A flocculation tensor to monitor water quality using a deep learning model

The increasing quantities of polluted waters are calling for advanced purification methods. Flocculation is an essential component of the water purification process, yet flocculation is commonly not optimal due to our poor understanding of the flocculation process. In particular, there is little knowledge on the mechanisms ruling the migration of pollutants during treatment. Here we have created the first tensor diagram, a mathematical framework for the flocculation process, analyzed its properties with a deep learning model, and developed a classification scheme for its relationship with pollutants. The tensor was constructed by combining pixel matrices from a variety of floc images, each with a particular flocculation period. Changing the factors used to make flocs images, such as coagulant dose and pH, resulted in tensors, which were used to generate matrices, that is the tensor diagram. Our deep learning algorithm employed a tensor diagram to identify pollution levels. Results show tensor map attributes with over 98% of sample images correctly classified. This approach offers potential to reduce the time delay of feedback from the flocculation process with deep learning categorization based on its clustering capabilities. The advantage of the tensor data from the flocculation process improves the efficiency and speed of response for commercial water treatment.

     

基于决策树的辽宁省北部沙漠化信息提取研究

以沙漠化问题较突出的辽宁省北部地区为例,选取2007年Landsat 5 TM遥感影像作为基本数据源,通过对影像中耕地、林地、草地、水域等常见地物及典型沙漠化土地进行光谱特征分析和波段间的相互运算,将修改型土壤调整植被指数（MSAVI）、归一化差异水体指数（NDWI）和遥感图像缨帽变换后的土壤亮度指数（SBI）、绿度植被指数（GVI）及湿度指数（WVI）等特征变量融入决策树分类模型后进行分层分离,从而实现对沙漠化信息的高精度提取。结果显示,决策树分类法可排除提取地物时的干扰信息,是保证沙漠化土地信息快速自动提取的方法之一。     

Coral reef habitat mapping: how much detail can remote sensing provide?

The capability of satellite and airborne remote-sensing methods for mapping Caribbean coral reefs is evaluated. Reef habitats were categorised into coarse, intermediate and fine detail, using hierarchical classification of field data (percent cover in 1 m quadrats and seagrass standing-crop). Habitats were defined as assemblages of benthic macro-organisms and substrata and were mapped using the satellite sensors Landsat MSS, Landsat TM, SPOT XS, SPOT Pan and merged Landsat TM/SPOT Pan. Habitats were also mapped using the high-resolution digital airborne sensor, CASI (compact airborne spectrographic imager). To map areas >60 km in any direction with coarse detail, Landsat TM was the most accurate and cost-effective satellite sensor (SPOT XS when <60 km). For maps with intermediate habitat detail, aerial photography (from a comparable study in Anguilla) exhibited similar accuracy to Landsat TM, SPOT XS, SPOT Pan and merged Landsat TM/SPOT Pan. Landsat MSS was consistently the least accurate sensor. Maps from CASI were significantly (p<0.001) more accurate than satellite sensors and aerial photographs. Maps with detailed habitat information (i.e. >9 reef classes) had a maximum accuracy of 37% when based on satellite imagery, but aerial photography and CASI achieved accuracies of 67 and 81%, respectively. Commissioning of new aerial photography does not appear to be a cost-effective option; satellites are cheaper for coarse habitat-mapping, and detailed habitat-mapping can be conducted more accurately and cheaply with CASI. The results will guide practitioners in matching survey objectives to appropriate remote-sensing methods. Received: 11 July 1997 / Accepted: 6 August 1997     

Identification of hot spots and well managed areas of Pichavaram mangrove using Landsat TM and Resourcesat—1 LISS IV: an example of coastal resource conservation along Tamil Nadu Coast, India

