Landsat TM Bands Research Articles

Determination of aerosol optical thickness through the derivation of an atmospheric correction for short-wavelength Landsat TM and ASTER image data: an application to areas located in the vicinity of airports at UK and Cyprus

Aerosol optical thickness is considered to be the most important unknown parameter of every atmospheric correction approach for removing atmospheric effects from satellite remotely sensed images. This study presents a description of the basics of the proposed atmospheric correction procedure, which combines the darkest object subtraction principle and the radiative transfer equations. The method considers the true reflectance values of the selected dark targets acquired in situ and the atmospheric parameters such as the aerosol single scattering phase function, single scattering albedo and water vapour absorption, which are also found from ground measurements. The proposed procedure is applicable to short wavelengths such as Landsat TM band 1, 2 and ASTER band 1 in which water vapour absorption is negligible. The proposed image processing method has been tested successfully to determine the aerosol optical thickness on Landsat-5/TM images of the Lower Thames Valley area located to West London (UK) in the vicinity of Heathrow Airport and to Landsat TM/ETM+ and ASTER images of an area located in the vicinity of Paphos International Airport (Cyprus). The determined aerosol optical thicknesses for the Heathrow Airport area were 0.60, 013 and 0.75 for the Landsat TM images (0.45–0.52 µm) acquired on 17th of May 1985, 2nd of June 1985 and 4th of July 1985. The determined aerosol optical thicknesses for the ASTER (0.52–0.60 µm) images acquired on the 4th of February 2008, 26th of February 2008, 17th of December and 24th of December 2007 were 0.18, 0.39, 0.49 and 0.90, respectively. The accuracy assessment applied using the in situ spectroradiometric and sun-photometer data during the satellite overpass acquired on July–August 2008 for the Paphos area in Cyprus shows satisfactory results both for removing the atmospheric effects and for determining the aerosol optical thickness. Indeed, the high correlation between the determined aerosol optical thickness and those extracted from the visibility values increases the potential of the proposed method.

Aerosol optical thickness (AOT) retrieval over land using satellite image-based algorithm

The aim of this study is to present and apply the proposed algorithm on archived time series Landsat TM over an urban area in the vicinity of Heathrow Airport (UK) acquired in 1984-1986 and to two up-to-dated Landsat TM images in the vicinity of Paphos Airport in Cyprus acquired in July-August 2008. The monitoring of aerosol concen- trations becomes a high environmental priority particularly in urban areas. The proposed algorithm has been developed to allow the quantification of the aerosol optical thickness (AOT) over land. The algorithm compares multitemporal satellite data sets and evaluates radiometric alterations due to the optical atmospheric effects of aerosols. Novel features of this algorithm which is based on the application of radiative transfer calculations are the inclusion of applying iteration procedures for selecting the suitable object for determining the aerosol optical thickness and the automatic division into working grid cells. AOT retrieval from Landsat TM band 1 images of Cyprus has been cross- validated satisfactorily by comparing AOT values found from handheld MICROTOPS II sun photometer.

Mapping the total phosphorus concentration of biosolid amended surface soils using LANDSAT TM data

Conventional methods for soil sampling and analysis for soil variability in chemical characteristics are too time-consuming and expensive for multi-seasonal monitoring over large-scale areas. Hence, the objectives of this study are: 1) to determine changes in chemical concentrations of soils that are amended with treated sewage sludge; and 2) to determine if LANDSAT TM data can be used to map surface chemical characteristics of such amended soils. For this study, we selected two fields in NW Ohio, designated as F34 and F11, that had been applied with 34 and 11 ton acre − 1 of biosolids, respectively. Soil samples from a total of 70 sampling locations across the two fields were collected one day prior to LANDSAT 5 overpass and were analyzed for several elemental concentrations. The accumulation of Ba, Cd, Cu, S and P were found to be significantly higher in the surface soils of field F34, compared to field F11. Regression equations were established to search for algorithms that could map these five elemental concentrations in the surface soils using six, dark-object-subtracted (DOS) LANDSAT TM bands and the 15 non-reciprocal spectral ratios derived from these six bands for the May 20, 2005, LANDSAT 5 TM image. Phosphorus (P) had the highest R 2 adjusted value (67.9%) among all five elements considered, and the resulting algorithm employed only spectral ratios. This model was successfully tested for robustness by applying it to another LANDSAT TM image obtained on June 5, 2005. Our results enabled us to conclude that LANDSAT TM imagery of bare-soil fields can be used to quantify and map the spatial variation of total phosphorous concentration in surface soils. This research has significant implications for identification and mapping of areas with high P, which is important for implementing and monitoring the best phosphorous management practices across the region.

