The spatio-temporal characteristics of remote sensing are considered to be the primary advantage in environmental studies. With long-term and frequent satellite observations, it is possible to monitor changes in key biophysical attributes such as phenological characteristics, and relate them to climate change by examining their correlations. Although a number of remote sensing methods have been developed to quantify vegetation seasonal cycles using time-series of vegetation indices, there is limited effort to explore and monitor changes and trends of vegetation phenology in the Monsoon Southeast Asia, which is adversely affected by changes in the Asian monsoon climate. In this study, MODIS EVI and TRMM time series data, along with field survey data, were analyzed to quantify phenological patterns and trends in the Monsoon Southeast Asia during 2001-2010 period and assess their relationship with climate change in the region. The results revealed a great regional variability and inter-annual fluctuation in vegetation phenology. The phenological patterns varied spatially across the region and they were strongly correlated with climate variations and land use patterns. The overall phenological trends appeared to shift towards a later and slightly longer growing season up to 14 days from 2001 to 2010. Interestingly, the corresponding rainy season seemed to have started earlier and ended later, resulting in a slightly longer wet season extending up to 7 days, while the total amount of rainfall in the region decreased during the same time period. The phenological shifts and changes in vegetation growth appeared to be associated with climate events such as EL Niño in 2005. Furthermore, rainfall seemed to be the dominant force driving the phenological changes in naturally vegetated areas and rainfed croplands, whereas land use management was the key factor in irrigated agricultural areas.
In the remote sensing domain, it is crucial to complete semantic segmentation on the raster images, e.g., river, building, forest, etc., on raster images. A deep convolutional encoder–decoder (DCED) network is the state-of-the-art semantic segmentation method for remotely sensed images. However, the accuracy is still limited, since the network is not designed for remotely sensed images and the training data in this domain is deficient. In this paper, we aim to propose a novel CNN for semantic segmentation particularly for remote sensing corpora with three main contributions. First, we propose applying a recent CNN called a global convolutional network (GCN), since it can capture different resolutions by extracting multi-scale features from different stages of the network. Additionally, we further enhance the network by improving its backbone using larger numbers of layers, which is suitable for medium resolution remotely sensed images. Second, “channel attention” is presented in our network in order to select the most discriminative filters (features). Third, “domain-specific transfer learning” is introduced to alleviate the scarcity issue by utilizing other remotely sensed corpora with different resolutions as pre-trained data. The experiment was then conducted on two given datasets: (i) medium resolution data collected from Landsat-8 satellite and (ii) very high resolution data called the ISPRS Vaihingen Challenge Dataset. The results show that our networks outperformed DCED in terms of F 1 for 17.48% and 2.49% on medium and very high resolution corpora, respectively.
Object segmentation of remotely-sensed aerial (or very-high resolution, VHS) images and satellite (or high-resolution, HR) images, has been applied to many application domains, especially in road extraction in which the segmented objects are served as a mandatory layer in geospatial databases. Several attempts at applying the deep convolutional neural network (DCNN) to extract roads from remote sensing images have been made; however, the accuracy is still limited. In this paper, we present an enhanced DCNN framework specifically tailored for road extraction of remote sensing images by applying landscape metrics (LMs) and conditional random fields (CRFs). To improve the DCNN, a modern activation function called the exponential linear unit (ELU), is employed in our network, resulting in a higher number of, and yet more accurate, extracted roads. To further reduce falsely classified road objects, a solution based on an adoption of LMs is proposed. Finally, to sharpen the extracted roads, a CRF method is added to our framework. The experiments were conducted on Massachusetts road aerial imagery as well as the Thailand Earth Observation System (THEOS) satellite imagery data sets. The results showed that our proposed framework outperformed Segnet, a state-of-the-art object segmentation technique, on any kinds of remote sensing imagery, in most of the cases in terms of precision, recall, and F1.
