to integrate the carbon fluxes for the entire Lek Vorkuta catchment. The upscaling indicated that the catchment (114 km2) lost 4 (±3.5) Gg C to the atmosphere in summer 2001. The results suggest that predicted warming in the tundra region would induce a substantial loss of carbon. In the warm summer of 2001, the carbon gas released from the whole northeast European tundra (area 205,000 km2) was 8 Tg C when calculated from the Lek Vorkuta data.

Enhancing Landsat TM to update the structural analysis of the Mirs Bay Basin, Hong Kong, China

The coastal provinces of South China have been uniquely shaped by various tectonic events. During the mid-late Mesozoic tectono-thermal event, the oblique subduction of the Paleo Kula-Pacific plate beneath the Eurasian plate has created a complicated tectonic setting for the whole region. However, the mechanism of this event is not completely understood. In this paper, we discuss the advantages of using LANDSAT TM satellite imagery over a small part of the region—the Mirs Bay Basin which is largely covered by dense vegetation and where limited outcrops is seen. The use of satellite imagery complements field mapping and the result shows a prominent sinistral offset along the eastern margin of the Mirs Bay Basin, which was not previously recognized on the ground.

Towards a spatial CO 2 budget of a metropolitan region based on textural image classification and flux measurements

A project on monitoring urban CO 2 budgets has been conducted since the year 2000 focusing on the Metropolitan area of Copenhagen, Denmark. Methodologically, the project combines remote sensing with CO 2 fluxes measured by eddy covariance technique from the top of a 40-m mast located in the center of Copenhagen. These data are supplemented by flux measurements from a mobile system on a 10-m mast which is moved between different urban types, including a major entrance road, and residential and industrial areas. By comparing the time series of vertical CO 2 exchange and the number of cars on the major entrance roads, it is demonstrated that the traffic intensity has a major impact on the urban carbon budget. The spatial distribution of the CO 2 emission rates is examined through texture-based classification of Landsat-TM satellite images. Local traffic intensity and local heating is seen as a function of specific local urban land use and activities, and the corresponding satellite image texture is used as a proxy for the CO 2 emission from these components. The urban scene is divided into urban land use classes that constitute homogenous areas in terms of main types of activity and these are linked to specific levels of CO 2 emission. For this purpose, a multi-scale approach based on co-occurrence matrices has been developed and applied. The paper outlines how the CO 2 exchange from the urban sources and sinks can be estimated from continuous flux measurements in central Copenhagen. It is shown that traffic is the largest single CO 2 source in the city. The mobile measurements demonstrate that the emission rates ranges from less than 0.8 g CO 2 m −2 h −1 in the residential areas up to a maximum of 16 g CO 2 m −2 h −1 along the major entrance roads in the city center. An average annual CO 2 exchange rate of 35 g CO 2 m −2 day −1 is calculated by assigning fluxes to each land use type and excluding the effect of remote sources (power plants, air and sea traffic). This value can be compared to a carbon budget recently calculated from national statistics showing that the local urban sources (road traffic, industry, service and household) have a comparable net emission rate of 38 g CO 2 m −2 day −1. The perspective of having more precise knowledge of the distribution of sources and sinks is finally discussed in relation to changing land use patterns.

Floristic patterns along a 43‐km long transect in an Amazonian rain forest

Summary 1 The floristic variation in Amazonian lowland forests is poorly understood, especially in the large areas of non-inundated (tierra firme) rain forest. Species composition may be either unpredictable as abundances fluctuate in a random walk, more-or-less uniform, or it may correspond to environmental heterogeneity. 2 We tested the three hypotheses by studying the floristic variation of two phylogenetically distant plant groups along a continuous 43-km long line transect that crossed tierra firme rain forest in northern Peru. 3 The observed floristic patterns were compared to patterns in the spectral reflectance characteristics of the forest as recorded in a Landsat TM satellite image. The topography of the transect was measured in the field, and surface soil samples were collected to document edaphic conditions. The two plant groups, pteridophytes and the Melastomataceae, were assessed in 2-m wide and 500-m long sampling units. 4 Floristic similarity (Jaccard index) between sampling units ranged from 0.01 to 0.71 (mean = 0.27), showing that some units were almost completely dissimilar while others were very alike. 5 Spatially constrained clustering produced very similar subdivisions of the transect when based separately on satellite image data, pteriophytes, and Melastomataceae, and the subdivisions were also related to topography and soil characteristics. Mantel tests showed that floristic similarity patterns of the two plant groups were highly correlated with each other and with similarities in reflectance patterns of the satellite image, and somewhat less correlated with geographical distance. 6 Our results lend no support to the uniformity hypothesis, but they partially support the random walk model, and are consistent with the hypothesis that species segregate edaphically at the landscape scale within the uniform-looking forest.