Fuzzy mapping of tropical land cover along an environmental gradient from remotely sensed data with an artificial neural network

Remote sensing is the only feasible means of mapping and monitoring land cover at regional to global scales. Unfortunately the maps are generally derived through the use of a conventional 'hard' classification algorithm and depict classes separated by sharp boundaries. Such approaches and representations are often inappropriate particularly when the land cover being represented may be considered to be fuzzy. The definition of boundaries between classes can therefore be difficult from remotely sensed data, particularly for continuous land cover classes which are separated by a fuzzy boundary which may also vary spatially in time. In this paper a neural network was used to derive fuzzy classifications of land cover along a transect crossing the transition from moist semi-deciduous forest to savanna in West Africa in February and December 1990. The fuzzy classifications revealed both sharp and gradual boundaries between classes located along the transect. In particular, the fuzzy classifications enabled the definition of important boundary properties, such as width and temporal displacement.     

Understanding spatial variations of malaria in Vietnam using remotely sensed data integrated into GIS and machine learning classifiers

Abstract

This study examines the potentials of remotely sensed data, GIS and some machine learning classifiers and ensemble techniques in the investigation of the non-linear relationship between malaria occurrences and socio-physical conditions in the Dak Nong province of Viet Nam. Accuracy assessment was determined with Receiver Operating Characteristic (ROC) curve and pair t-test. The results showed that the area under ROC of Random Subspace ensemble model performed better than the other models based on statistical indicators. Comparing pair t-test with Area Under Curve values showed a slight difference of about 1%. Therefore ensemble techniques had significantly improved the performance of the base classifier. However, the performances might vary according to geographic locations. It is concluded that the machine learning classifiers combined with remotely sensed data and GIS is promising for malaria vulnerability mapping, and the derived maps can be used as a fundamental basis for programmes on spatial disease control.     

Characterizing hydrothermal alteration zones in Hamama area in the central Eastern Desert of Egypt by remotely sensed data

Abstract

Remote sensing techniques provide meaningful information to mineral exploration by identifying the hydrothermally altered minerals and the fracture/fault systems. In this article, Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data were processed to detect the hydrothermal alteration zones in Hamama area in the central part of the Eastern Desert of Egypt. Band ratios and principal component analyses successfully revealed the extent and the geometry of the hydrothermal alteration zones that trend in an NE–SW direction. Matching pixel spectrum derived from Minimum Noise Fraction, Pixel Purity Index, and n-dimensional visualization with reference spectra allowed characterizing key hydrothermal alteration minerals, including chlorite, kaolinite-smectite, muscovite, and haematite, in a successive alteration pattern. Field investigations and X-Ray Diffraction analysis validated the results revealed by ASTER data. In addition, the present prospects of significant gold and massive sulphide mineralizations are consistent with the detected hydrothermal alteration zone.     

Chronology and backtracking of oil slick trajectory to source in offshore environments using ultraspectral to multispectral remotely sensed data

Offshore natural seepage confirms the occurrence of an active petroleum system with thermal maturation and migration, regardless its economic viability for petroleum production. Ocean dynamics, however, impose a challenge for correlation between oil seeps detected on the water surface and its source at the ocean floor. This hinders the potential use of seeps in petroleum exploration. The present study aims to estimate oil exposure time on the water surface via remote sensing in order to help locating ocean floor seepage sources. Spectral reflectance properties of a variety of fresh crude oils, oil films on water and oil–water emulsions were determined. Their spectral identity was used to estimate the duration of exposure of oil–water emulsions based on their temporal spectral responses. Laboratory models efficiently predicted oil status using ultraspectral (>2000 bands), hyperspectral (>300 bands), and multispectral (<10 bands) sensors covering near infrared and shortwave infrared wavelengths. An oil seepage recorded by the ASTER sensor on the Brazilian coast was used to test the designed predictive model. Results indicate that the model can successfully forecast the timeframe of crude oil exposure in the ocean (i.e., the relative “age” of the seepage). The limited spectral resolution of the ASTER sensor, though, implies less accurate estimates compared to higher resolution sensors. The spectral libraries and the method proposed here can be reproduced for other oceanic areas in order to approximate the duration of exposure of noticeable natural oil seepages. This type of information is optimal for seepage tracing and, therefore, for oceanic petroleum exploration and environmental monitoring.     