Abstract:Object segmentation on remotely-sensed images: aerial (or very high resolution, VHS) images and satellite (or high resolution, HR) images, has been applied to many application domains, especially road extraction in which the segmented objects are served as a mandatory layer in geospatial databases. Several attempts in applying deep convolutional neural network (DCNN) to extract roads from remote sensing images have been made; however, the accuracy is still limited. In this paper, we present an enhanced DCNN framework specifically tailored for road extraction on remote sensing images by applying landscape metrics (LMs) and conditional random fields (CRFs). To improve DCNN, a modern activation function, called exponential linear unit (ELU), is employed in our network resulting in a higher number of and yet more accurate extracted roads. To further reduce falsely classified road objects, a solution based on an adoption of LMs is proposed. Finally, to sharpen the extracted roads, a CRF method is added to our framework. The experiments were conducted on Massachusetts road aerial imagery as well as THEOS satellite imagery data sets. The results showed that our proposed framework outperformed Segnet, the state-of-the-art object segmentation technique on any kinds of remote sensing imagery, in most of the cases in terms of precision, recall, and F1.
Background Tak Province, at the Thai–Myanmar border, is one of three high malaria incidence areas in Thailand. This study aimed to describe and identify possible factors driving the spatiotemporal trends of disease incidence from 2012 to 2015. Methods Climate variables and forest cover were correlated with malaria incidence using Pearson’s r . Statistically significant clusters of high (hot spots) and low (cold spots) annual parasite incidence per 1000 population (API) were identified using Getis-Ord Gi* statistic. Results The total number of confirmed cases declined by 76% from 2012 to 2015 ( Plasmodium falciparum by 81%, Plasmodium vivax by 73%). Incidence was highly seasonal with two main annual peaks. Most cases were male (62.75%), ≥ 15 years (56.07%), and of Myanmar (56.64%) or Thai (39.25%) nationality. Median temperature (1- and 2-month lags), average temperature (1- and 2-month lags) and average relative humidity (2- and 3-month lags) correlated positively with monthly total, P. falciparum and P. vivax API. Total rainfall in the same month correlated with API for total cases and P. vivax but not P. falciparum . At sub-district level, percentage forest cover had a low positive correlation with P. falciparum , P. vivax , and total API in most years. There was a decrease in API in most sub-districts for both P. falciparum and P. vivax . Sub-districts with the highest API were in the Tha Song Yang and Umphang Districts along the Thai–Myanmar border. Annual hot spots were mostly in the extreme north and south of the province. Conclusions There has been a large decline in reported clinical malaria from 2012 to 2015 in Tak Province. API was correlated with monthly climate and annual forest cover but these did not account for the trends over time. Ongoing elimination interventions on one or both sides of the border are more likely to have been the cause but it was not possible to assess this due to a lack of suitable data. Two main hot spot areas were identified that could be targeted for intensified elimination activities. Electronic supplementary material The online version of this article (10.1186/s12936-019-2871-2) contains supplementary material, which is available to authorized users.
. In moving toward malaria elimination, finer scale malaria risk maps are required to identify hotspots for implementing surveillance–response activities, allocating resources, and preparing health facilities based on the needs and necessities at each specific area. This study aimed to demonstrate the use of multi-criteria decision analysis (MCDA) in conjunction with geographic information systems (GISs) to create a spatial model and risk maps by integrating satellite remote-sensing and malaria surveillance data from 18 counties of Yunnan Province along the China–Myanmar border. The MCDA composite and annual models and risk maps were created from the consensus among the experts who have been working and know situations in the study areas. The experts identified and provided relative factor weights for nine socioeconomic and disease ecology factors as a weighted linear combination model of the following: ([Forest coverage × 0.041] + [Cropland × 0.086] + [Water body × 0.175] + [Elevation × 0.297] + [Human population density × 0.043] + [Imported case × 0.258] + [Distance to road × 0.030] + [Distance to health facility × 0.033] + [Urbanization × 0.036]). The expert-based model had a good prediction capacity with a high area under curve. The study has demonstrated the novel integrated use of spatial MCDA which combines multiple environmental factors in estimating disease risk by using decision rules derived from existing knowledge or hypothesized understanding of the risk factors via diverse quantitative and qualitative criteria using both data-driven and qualitative indicators from the experts. The model and fine MCDA risk map developed in this study could assist in focusing the elimination efforts in the specifically identified locations with high risks.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.