Mapping Karst Rock in Southwest China

Monitoring and assessing karst rock resulting from desertification is important to local sustainable development in the karst mountain regions of Southwest China. A new derived index, the Normalized Difference Rock Index (NDRI), is proposed to map karst rock. It takes advantage of the unique spectral response of karst rock and other land cover forms. Supervised maximum likelihood classification trials using different input bands (Landsat TM band 3–5, NDVI and NDRI derived from TM imagery) were assessed for accuracy of karst rock mapping in Bijie County, Southwest China. Results were compared with land use maps for 2000, with the best result obtained using a combination of the NDVI and NDRI indices. This paper offers a tool for land cover mapping.

Combined Methodology Based on Field Spectrometry and Digital Photography for Estimating Fire Severity

Fire severity can be considered one of the most influential factors in the postfire development of burnt areas. Ground level studies documenting changes in reflectance values improve the discrimination of spatial fire severity across large areas. The objective of this study was to determine those spectral regions most sensitive to fire severity levels by investigating the relationship between postfire surface materials and reflectance values. A total of 34 field plots were analyzed immediately following two natural fires in Spain. To obtain postfire data, we used a portable structure to obtain vertical photographs and reflectance values in the visible-near-infrared (VIS-NIR) range. Spectral reflectance and convolved Landsat Thematic Mapper (TM) bands were statistically correlated with individual postfire materials and with a fire severity index, the Combustion Products Index (CPI), derived from them. The wavelengths most sensitive to postfire materials were 450.6, 758.2, and 797 nm for ash, black carbon, and vegetation, respectively, although these wavelengths were included within ranges of comparable behavior. Finally, stepwise multiple regression models (SMLRs) were applied to the data. SMLR predicted ash, black carbon, and vegetation levels from reflectance data with accuracies higher than 75%. Estimation from Landsat-TM bands yielded slightly lower accuracies. The CPI severity index was also well estimated using either reflectance data or TM bands ( <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">r</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.925 and <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">r</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> = 0.840 , respectively).

Multi‐scale thermal pattern monitoring of a large lake (Lake Geneva) using a multi‐sensor approach

The applicability of satellite imagery products from different sensors (AVHRR‐derived multi‐channel sea surface temperature (MCSST), MODIS sea surface temperature (SST) products 5‐Min L2 Swath 1 km and Landsat TM band 6 thermal signature) for the comprehensive monitoring of temperature and its temporal patterns over a large lake is tested in this study. The coverage of cloud‐free satellite data for Lake Geneva is reported throughout a year and, more specifically, during a 13 day period in summer 2003. In a second step, we demonstrate the feasibility of the AVHRR/MODIS imagery to discern day and night temperature patterns, by generating day and night climatologies and various spatial statistics over the 13 day period. The different day and night surface thermal patterns observed by satellite imagery could be linked to the thermal structure existing in deeper water using the concept of the diurnal decoupled layer. The forcing of the persistent patterns, two warm cores divided by a saddle‐shaped cold anomaly, is explained by wind periodicity and insolation conditions. The patterns can be matched to features postulated by findings of different limnologists in the past. Other surface temperature related phenomena such as water upwelling and downwelling and the occurrence of plumes are related to meteorological and hydrological events. The lakewide average lake surface water temperature (LSWT) trends for day and night during the study period are roughly parallel. A sudden loss of stored heat can be explained by episodes of long fetch, synoptic wind (bise) that interrupted the predominant breeze regime.

