“…Central pivot systems are commonly utilized in agricultural irrigation practices across the United States. In the past, many studies have focused more on fields served by central pivot and less on fields served by other types of irrigation, which may be because of its obvious characteristics and because it is relatively easily identified [16][17][18][19]53,62]. This study demonstrates that deep learning techniques could efficiently extract irrigated croplands, regardless of the type of irrigation being utilized.…”
Section: Discussionmentioning
confidence: 90%
“…Moreover, Xie and Lark [18] refined the process of collecting training data by estimating the ideal thresholds for crop greenness and gathering data on the status of center-pivot-irrigated and non-irrigated fields. Ren et al [19] mapped irrigated and non-irrigated corn for Nebraska at a 30 m resolution by employing training data generated from the MODIS-derived irrigated and non-irrigated map and images from Landsat data. Magidi et al [20] mapped irrigated areas using Landsat and Sentinel-2 images as well as Google Earth Engine.…”
Understanding the spatial distribution of irrigated croplands is crucial for food security and water use. To map land cover classes with high-spatial-resolution images, it is necessary to analyze the semantic information of target objects in addition to the spectral or spatial–spectral information of local pixels. Deep convolutional neural networks (DCNNs) can characterize the semantic features of objects adaptively. This study uses DCNNs to extract irrigated croplands from Sentinel-2 images in the states of Washington and California in the United States. We integrated the DCNNs of 101 layers, discarded pooling layers, and employed dilation convolution to preserve location information; these are models which were used based on fully convolutional network (FCN) architectures. The findings indicated that irrigated croplands may be effectively detected at various phases of crop growth in the fields. A quantitative analysis of the trained models revealed that the three models in the two states had the lowest values of Intersection over Union (IoU) and Kappa, i.e., 0.88 and 0.91, respectively. The deep models’ temporal portability across different years was acceptable. The lowest values of recall and OA (overall accuracy) from 2018 to 2021 were 0.91 and 0.87, respectively. In Washington, the lowest OA value from 10 to 300 m resolution was 0.76. This study demonstrates the potential of FCNs + DCNNs approaches for mapping irrigated croplands across large regions, providing a solution for irrigation mapping. The spatial resolution portability of deep models could be improved further by designing model architectures.
“…Central pivot systems are commonly utilized in agricultural irrigation practices across the United States. In the past, many studies have focused more on fields served by central pivot and less on fields served by other types of irrigation, which may be because of its obvious characteristics and because it is relatively easily identified [16][17][18][19]53,62]. This study demonstrates that deep learning techniques could efficiently extract irrigated croplands, regardless of the type of irrigation being utilized.…”
Section: Discussionmentioning
confidence: 90%
“…Moreover, Xie and Lark [18] refined the process of collecting training data by estimating the ideal thresholds for crop greenness and gathering data on the status of center-pivot-irrigated and non-irrigated fields. Ren et al [19] mapped irrigated and non-irrigated corn for Nebraska at a 30 m resolution by employing training data generated from the MODIS-derived irrigated and non-irrigated map and images from Landsat data. Magidi et al [20] mapped irrigated areas using Landsat and Sentinel-2 images as well as Google Earth Engine.…”
Understanding the spatial distribution of irrigated croplands is crucial for food security and water use. To map land cover classes with high-spatial-resolution images, it is necessary to analyze the semantic information of target objects in addition to the spectral or spatial–spectral information of local pixels. Deep convolutional neural networks (DCNNs) can characterize the semantic features of objects adaptively. This study uses DCNNs to extract irrigated croplands from Sentinel-2 images in the states of Washington and California in the United States. We integrated the DCNNs of 101 layers, discarded pooling layers, and employed dilation convolution to preserve location information; these are models which were used based on fully convolutional network (FCN) architectures. The findings indicated that irrigated croplands may be effectively detected at various phases of crop growth in the fields. A quantitative analysis of the trained models revealed that the three models in the two states had the lowest values of Intersection over Union (IoU) and Kappa, i.e., 0.88 and 0.91, respectively. The deep models’ temporal portability across different years was acceptable. The lowest values of recall and OA (overall accuracy) from 2018 to 2021 were 0.91 and 0.87, respectively. In Washington, the lowest OA value from 10 to 300 m resolution was 0.76. This study demonstrates the potential of FCNs + DCNNs approaches for mapping irrigated croplands across large regions, providing a solution for irrigation mapping. The spatial resolution portability of deep models could be improved further by designing model architectures.
“…Indeed, as IAP tend to avoid the overlap of blooming and vegetative periods with native vegetation, temporal variability of spectral values may allow the discrimination between IAP and native species [66,67]. Most of the multispectral satellitebased studies of IAP on coastal areas are supported by Landsat data [68][69][70][71]. Landsat mission with 30 m spatial resolution, launched for the first time in the early 1970s, offers the longest temporal series with adequate spectral resolutions (11 bands in Landsat-8) [63].…”
Coastal environments are highly threatened by invasive alien plants (IAP), and Remote Sensing (RS) may offer a sound support for IAP detection and mapping. There is still a need for an overview of the progress and extent of RS applications on invaded coasts that can help the development of better RS procedures to support IAP management. We conducted a systematic literature review of 68 research papers implementing, recommending, or discussing RS tools for IAP mapping in coastal environments, published from 2000 to 2021. According to this review, most research was done in China and USA, with Sporobolus (17.3%) being the better studied genus. The number of studies increased at an accelerated rate from 2015 onwards, coinciding with the transition from RS for IAP detection to RS for invasion modeling. The most used platforms in the 2000s were aircraft, with satellites that increased from 2005 and unmanned aerial vehicles after 2014. Frequentist inference was the most adopted classification approach in the 2000s, as machine learning increased after 2009. RS applications vary with coastal ecosystem types and across countries. RS has a huge potential to further improve IAP monitoring. The extension of RS to all coasts of the world requires advanced applications that bring together current and future Earth observation data.
“…The Support Vector Machine (SVM) algorithm is also found to be one of the most widely used algorithms in land use and land cover classification (Huang et al, 2002;Otukei and Blaschke, 2010). For a given land cover mapping task, it is often a good practice to examine multiple machine learning algorithms to compare and select a better performer (Ren et al, 2021).…”
Section: Land Use and Land Cover (Lulc) Classification And Accuracy A...mentioning
Coastal erosion is one of the most significant environmental threats to coastal communities globally. In Bangladesh, coastal erosion is a regularly occurring and major destructive process, impacting both human and ecological systems at sea level. The Lower Meghna estuary, located in southern Bangladesh, is among the most vulnerable landscapes in the world to the impacts of coastal erosion. Erosion causes population displacement, loss of productive land area, loss of infrastructure and communication systems, and, most importantly, household livelihoods. With an aim to assess the impacts of historical and predicted shoreline change on different land use and land cover, this study estimated historical shoreline movement, predicted shoreline positions based on historical data, and quantified and assessed past land use and land cover change. Multi-temporal Landsat images from 1988–2021 were used to quantify historical shoreline movement and past land use and land cover. A time-series classification of historical land use and land cover (LULC) were produced to both quantify LULC change and to evaluate the utility of the future shoreline predictions for calculating amounts of lost or newly added land resources by LULC type. Our results suggest that the agricultural land is the most dominant land cover/use (76.04% of the total land loss) lost over the studied period. Our results concluded that the best performed model for predicting land loss was the 10-year time depth and 20-year time horizon model. The 10-year time depth and 20-year time horizon model was also most accurate for agricultural, forested, and inland waterbody land use/covers loss prediction. We strongly believe that our results will build a foundation for future research studying the dynamics of coastal and deltaic environments.
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.