Employing ground and satellite-based QuickBird data and random forest to discriminate five tree species in a Southern African Woodland

With the emergence of very high spatial and spectral resolution data set, the resolution gap that existed between remote-sensing data set and aerial photographs has decreased. The decrease in resolution gap has allowed accurate discrimination of different tree species. In this study, discrimination of indigenous tree species (n?=?5) was carried out using ground based hyperspectral data resampled to QuickBird bands and the actual QuickBird imagery for the area around Palapye, Botswana. The purpose of the study was to compare the accuracies of resampled hyperspectral data (resampled to QuickBird sensors) with the actual image (QuickBird image) in discriminating between the indigenous tree species. We performed Random Forest (RF) using canopy reflectance taking from ground-based hyperspectral sensor and the reflectance delineated regions of the tree species. The overall accuracies for classifying the five tree species was 79.86 and 88.78% for both the resampled and actual image, respectively. We observed that resampled data set can be upscale to actual image with the same or even greater level of accuracy. We therefore conclude that high spectral and spatial resolution data set has substantial potential for tree species discrimination in savannah environments.     

Thematic mapping from remotely sensed data with neural networks: MLP, RBF and PNN based approaches

 Neural networks are attractive tools for the derivation of thematic maps from remotely sensed data. Most attention has focused on the multilayer perceptron (MLP) network but other network types are available and have different properties that may sometimes be more appropriate for some applications. Here a MLP, radial basis function (RBF) and probabilistic neural network (PNN) were used to classify remotely sensed data of an agricultural site. The accuracy of these classifications ranged from 86.25–91.25%. The accuracy of the PNN classification could be increased through the incorporation of prior probabilities of class membership but the accuracy of each classification could also be degraded by the presence of an untrained class. Post-classification analyses, however, could be used to identify potentially misclassified cases, including those belonging to an untrained class, to increase accuracy. The effect of the post-classification analysis on the accuracy of the classification derived from each of the three network types investigated differed and it is suggested that network type be selected carefully to meet the requirements of the application in-hand. Received: 23 March 2000 / Accepted: 9 July 2000     

Generic rule-sets for automated detection of urban tree species from very high-resolution satellite data

The sustainable management and monitoring of urban forests is an important activity in the urbanized world, and operational approaches require information about the status of urban trees to determine the best strategy. One limitation in urban forest studies is the detection and discrimination of tree species using limited training data. Thus, this study focuses on developing generic rule sets from high-resolution WorldView-2 imagery in conjunction with spectral, spatial, colour and textural information for automated urban tree species detection. The object-based image analysis and its combination with statistical analysis of object features is utilized for this purpose. Results of attribute selection indicated that from 55 attributes, only 26 were useful to discriminate urban tree species, namely Messua ferrea L., Samanea saman and Casuarina sumatrana. Finally, the high overall accuracy, approximately 86.87% with kappa of 0.75 confirmed the transferability of the generic model.     

Bayesian and geostatistical approaches to combining categorical data derived from visual and digital processing of remotely sensed images

This paper seeks a synthesis of Bayesian and geostatistical approaches to combining categorical data in the context of remote sensing classification. By experiment with aerial photographs and Landsat TM data, accuracy of spectral, spatial, and combined classification results was evaluated. It was confirmed that the incorporation of spatial information in spectral classification increases accuracy significantly. Secondly, through test with a 5-class and a 3-class classification schemes, it was revealed that setting a proper semantic framework for classification is fundamental to any endeavors of categorical mapping and the most important factor affecting accuracy. Lastly, this paper promotes non-parametric methods for both definition of class membership profiling based on band-specific histograms of image intensities and derivation of spatial probability via indicator kriging, a non-parametric geostatistical technique.     