La pollution atmosphérique et le risque de maladie thrombo-embolique

The Boeing Landing Suitability Index (BLSI) is a computer algorithm designed at Bowling Green State University for The Boeing Company for the purpose of mapping candidate aircraft operating sites of different “runway” lengths in natural terrain from LANDSAT TM data. Four LANDSAT TM frames (each 115 miles × 115 miles in area) from parts of California, New Mexico, South Dakota, and Louisiana (with annual rainfalls averaging approximately 3.5, 7, 17, and 53 in., respectively, though the Louisiana frame was collected during a drought year that had 28 in. of annual rainfall) were entered as inputs to the BLSI algorithm, which identified pixels in each of the frames that were sufficiently flat, treeless, and absent of surface water to land an aircraft on them, present weather conditions permitting, and perform off-loading, on-loading, and take-off operations. Adjustments were made to the algorithm for differences in annual rainfall. Although no attempt was made before these field trips to force BLSI to map surface firmness, the California Bearing Ratio (CBR) was measured in the field on-site and off-site with dynamic cone penetrometers by a contractor for the Air Force Research Lab (Tyndall AFB in Florida) for several of these candidate aircraft operating sites. We found that all of the sites selected by BLSI that were visited in the field were suitably flat for landing and had no surface water or trees on them (except for one small tree that was younger than the LANDSAT frame employed for that region). Fortunately, most of these candidate aircraft operating sites found by BLSI had firmer ground than their immediate surroundings, as measured from the dynamic cone penetrometers. Our field observations support the hypothesis that BLSI is primarily mapping very subtle, local topographic highs. Such a subtle topographic high, which may rise only few feet from the middle to the edge of a candidate aircraft operating site, usually has a hard pan beneath it, owing to its relatively better drainage, compared with its immediate surrounds. This may account for its greater firmness, compared to its surrounds. Because the slopes are so small and the feature is slightly convex upward, there are no gullies on candidate aircraft operating sites. This causes the subtle topographic highs to be very flat and without rocks, unlike higher-sloped regions, where water in gullies tend to wash rocks. This seems the most reasonable explanation for why a pixel size as gross as 28.5 m (LANDSAT TM bands 1–5 and 7) is adequate to map such features. BLSI displayed as a continuous-toned image has potential for mapping areas of highest trafficability for wheeled vehicles, with lower BLSI values (shown as purple or blue areas) having the greatest trafficability.

Seasonal reflectance course of some forest types in Estonia from a series of Landsat TM and SPOT images and via simulation

A time series of Landsat TM and SPOT images over the Järvselja forest region of southeast Estonia was compiled from a set of 23 images acquired in the years 1986 to 2003 in order to study the seasonal changes in reflectance among different forest types. The images were transformed into ground‐level reflectance factor units and seasonal courses of reflectance were constructed for several forest types in all six reflective spectral bands of Landsat TM. For better comparability between different years, the seasonal time course was presented as a function of temperature (phenology) time. A smoothing of the time series was also applied. The smoothed seasonal series were compared with model simulations for the same forest types using the Kuusk–Nilson forest reflectance model. The input for the model was formed using the data from the respective forestry database. An expert estimate was used to obtain the values for the biochemical and biophysical parameters lacking in the forestry database and of their change in course in the season. The simulations showed that, as main driving factors in the seasonal course, the course of leaf area index (LAI) of overstorey and understorey, chlorophyll content in leaves and ground wetness could be pointed out. The course of sun elevation during the season appeared also to be important in forming the image‐based seasonal course. For the studied forest types, the seasonal average normalised difference vegetation index (NDVI) was calculated from the smoothed seasonal courses of reflectance in the red and near infrared bands. The validity of the Monteith hypothesis was tested by establishing a linear relation between the yearly trunk mass increment and seasonal average NDVI.