The present work is a multi-temporal satellite based study on the spatial dynamic of an important coastal habitat, the Pichavaram mangrove ecosystem, over a period of 15 years. The Pichavaram mangrove forest near Chidambaram, South India is the second largest mangrove forest in the world. Unsupervised classification, the Iterative Self Organising Data Analysis Technique (ISODATA), has been used to classify the mangrove cover into the open and dense classes. The status of the classes has been monitored using Landsat TM of 1991, 2001, and Resourcesat–1 LISS IV of 2006 satellite data. The study demonstrated that by classifying mangrove ecosystem into just the 3 classes using remote sensing data and by studying their temporal variations, it is possible to get a reasonably accurate picture of the extent and condition of the mangrove ecosystem. The total area of the Pichavaram mangrove showed a net increase of 2.51 km² within a span of 15 years (1991 to 2006). The hot spots that are at a risk of being degraded, and on the other hand, the mangrove areas that are well managed are identified using Geographical Information System (GIS) tools for the restoration and conservation measures.     

Quantitative Determination of Rarity of Freshwater Fishes and Implications for Imperiled‐Species Designations

Abstract: Conserving rare species and protecting biodiversity and ecosystem functioning depends on sound information on the nature of rarity. Rarity is multidimensional and has a variety of definitions, which presents the need for a quantitative classification scheme with which to categorize species as rare or common. We constructed such a classification for North American freshwater fishes to better describe rarity in fishes and provide researchers and managers with a tool to streamline conservation efforts. We used data on range extents, habitat specificities, and local population sizes of North American freshwater fishes and a variety of quantitative methods and statistical decision criteria, including quantile regression and a cost‐function algorithm to determine thresholds for categorizing a species as rare or common. Species fell into eight groups that conform to an established framework for rarity. Fishes listed by the American Fisheries Society (AFS) as endangered, threatened, or vulnerable were most often rare because their local population sizes were low, ranges were small, and they had specific habitat needs, in that order, whereas unlisted species were most often considered common on the basis of these three factors. Species with large ranges generally had few specific habitat needs, whereas those with small ranges tended to have narrow habitat specificities. We identified 30 species not designated as imperiled by AFS that were rare along all dimensions of rarity and may warrant further study or protection, and we found three designated species that were common along all dimensions and may require a review of their imperilment status. Our approach could be applied to other taxa to aid conservation decisions and serve as a useful tool for future revisions of listings of fish species.     

A Bayesian approach to multi-source forest area estimation

In efforts such as land use change monitoring, carbon budgeting, and forecasting ecological conditions and timber supply, there is increasing demand for regional and national data layers depicting forest cover. These data layers must permit small area estimates of forest area and, most importantly, provide associated error estimates. This paper presents a model-based approach for coupling mid-resolution satellite imagery with plot-based forest inventory data to produce estimates of probability of forest and associated error at the pixel-level. The proposed Bayesian hierarchical model provides access to each pixel’s posterior predictive distribution allowing for a highly flexible analysis of pixel and multi-pixel areas of interest. The paper presents a trial using multiple dates of Landsat imagery and USDA Forest Service Forest Inventory and Analysis plot data. The results describe the spatial dependence structure within the trial site, provide pixel and multi-pixel summaries of probability of forest land use, and explore discretization schemes of the posterior predictive distributions to forest and non-forest classes. Model prediction results of a holdout set analysis suggest the proposed model provides high classification accuracy, 88%, for the trial site.

Refining Biodiversity Conservation Priorities

Abstract:  Although there is widespread agreement about conservation priorities at large scales (i.e., biodiversity hotspots), their boundaries remain too coarse for setting practical conservation goals. Refining hotspot conservation means identifying specific locations (individual habitat patches) of realistic size and scale for managers to protect and politicians to support. Because hotspots have lost most of their original habitat, species endemic to them rely on what remains. The issue now becomes identifying where this habitat is and these species are. We accomplished this by using straightforward remote sensing and GIS techniques, identifying specific locations in Brazil's Atlantic Forest hotspot important for bird conservation. Our method requires a regional map of current forest cover, so we explored six popular products for mapping and quantifying forest: MODIS continuous fields and a MODIS land cover (preclassified products), AVHRR, SPOT VGT, MODIS (satellite images), and a GeoCover Landsat thematic mapper mosaic (jpg). We compared subsets of these forest covers against a forest map based on a Landsat enhanced thematic mapper. The SPOT VGT forest cover predicted forest area and location well, so we combined it with elevation data to refine coarse distribution maps for forest endemic birds. Stacking these species distribution maps enabled identification of the subregion richest in threatened birds—the lowland forests of Rio de Janeiro State. We highlighted eight priority fragments, focusing on one with finer resolved imagery for detailed study. This method allows prioritization of areas for conservation from a region >1 million km² to forest fragments of tens of square kilometers. To set priorities for biodiversity conservation, coarse biological information is sufficient. Hence, our method is attractive for tropical and biologically rich locations, where species location information is sparse.     