Mapping species of submerged aquatic vegetation with multi-seasonal satellite images and considering life history information

Spatial information of the dominant species of submerged aquatic vegetation (SAV) is essential for restoration projects in eutrophic lakes, especially eutrophic Taihu Lake, China. Mapping the distribution of SAV species is very challenging and difficult using only multispectral satellite remote sensing. In this study, we proposed an approach to map the distribution of seven dominant species of SAV in Taihu Lake. Our approach involved information on the life histories of the seven SAV species and eight distribution maps of SAV from February to October. The life history information of the dominant SAV species was summarized from the literature and field surveys. Eight distribution maps of the SAV were extracted from eight 30 m HJ-CCD images from February to October in 2013 based on the classification tree models, and the overall classification accuracies for the SAV were greater than 80%. Finally, the spatial distribution of the SAV species in Taihu in 2013 was mapped using multilayer erasing approach. Based on validation, the overall classification accuracy for the seven species was 68.4%, and kappa was 0.6306, which suggests that larger differences in life histories between species can produce higher identification accuracies. The classification results show that Potamogeton malaianus was the most widely distributed species in Taihu Lake, followed by Myriophyllum spicatum, Potamogeton maackianus, Potamogeton crispus, Elodea nuttallii, Ceratophyllum demersum and Vallisneria spiralis. The information is useful for planning shallow-water habitat restoration projects.     

Bayesian and Geostatistical Approaches to Combining Categorical Data Derived from Visual and Digital Processing of Remotely Sensed Images

This paper seeks a synthesis of Bayesian and geostatistical approaches to combining categorical data in the context of remote sensing classification. By experiment with aerial photographs and Landsat TM data, accuracy of spectral, spatial, and combined classification results was evaluated. It was confirmed that the incorporation of spatial information in spectral classification increases accuracy significantly. Secondly, through test with a 5-class and a 3-class classification schemes, it was revealed that setting a proper semantic framework for classification is fundamental to any endeavors of categorical mapping and the most important factor affecting accuracy. Lastly, this paper promotes non-parametric methods for both definition of class membership profiling based on band-specific histograms of image intensities and derivation of spatial probability via indicator kriging, a non-parametric geostatistical technique.     

Quantifying spatiotemporal dynamics of agricultural landscapes using remotely sensed data and landscape metrics

This study focuses on the spatiotemporal dynamics of agricultural lands and differences in rapidly developing urban and declining rural counties in Iowa, USA between 1984 and 2000. The study presents an analysis of land-cover maps derived from Landsat TM and ETM+ satellite imagery and different landscape metrics using FRAGSTATS and IDRISI software. The study provides evidence of both loss of croplands and change in fragmentation between 1984 and 2000. Fragmentation in agriculture-dominated areas increased with the development of urban centres and diversification of land uses. Fragmentation of landscapes, including agricultural land, was found to be higher in the urbanized counties, but was stable or even declined over time in these counties. In contrast, in the context of remote rural areas, agricultural landscapes experienced rapid increase in fragmentation and farmland loss. The urban–rural gradient analysis used in this study showed that the highest fragmentation occurred on the city edges. These findings suggest that farmland fragmentation is a complex process associated with socio-economic trends at regional and local scales. In addition, socio-economic determinants of landscape fragmentation differ between areas with diverging development trajectories. Intensive cropland fragmentation in remote agricultural regions, detected by this research, should be further studied and its possible effects on both agricultural productivity and biodiversity should be carefully considered.     

Modelling electrical conductivity of soil from backscattering coefficient of microwave remotely sensed data using artificial neural network

Soil salinity is one of the main agricultural problems which expand to larger areas. Soil scientists categorize salinity level by electrical conductivity (EC) measurement. However, field measurements of EC require extensive time, cost and experiences. Remote sensing is one suitable option to investigate and collect spatial data in larger areas. Many researches estimated soil moisture through microwave, but there are fewer studies which mentioned about direct relationship between EC and backscattering coefficient (BC). Thus, this study aims to propose the estimation of EC directly from BC of microwave. The relationship between EC obtained from field survey and BC from microwave is non-linear, artificial neural network (ANN) is one technique proposed in this study to figure out EC and BC relationship. ANN uses multilayer of interconnected processing resulting in EC value with high accuracy which is acceptable. For this reason, ANN model can be successfully utilized as an effective tool for EC estimation from microwave.     