Performance of Landsat TM in ship detection in turbid waters

The visible and near infrared bands of Landsat have limitations for detecting ships in turbid water. The potential of TM middle infrared bands for ship detection has so far not been investigated. This study analyzed the performance of the six Landsat TM visible and infrared bands for detecting dredging ships in the turbid waters of the Poyang Lake, China. A colour composite of principal components analysis (PCA) components 3, 2 and 1 of a TM image was used to randomly select 81 dredging ships. The reflectance contrast between ships and adjacent water was calculated for each ship. A z-score and related p-value were used to assess the ship detection performance of the six Landsat TM bands. The reflectance contrast was related to water turbidity to analyze how water turbidity affected the capability of ship identification. The results revealed that the TM middle infrared bands 5 and 7 better discriminated vessels from surrounding waters than the visible and near infrared bands 1–4. A significant relation between reflectance contrast and water turbidity in bands 1–4 could explain the limitations of bands 1–4; while water turbidity has no a significant relation to the reflectance contrast of bands 5 and 7. This explains why bands 5 and 7 detect ships better than bands 1–4.

An image-based diversity index for assessing land degradation in an arid environment in South Australia

One of the major indirect effects of artificial watering points in a grazed landscape is the development around them of a zone of extreme degradation called a piosphere. We investigated the use of the moving standard deviation index (MSDI) applied to Landsat TM band 3 data for detection and assessment of these zones in the arid grazing lands of South Australia. We found that watering points had significantly higher MSDI values ( p<0.001) than non-degraded reference areas in four different land systems. Comparison of two vegetation indices, the normalised difference vegetation index (NDVI) and perpendicular distance vegetation index (PD54), used as reference indices, showed that the PD54 was more appropriate than NDVI in this perennial-dominated arid environment. Piospheres were found to be more heterogeneous in vegetation cover, with higher MSDI values, compared with non-degraded areas, and spatial heterogeneity in cover decreased with increasing distance from water points. This study indicates that MSDI can be used as an appropriate method for land degradation assessment in naturally heterogeneous arid lands of South Australia.

Integrating Landsat TM and SRTM-DEM derived variables with decision trees for habitat classification and change detection in complex neotropical environments

Remotely sensed images and processing techniques are a primary tool for mapping changes in tropical forest types important to biodiversity and environmental assessment. Detailed land cover data are lacking for most wet tropical areas that present special challenges for data collection. For this study, we utilize decision tree (DT) classifiers to map 32 land cover types of varying ecological and economic importance over an 8000 km2 study area and biological corridor in Costa Rica. We assess multivariate QUEST DTs with unbiased classification rules and linear discriminant node models for integrated vegetation mapping and change detection. Predictor variables essential to accurate land cover classification were selected using importance indices statistically derived with classification trees. A set of 35 variables from SRTM-DEM terrain variables, WorldClim grids, and Landsat TM bands were assessed.Of the techniques examined, QUEST trees were most accurate by integrating a set of 12 spectral and geospatial predictor variables for image subsets with an overall cross-validation accuracy of 93%±3.3%. Accuracy with spectral variables alone was low (69%±3.3%). A random selection of training and test set pixels for the entire landscape yielded lower classification accuracy (81%) demonstrating a positive effect of image subsets on accuracy. A post-classification change comparison between 1986 and 2001 reveals that two lowland forest types of differing tree species composition are vulnerable to agricultural conversion. Tree plantations and successional vegetation added forest cover over the 15-year time period, but sometimes replaced native forest types, reducing floristic diversity. Decision tree classifiers, capable of combining data from multiple sources, are highly adaptable for mapping and monitoring land cover changes important to biodiversity and other ecosystem services in complex wet tropical environments.