Optimizing model for land use/land cover retrieval from remote sensing imagery based on variable precision rough sets

The suitable spectral mode in remote sensing is often desirable to facilitate the inversion of ecological environment and landscape. This paper put forward an optimizing model based on variable precision rough sets (VPRS) for the land cover discrimination in wetland inventory. In the case study of Lake Baiyangdian which has important ecological functions to the northern China, this model is established successfully according to the domain-experts knowledge. The procedure is as follows. First step is data collection, including remote-sensing data (e.g., Landsat-5 TM bands), the digitized relief maps, and statistical yearbooks. Second, the remote sensing imagery (RSI) and relief maps are co-registered into the same resolution. Third, a condition set, including various attributes is derived from spectral bands, band math or ratio indices based on previous studies, at the same time, the decision set is derived from true land types after investigation and validation. Then, the remote sensing decision table (RSDT) is constructed by linking condition set with decision set according to the sequential pixels in RSI. Fourth, we create one forward greedy searching algorithm based on VPRS to handle this RSDT. After adjusting parameters such as β and knowledge granularity diameter (KGD), we obtain the stable optimized results. Comparative experiments and evaluation show that the discrimination or retrieval accuracy of VPRS model is satisfying (overall accuracy: 87.32% and KHAT: 0.84) and better than original data. Moreover, data dimension has been decreased dramatically (from 12 to 3) and key attributes found by the model may be useful for specific retrieval in wetland inventories.     

Landsat Thematic Mapper: Detection of Shifts in Community Composition of Coral Reefs

Abstract: We assembled a time series of 20 Landsat thematic mapper images from 1982 to 1996 for Key Largo, Florida, to ascertain whether satellite imagery can detect temporal changes in coral reef communities. Selected reef and control areas were examined for changes in brightness, spectral reflectance, band ratios, spatial texture, and temporal texture (  pixel-to-pixel change over time). We compared the data to known changes in the reef ecosystem of Carysfort Reef and terrestrial sample sites. Changes in image brightness and spectral-band ratios were suggestive of shifts from coral- to algal-dominated community structure, but the trends were not statistically significant. The spatial heterogeneity of the reef community decreased in the early 1980s at scales consistent with known ecological changes to the coral community on Carysfort Reef. An analysis of pixel-scale variation through time, termed temporal texture, revealed that the shallow reef areas are the most variable in regions of the reef that have experienced significant ecological decline. Thus, the process of reef degradation, which alters both the spatial patterning and variability of pixel brightness, can be identified in unclassified thematic mapper images.     

Classification of floristic composition of mangrove forests using hyperspectral data: case study of Bhitarkanika National Park, India