机载LiDAR和高光谱融合实现温带天然林树种识别

将机载LiDAR(Light Detection and Ranging)与高光谱CASI(Compact Airborne Spectrographic Imager)数据融合,充分利用垂直结构信息和光谱信息进行温带森林树种分类,并与仅用高光谱数据的分类结果相比较,评估融合数据的树种分类能力。结合样地实测数据,首先用LiDAR获得的3维垂直结构信息对CASI影像上的林间空隙进行掩膜,提取林木冠层子集;然后对冠层子集分层掩膜,利用光谱曲线的一阶微分及曲线匹配技术,实现各树种训练样本的自动提取;利用SVM分类器对两种数据分类并比较精度。结果表明,融合数据的树种分类总体精度和Kappa系数(83.88%,0.80)优于仅使用CASI数据(76.71%、0.71),优势树种的制图精度为78.43%—89.22%,用户精度为75.15%—95.65%,整体也优于仅使用CASI的制图精度(68.51%—84.69%)和用户精度(63.34%—95.45%)。结果表明,机载LiDAR与CASI基于像元的融合对温带森林树种识别的精度较仅高光谱数据有较大提高。     

Using ground observations of a digital camera in the VIS-NIR range for quantifying the phenology of Mediterranean woody species

The spectral reflectance of most plant species is quite similar, and thus the feasibility of identifying most plant species based on single date multispectral data is very low. Seasonal phenological patterns of plant species may enable to face the challenge of using remote sensing for mapping plant species at the individual level. We used a consumer-grade digital camera with near infra-red capabilities in order to extract and quantify vegetation phenological information in four East Mediterranean sites. After illumination corrections and other noise reduction steps, the phenological patterns of 1839 individuals representing 12 common species were analyzed, including evergreen trees, winter deciduous trees, semi-deciduous summer shrubs and annual herbaceous patches. Five vegetation indices were used to describe the phenology: relative green and red (green\red chromatic coordinate), excess green (ExG), normalized difference vegetation index (NDVI) and green-red vegetation index (GRVI). We found significant differences between the phenology of the various species, and defined the main phenological groups using agglomerative hierarchical clustering. Differences between species and sites regarding the start of season (SOS), maximum of season (MOS) and end of season (EOS) were displayed in detail, using ExG values, as this index was found to have the lowest percentage of outliers. An additional visible band spectral index (relative red) was found as useful for characterizing seasonal phenology, and had the lowest correlation with the other four vegetation indices, which are more sensitive to greenness. We used a linear mixed model in order to evaluate the influences of various factors on the phenology, and found that unlike the significant effect of species and individuals on SOS, MOS and EOS, the sites' location did not have a direct significant effect on the timing of phenological events. In conclusion, the relative advantage of the proposed methodology is the exploitation of representative temporal information that is collected with accessible and simple devices, for the subsequent determination of optimal temporal acquisition of images by overhead sensors, for vegetation mapping over larger areas.     

The UK land cover map 2000: Planning,construction and calibration of a remotely sensed,user-oriented map of broad habitats

Land cover map 2000 (LCM2000) is a comprehensive survey of UK broad habitats giving vector digital maps from segment-based classification of remotely sensed satellite data. This paper examines the influence of users in designing LCM2000 and the difficulties in applying a user-defined classification. It assesses problems and successes through comparisons with a sample-based field survey. These suggest that LCM2000 accuracy at broad habitat level may be around 80–85%; however, it was not possible fully to discriminate errors in LCM2000 from those of the field survey or from mismatches in scales, resolutions and survey dates. Calibration generated broad habitat cover statistics from LCM2000 data to field survey equivalence. These take full account of the heterogeneity of a study area, helping to generate accurate statistics, including those at local level where the field survey cannot operate effectively. The paper concludes that the comprehensive and extensive coverage from remote sensing comes closer than alternative methods to meeting users needs. However, it recognises that producers of remotely sensed information need to understand better the needs of users, and users need to appreciate what the technology can and cannot deliver. This paper adds some benefits of hindsight to the process of communication.     

Development of a modified neural network-based land cover classification system using automated data selector and multiresolution remotely sensed data

Integrating multiple images with artificial neural networks (ANN) improves classification accuracy. ANN performance is sensitive to training datasets. Complexity and errors compound when merging multiple data, pointing to needs for new techniques. Kohonen's self-organizing mapping (KSOM) neural network was adapted as an automated data selector (ADS) to replace manual training data processes. The multilayer perceptron (MLP) network was then trained using automatically extracted datasets and used for classification. Two hypotheses were tested: ADS adapted from the KSOM network provides adequate and reliable training datasets, improving MLP classification performance; and fusion of Landsat thematic mapper (TM) and SPOT images using the modified ANN approach increases accuracy. ADS adapted from the KSOM network improved training data quality and increased classification accuracy and efficiency. Fusion of compatible multiple data can improve performance if appropriate training datasets are collected. This proved to be a viable classification scheme particularly where acquiring sufficient and reliable training datasets is difficult.     