Water Quality Monitoring to Support the European Commission's Water Framework Directive Reporting Requirements

This article aims at providing a simple way for water quality monitoring in a set of reservoirs using an earth observation-based approach and the assessment of the use of this technique for a monitoring network in order to meet the requirements and objectives of the Water Framework Directive (WFD) by the Member States of the European Union. The study carried out was preformed in 42 reservoirs of the Ebro River Basin, in the northeast of Spain. The proposed methodology is based on the development of an algorithm for the estimation of water quality by means of LANDSAT TM band reflectance. Some band ratios were used in the model as well. Trophic State Index (TSI), calculated by means of Secchi Disc Transparency data, was estimated using a forward stepwise multiple regression analysis and the TM2 band and TM1/TM2 ratio. The final results showed a large variability in water quality across reservoirs. Moreover, substantial spatial heterogeneity was also observed in the water bodies. These results demonstrate the likelihood of developing a monitoring network based on remote sensing techniques for the implementation of the WFD.

Haze reduction from the visible bands of LANDSAT TM and ETM+ images over a shallow water reef environment

A method for haze reduction in the visible bands of Landsat TM and ETM+ images over a shallow water marine environment is presented in this paper. This method uses the near infrared (NIR) band to estimate the spatial distribution of haze intensity in each visible band through a linear regression model established over deep water areas. As a first order approximation, the signal received at the sensor is assumed to be the arithmetic sum of radiance contributed by haze and the radiance leaving the water surface. Reduction of haze is then carried out by a simple subtraction procedure. Images acquired over the Southern Tip of Palawan, Philippines are used for the experiments. Results show that the method works well for compensating signals contaminated by optically thin haze. Overcorrection occurs when haze is optically thick and geometrically complex. When images are acquired under hazy conditions the method can be applied to drastically improve image interpretability and may also be considered as a necessary pre-processing step for subsequent analyses and information extraction.

Spectral Reflectance Measurements of a Microcystis Bloom in Upper Klamath Lake, Oregon

The objective of this study was to measure the in situ spectral reflectance of lake water that contains a bloom of Microcystis, a species of cyanobacteria. Reflectance spectra (350–2,500 nm) of lake water near a boat dock in Upper Klamath Lake, Oregon, were collected with a portable spectroradiometer on a cloud-free day with sunlight as a source of illumination between 0845 to 0915 hours, Pacific Daylight Time (PDT) on 17 August 2004, at a near-normal angle of observation. The averaged spectrum of the lake water containing the Microcystis bloom exhibits reflectance maxima from 550-590 nm and near 710 nm wavelengths and reflectance minima near 630 nm and 675 nm wavelengths. The reflectance gradually decreases from 810–1,000 nm and has very low reflectance in the 1,000–2,500 nm wavelength region. Our results show that the spectral reflectance of Microcystis at this stage of its bloom remains low for wavelengths longer than 1,000 nm in the near-infrared region of the spectrum. These spectral results have implications in selecting the spectral ratios and refining the algorithms that will be used to estimate phycocyanin content using satellite models. Microcystis is the predominant species of cyanobacteria blooms in Lake Erie, which makes these spectral data as important to the Great Lakes as it is to Upper Klamath Lake. Satellite algorithms have been published that have mapped phycocyanin, a pigment more uniquely associated with cyanobacteria than is chlorophyll a, in Lake Erie with data from LANDSAT TM bands 1, 3, 4, 5, and 7, and the reflectance spectra reported here are the first that cover the entire spectral range of all those LANDSAT TM spectral bands for a Microcystis bloom.

Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery

The normalized difference water index (NDWI) of McFeeters (1996) was modified by substitution of a middle infrared band such as Landsat TM band 5 for the near infrared band used in the NDWI. The modified NDWI (MNDWI) can enhance open water features while efficiently suppressing and even removing built‐up land noise as well as vegetation and soil noise. The enhanced water information using the NDWI is often mixed with built‐up land noise and the area of extracted water is thus overestimated. Accordingly, the MNDWI is more suitable for enhancing and extracting water information for a water region with a background dominated by built‐up land areas because of its advantage in reducing and even removing built‐up land noise over the NDWI.