The aim of the present work is to unveil the potential of some of the unexplored remote sensing techniques for mangrove studies. The paper deals with the classification of an Earth Observing–1 Hyperion image of the mangrove area of Bhitarkanika National Park, Odisha, India into mangrove floristic composition classes. Out of 196 calibrated bands of the image, 56 were found to be highly uncorrelated and contained maximum information; therefore, these 56 bands were used for classification. Amongst the three full–pixel classifiers tested in the investigation, Support Vector Machine produced the best results in terms of training pixel accuracy with overall precision of 96.85 %, in comparison to about 70–72.0 % for the other two classifiers. A total of five mangrove classes were obtained – pure or dominant class of Heritiera fomes, mixed class of H. fomes, mixed Excoecaria agallocha with Avicennia officinalis, mixed class of fringing Sonneratia apetala and class comprising of mangrove associates with salt resistant grasses. Post–classification field data also established the same. Pure or dominant classes of H. fomes occupied more than 50 % of the total mangrove vegetation in the forest blocks of the National Park. Spectral profile matching of image pixels with that of in–situ collected canopy reflectance profile revealed good match for H. fomes (pure or dominant stands). Red–edge index, which was a preferred criterion for matching was notably correlated in case of H. fomes and E. agallocha. The outcomes indicated the efficacy of hyperspectral canopy reflectance library for such kind of work. It is hoped that the methodology presented in this paper will prove to be useful and may be followed for producing mangrove floristic maps at finer levels.     

Study on the utility of IRS 1D LISS-III data and the classification techniques for mapping of Sunderban mangroves

Mangrove conservation and management is a stupendous task chiefly due to the inaccessibility and the hostile substrate conditions. Remote sensing technology serves as an important tool in providing fast, accurate and up-to-date baseline information on the status of mangroves. It is almost impossible to carry out conventional field surveys in these swampy areas. The present study aims at the classification and mapping of the mangroves in Sunderban Biosphere Reserve (SBR) in the West Bengal province of India using IRS 1D LISS-III satellite data. Different classification approaches, viz., on-screen visual interpretation, supervised and unsupervised classifications were tried. The study showed that four mangroves classes, viz., Avicennia, Phoenix, mixed mangroves, and mangrove scrub and eight non-mangrove classes could be delineated using all the three approaches. All the mangrove and non-mangrove classes were field verified and the overall accuracy as well as user’s and producer’s accuracies for each category were determined. It was observed that among the three approaches, on-screen visual interpretation yielded higher classification accuracy (91.67%) compared to supervised (79.90%) and unsupervised classifications (71.08%). The results obtained through on-screen visual interpretation showed that all mangrove categories together cover 23.21% of the total geographical area of SBR, of which the mixed mangrove category covers maximum area (18.31%). Among the non-mangrove classes, the waterbody occupies largest area (35.36%) followed by agriculture (34.51%).     

A study on abundance and distribution of mangrove species in Indian Sundarban using remote sensing technique

Conservation and management of Sundarban mangrove forest is difficult chiefly due to inaccessibility and hostile condition. Remote sensing serves as an important tool to provide up-to date baseline information which is the primary requirement for the conservation planning of mangroves. In this study, supervised classification by maximum likelihood classifier (MLC) has been used to classify LANDSAT TM and LANDSAT ETM satellite data. This algorithm is used for computing likelihood of unknown measurement vector belonging to unknown classes based on Bayesian equation. Image spectra for various mangrove species were also generated from hyperspectral image. During field visits, GPS locations of five dominant mangrove species with appreciable distribution were taken and image spectra were generated for the same points from hyperion image. The result of this classification shows that, in 1999 total mangrove forest accounted for 55.01 % of the study area which has been reduced to 50.63 % in the year 2010. Avicennia sp. is found as most dominating species followed by Excoecaria sp. and Phoenix sp. but the aerial distribution of Avicennia sp., Bruguiera sp. and Ceriops sp. has reduced. In this classification technique the overall accuracy and Kappa value for 1999 and 2010 are 80 % and 0.77, 85.71 % and 0.81 respectively.     