Assessing the utility WorldView-2 imagery for tree species mapping in South African subtropical humid forest and the conservation implications: Dukuduku forest patch as case study

Indigenous forest biome in South Africa is highly fragmented into patches of various sizes (most patches < 1 km²). The utilization of timber and non-timber resources by poor rural communities living around protected forest patches produce subtle changes in the forest canopy which can be hardly detected on a timely manner using traditional field surveys. The aims of this study were to assess: (i) the utility of very high resolution (VHR) remote sensing imagery (WorldView-2, 0.5–2 m spatial resolution) for mapping tree species and canopy gaps in one of the protected subtropical coastal forests in South Africa (the Dukuduku forest patch (ca.3200 ha) located in the province of KwaZulu-Natal) and (ii) the implications of the map products to forest conservation. Three dominant canopy tree species namely, Albizia adianthifolia, Strychnos spp. and Acacia spp., and canopy gap types including bushes (grass/shrubby), bare soil and burnt patches were accurately mapped (overall accuracy = 89.3 ± 2.1%) using WorldView-2 image and support vector machine classifier. The maps revealed subtle forest disturbances such as bush encroachment and edge effects resulting from forest fragmentation by roads and a power-line. In two stakeholders’ workshops organised to assess the implications of the map products to conservation, participants generally agreed amongst others implications that the VHR maps provide valuable information that could be used for implementing and monitoring the effects of rehabilitation measures. The use of VHR imagery is recommended for timely inventorying and monitoring of the small and fragile patches of subtropical forests in Southern Africa.     

无人机遥感与XGBoost的红树林物种分类

无人机遥感数据会衍生大量的光谱、纹理与结构特征,如何提取优势特征是提高红树林物种分类效率和精度的关键问题.针对深圳福田红树林自然保护区缓冲区获取的无人机高光谱影像和LiDAR点云数据,本研究旨在利用极端梯度提升算法(XGBoost)的“特征重要性”属性筛选出适合红树林物种分类的8类优势特征:基于无人机高光谱影像的单一特...     

Individual tree crown approach for predicting site index in boreal forests using airborne laser scanning and hyperspectral data

Site productivity is essential information for sustainable forest management and site index (SI) is the most common quantitative measure of it. The SI is usually determined for individual tree species based on tree height and the age of the 100 largest trees per hectare according to stem diameter. The present study aimed to demonstrate and validate a methodology for the determination of SI using remotely sensed data, in particular fused airborne laser scanning (ALS) and airborne hyperspectral data in a forest site in Norway. The applied approach was based on individual tree crown (ITC) delineation: tree species, tree height, diameter at breast height (DBH), and age were modelled and predicted at ITC level using 10-fold cross validation. Four dominant ITCs per 400 m² plot were selected as input to predict SI at plot level for Norway spruce (Picea abies (L.) Karst.) and Scots pine (Pinus sylvestris L.). We applied an experimental setup with different subsets of dominant ITCs with different combinations of attributes (predicted or field-derived) for SI predictions. The results revealed that the selection of the dominant ITCs based on the largest DBH independent of tree species, predicted the SI with similar accuracy as ITCs matched with field-derived dominant trees (RMSE: 27.6% vs 23.3%). The SI accuracies were at the same level when dominant species were determined from the remotely sensed or field data (RMSE: 27.6% vs 27.8%). However, when the predicted tree age was used the SI accuracy decreased compared to field-derived age (RMSE: 27.6% vs 7.6%). In general, SI was overpredicted for both tree species in the mature forest, while there was an underprediction in the young forest. In conclusion, the proposed approach for SI determination based on ITC delineation and a combination of ALS and hyperspectral data is an efficient and stable procedure, which has the potential to predict SI in forest areas at various spatial scales and additionally to improve existing SI maps in Norway.