Modeling woody vegetation resources using Landsat TM imagery in northern Namibia

In 1995 a forest inventory covering northern Namibia was initiated, based on stratified systematic field sampling of plots with a radius of up to 30 m. In these plots detailed tree parameters were measured. Due to security problems the most important wooded parts of the area could not be covered. This study investigated whether Landsat TM imagery could be used to estimate woody vegetation parameters. As the existing field sampling method did not result in significant relationships between pixel values of different bands and tree cover, two sampling methods of different design were tested. Both resulted in statistically significant relationships between tree cover and pixel values of mainly band 4 reflectance of Landsat TM. Regression equations to estimate cover and volumes were obtained. The increased size of the sample plots in both methods was the main reason for improved correlations. The relation between tree cover and Landsat TM band 4 was influenced by fire scars and patches of heavy grazing. Estima...

Retrieval of ocean colour from high resolution multi‐spectral imagery for monitoring highly dynamic ocean features

Retrieval of ocean colour information from a space borne Multi‐spectral Camera (MSC) on KOMPSAT‐2 is investigated to study and characterize small‐scale biogeophysical features that are very rich and dynamic in nature in the coastal oceans rather than the interior. Prior to the derivation of this information from space‐borne ocean colour observations, the path radiance largely from the atmospheric path and air–sea interface should be removed from the total signal recorded at the top of the atmosphere (TTOA ). In this study, the ‘path extraction’ method is introduced for the atmospheric correction of ocean colour images. The potential use of path extraction was demonstrated on Landsat TM and SeaWiFS images of highly turbid coastal waters of Korea. The path‐extracted water‐leaving radiance was then compared with the water‐leaving radiance spectra derived from the standard SeaWiFS atmospheric correction algorithm. It was noticed that the path‐extracted water‐leaving radiance resembled in situ spectra while the same was found highly degraded throughout the visible wavebands by adopting the standard SeaWiFS atmospheric correction algorithm. Algorithms for the retrieval of ocean colour information are explored from remotely sensed reflectance (Rrs ) in the visible wavelength bands of a Multi‐spectral Camera. A large set of remote sensing reflectances are generated by random number functions using an Rrs model, which relates bb /(a+bb ) to Rrs as functions of inherent optical properties, such as absorption and backscattering coefficients of six water components including water, phytoplankton (chl), dissolved organic matter (DOM), suspended sediment (SS) concentration, heterotropic organisms (he) and an unknown component, possibly represented by bubbles or other particulates unrelated to the first five components. Since the Kompsat‐2 MSC and Landsat‐5 TM bands are spectrally similar, these Rrs values are then modelled to the equivalent remote sensing reflectances at MSC and Landsat TM bands using a spectral band model. The empirical relationships between the spectral ratios of modelled Rrs (e.g. Rrs (MSC band1)/Rrs (MSC band2) and Rrs (MSC band1‐centre)/Rrs (MSC band2‐centre)) and chlorophyll concentrations are established in order to derive ⟨chl⟩ algorithms for both Landsat TM and MSC bands. Similarly, ⟨SS⟩ algorithms are obtained by relating a single band reflectance (e.g. Rrs (MSC band2) and Rrs (MSC Band2‐centre)) to the suspended sediment concentrations. Finally, a comparative analysis is made between the Landsat TM and MSC bands as well as narrow (centre‐wavelength) and broad band (full bandwidth) width of algorithms. From this study, it was observed that the Rrs spectra of three MSC spectral bands are found to be slightly superior to the Landsat TM bands in terms of spectral sensitivity to varying constituent concentrations. A small discrepancy between the reflectance ratios of broad and narrow bands was noticed in MSC and Landsat TM. The coefficient of determination (R 2) for log‐transformed data [⟨chl⟩ N = 500] was interestingly found to be R 2 = 0.90 for both Landsat TM and MSC. Similarly, the R 2 value for log‐transformed data [⟨SS⟩ N = 500] was 0.93 and 0.92 for Landsat TM and MSC, respectively. The modelled Rrs spectra were in good agreement with our in situ spectra obtained from the southern coastal Sea of Korea during 1998 and 1999. The algorithms presented are expected to explore the fine details of the complex coastal oceanic features from the ocean colour images of Multi‐spectral Camera and Landsat TM.