Supporting habitat conservation with automated change detection in Google Earth Engine

A significant limitation in biodiversity conservation has been the effective implementation of laws and regulations that protect species’ habitats from degradation. Flexible, efficient, and effective monitoring and enforcement methods are needed to help conservation policies realize their full benefit. As remote sensing data become more numerous and accessible, they can be used to identify and quantify land-cover changes and habitat loss. However, these data remain underused for systematic conservation monitoring in part because of a lack of simple tools. We adapted 2 algorithms that automatically identify differences between pairs of images. We used free, publicly available satellite data to evaluate their ability to rapidly detect land-cover changes in a variety of land-cover types. We compared algorithm predictions with ground-truthed results at 100 sites of known change in the United States. We also compared algorithm predictions to manually created polygons delineating anthropogenic change in 4 case studies involving imperiled species’ habitat: oil and gas development in the range of the Greater Sage Grouse (Centrocercus urophasianus); sand mining operations in the range of the dunes sagebrush lizard (Sceloporus arenicolus); loss of Piping Plover (Charadrius melodus) coastal habitat after Hurricane Michael (2018); and residential development in St. Andrew beach mouse (Peromyscus polionotus peninsularis) habitat. Both algorithms effectively discriminated between pixels corresponding to land-cover change and unchanged pixels as indicated by area under a receiver operating characteristic curve >0.90. The algorithm that was most effective differed among the case-study habitat types, and both effectively delineated habitat loss as indicated by low omission (min. = 0.0) and commission (min. = 0.0) rates, and moderate polygon overlap (max. = 47%). Our results showed how these algorithms can be used to help close the implementation gap of monitoring and enforcement in biodiversity conservation. We provide a free online tool that can be used to run these analyses ( https://conservationist.io/habitatpatrol ).     

Study of the dynamics of estuarine and coastal waters using remote sensing: the Tinto-Odiel estuary,SW Spain

This paper deals with the application of satellite images to characterize some aspects of the circulation dynamics of the Tinto-Odiel estuary using turbidity patterns as ‘natural tracers’. 15 images (Landsat TM and Spot HRV) were processed to provide synoptic, instantaneous views of the circulation patterns under different environmental conditions. In addition, a comparison was made between results of oceanographic field work, using biplanes and fluorescent tracers, and satellite image turbidity patterns used as ‘ground truth’ data for specific hydroclimatic situations. This approach allowed (1) the identification and mapping of dynamic processes of interest during a theoretical tidal cycle, (2) the elaboration of additional information on the ‘flow schemes’ at the mouth of the estuary with improved spatial and temporal resolution, and (3) the supply of basic data to improve the knowledge of exchange processes between estuarine and coastal waters. The results of this study are considered to be useful for the management of the estuarine system.     

Study of the dynamics of estuarine and coastal waters using remote sensing: the Tinto-Odiel estuary, SW Spain

This paper deals with the application of satellite images to characterize some aspects of the circulation dynamics of the Tinto-Odiel estuary using turbidity patterns as ‘natural tracers’. 15 images (Landsat TM and Spot HRV) were processed to provide synoptic, instantaneous views of the circulation patterns under different environmental conditions. In addition, a comparison was made between results of oceanographic field work, using biplanes and fluorescent tracers, and satellite image turbidity patterns used as ‘ground truth’ data for specific hydroclimatic situations. This approach allowed (1) the identification and mapping of dynamic processes of interest during a theoretical tidal cycle, (2) the elaboration of additional information on the ‘flow schemes’ at the mouth of the estuary with improved spatial and temporal resolution, and (3) the supply of basic data to improve the knowledge of exchange processes between estuarine and coastal waters. The results of this study are considered to be useful for the management of the estuarine system.     

Crustacean habitat potential mapping in a tidal flat using remote sensing and GIS