Landsat-based Remote Sensing of Lake Water Quality Characteristics, Including Chlorophyll and Colored Dissolved Organic Matter (CDOM)

Ground-based measurements on 15 Minnesota lakes with wide ranges of optical properties and Landsat TM data from the same lakes were used to evaluate the effect of humic color on satellite-inferred water quality conditions. Color (C440), as measured by absorbance at 440 nm, causes only small biases in estimates of Secchi disk transparency (SDT) from Landsat TM data, except at very high values (> ~ 300 chloroplatinate units, CPU). Similarly, when chlorophyll a (chl a) levels are moderate or high (> 10 μg/L), low-to-moderate levels of humic color have only a small influence on the relationship between SDT and chl a concentration, but it has a pronounced influence at high levels of C440 (e.g., > ~200 CPU). However, deviations from the general chl a-SDT relationship occur at much lower C440 values (~ 60 CPU) when chl a levels are low. Good statistical relationships were found between optical properties of lake water generally associated with algal abundance (SDT, chl a, turbidity) and measured brightness of various Landsat TM bands. The best relationships for chl a (based on R2 and absence of statistical outliers or lakes with large leverage) were combinations of bands 1, 2, or 4 with the band ratio 1:3 (R2 = 0.88). Although TM bands 1–4 individually or as simple ratios were poor predictors of C440, multiple regression analyses between ln(C440) and combinations of bands 1–4 and band ratios yielded several relationships with R2 ≥ 0.70, suggesting that C440 can be estimated with fair reliability from Landsat TM data.

CLASSIFICAÇÃO DO USO/COBERTURA DA TERRA POR MEIO DE ALGORITMOS NÃO PARAMÉTRICOS COMPARADOSAO CLASSIFICADOR DE MÁXIMA VEROSSIMILHANÇA

Mapas de uso/cobertura da terra gerados por meio de estratégias de classificação não paramétrica de uma imagem Landsat TM foram comparados ao resultado gerado pelo tradicional classificador de máxima verossimilhança (MAXVER). Algumas estratégias não paramétricas, desenvolvidas neste trabalho, tiveram a finalidade de diminuir a limitação do classificador não paramétrico de Skidmore/Turner, que requer um tamanho de amostra de treinamento muito grande, comparado ao MAXVER, para reduzir a área não classificada na imagem. Os resultados mostraram a superioridade do MAXVER, na combinação entre exatidão e área não classificada na imagem. De qualquer forma, a classificação não paramétrica tem a propriedade de não necessitar da pressuposição de normalidade dos dados, utilizando a real informação das amostras de treinamento para classificação, o que lhe assegura maior exatidão.

Using Aerial Video to Train the Supervised Classification of Landsat Tm Imagery for Coral Reef Habitats Mapping

Management of coral reef resources is a challenging task, in many cases, because of the scarcity or inexistence of accurate sources of information and maps. Remote sensing is a not intrusive, but powerful tool, which has been successfully used for the assessment and mapping of natural resources in coral reef areas. In this study we utilized GIS to combine Landsat TM imagery, aerial photography, aerial video and a digital bathymetric model, to assess and to map submerged habitats for Alacranes reef, Yucatán, México. Our main goal was testing the potential of aerial video as the source of data to produce training areas for the supervised classification of Landsat TM imagery. Submerged habitats were ecologically characterized by using a hierarchical classification of field data. Habitats were identified on an overlaid image, consisting of the three types of remote sensing products and the bathymetric model. Pixels representing those habitats were selected as training areas by using GIS tools. Training areas were used to classify the Landsat TM bands 1, 2 and 3 and the bathymetric model by using a maximum likelihood algorithm. The resulting thematic map was compared against field data classification to improve habitats definition. Contextual editing and reclassification were used to obtain the final thematic map with an overall accuracy of 77%. Analysis of aerial video by a specialist in coral reef ecology was found to be a suitable source of information to produce training areas for the supervised classification of Landsat TM imagery in coral reefs at a coarse scale.