To make a macrofaunal (crustacean) habitat potential map, the spatial distribution of ecological variables in the Hwangdo tidal flat, Korea, was explored. Spatial variables were mapped using remote sensing and a geographic information system (GIS) combined with field observations. A frequency ratio (FR) and logistic regression (LR) model were employed to map the macrofauna potential area for the Ilyoplax dentimerosa, a crustacean species. Spatial variables affecting the tidal macrofauna distribution were selected based on abundance and biomass and used within a spatial database derived from remotely sensed data of various types of sensors. The spatial variables included the intertidal digital elevation model (DEM), slope, distance from a tidal channel, tidal channel density, surface sediment facies, spectral reflectance of the near infrared (NIR) bands and the tidal exposure duration. The relation between the I. dentimerosa and each spatial variable was calculated using the FR and LR. The species was randomly divided into a training set (70%) to analyse habitat potential using FR and LR and a test set (30%) to validate the predicted habitat potential map. The relations were overlaid to produce a habitat potential map with the species potential index (SPI) value for each pixel. The potential habitat maps were compared with the surveyed habitat locations such as validation data set. The comparison results showed that the LR model (accuracy is 85.28%) is better in prediction than the FR (accuracy is 78.96%) model. The performance of models gave satisfactory accuracies. The LR provides the quantitative influence of variables on a potential habitat of species; otherwise, the FR shows the quantitative influence of a class in each variable. The combination of a GIS-based frequency ratio and logistic regression models and remote sensing with field observations is an effective method to determine locations favorable for macrofaunal species occurrences in a tidal flat.     

Modeling the dynamic habitat and breeding population of Southwestern Willow Flycatcher

To aid in the management and conservation of Southwestern Willow Flycatcher (Empidonax traillii extimus, hereafter “Flycatcher”), we developed numerous models of flycatcher breeding habitat at Roosevelt Lake, AZ. For model development and testing, we compiled 10 years of flycatcher territory data that were obtained from intensive fieldwork between 1996 and 2005. We identified riparian vegetation annually in the project area from Landsat Thematic Mapper images, and extracted floodplain features from a digital elevation model. We created a novel class of temporal (i.e., multiyear) variables by characterizing the stability and variability in breeding habitat over a 6-year time interval. We used logistic regression to determine associations between environmental variables and flycatcher territory occurrence, and to test specific hypotheses. We mapped the probability of territory occurrence with a GIS and determined model accuracies with a classification table and a 10-year population database. Environmental features that were associated with breeding flycatchers included floodplain size, proximity to water, and the density, heterogeneity, age and stability of riparian vegetation. Our best model explained 79% of the variability in the flycatcher breeding population at Roosevelt Lake. The majority of predicted flycatcher habitat formed between 1996 and 2004 on an exposed lakebed ～3 years after water levels receded during a prolonged drought. A high correlation between annual reservoir levels and predicted breeding habitat (r = ?0.82) indicates that we can create and manage habitat for conservation purposes. Our predictive models quantify and assess the relative quality of flycatcher breeding habitat remotely, and can be used to evaluate the effectiveness of habitat restoration activities. Numerous techniques we developed can be used to characterize riparian vegetation and patch dynamics directly off of satellite imagery, thereby increasing its utility for conservation purposes.     

Using Presence-Absence Data to Build and Test Spatial Habitat Models for the Fisher in the Klamath Region, U.S.A.

Abstract: Forest carnivores such as the fisher ( Martes pennanti ) have frequently been the target of conservation concern because of their association in some regions with older forests and sensitivity to landscape-level habitat alteration. Although the fisher has been extirpated from most of its former range in the western United States, it is still found in northwestern California. Fisher distribution, however, is still poorly known in most of this region where surveys have not been conducted. To predict fisher distribution across the region, we created a multiple logistic regression model using data from 682 previously surveyed locations and a vegetation layer created from satellite imagery. A moving-window function in a geographic information system was used to derive landscape-level indices of canopy closure, tree size class, and percent conifer. The model was validated with new data from 468 survey locations. The correct classification rate of 78.6% with the new data was similar to that achieved with the original data set (80.4%). Whereas several fine-scale habitat attributes were significantly correlated with fisher presence, the multivariate model containing only landscape- and regional-scale variables performed as well as one incorporating fine-scale data, suggesting that habitat selection by fishers may be dominated by factors operating at the home-range scale and above. Fisher distribution was strongly associated with landscapes with high levels of tree canopy closure. Regional gradients such as annual precipitation were also significant. At the plot level, the diameter of hardwoods was greater at sites with fisher detections. A comparison of regional fisher distribution with land-management categories suggests that increased emphasis on the protection of biologically productive, low- to mid-elevation forests is important to ensuring the long-term viability of fisher populations.