Producer's Accuracy Research Articles

EXPLORING THE HARMONIZED LANDSAT SENTINEL (HLS) DATACUBE TO MAP AN AGRICULTURAL LANDSCAPE IN THE BRAZILIAN SAVANNA

Abstract. Brazil has established itself as one of the world leaders in food production. Different types of remote sensing mapping techniques have been undertaken to support rural planning in the country. However, due to the complex dynamics of Brazilian agriculture, especially in the Cerrado biome (tropical savanna), there is a need for more feasible crop discrimination and monitoring initiatives, which require a consistent time series of remote sensing data at medium meter and potentially up to 3 day Landsat 8 and Sentinel-2 satellite time series, minimizing the cloud cover limitations for rainfed agricultural monitoring. This paper aims to explore the potential of the Harmonized Landsat 8 Sentinel-2 (HLS) data cube to map agricultural landscapes in the Brazilian Cerrado. The HLS multispectral bands from 27 scenes with less than 10% cloud cover, from October 2020 to September 2021, encompassing one entire crop growing season, were processed by the Random Forest algorithm to produce a map with four land use/cover classes (annual crops, sugarcane, renovated sugarcane fields, cultivated pastures, and native Cerrado). We performed accuracy assessment through 10-fold cross-validation and confusion matrix analyses. The results showed a high level of overall accuracy and Kappa coefficient, both with 99%, as well as high user's and producer's accuracies of at least 99%. The HLS dataset has been continuously improved, showing very promising results for rainfed agricultural mapping and monitoring.

Assessment of Sentinel‐2A images for estimating rosemary land cover through an object‐based image analysis approach

AbstractRosemary is a medicinal plant widely used in the food and pharmaceutical industry. In Morocco, rosemary harvesting generates significant benefits for the local economy. Even though, decision makers are short of accurate information about its geographical distribution area. The aim of this study is to delineate the rosemary shrubs in the Oriental region in Morocco in order to provide guidance for the sustainable use of these natural resources. We therefore used the freely available images of Sentinel‐2A for binary classification (rosemary shrubs and the other lands) through an object‐based image‐analysis approach. The random forest classifier was deployed and trained with different object features likewise spectral values and vegetation indices. As a result, the distribution area of rosemary is 240,747 ha, which represents 18% of the studied area. Besides, an encouraging overall accuracy of 94.24% was attained; the user's accuracy for the class of “rosemary” and “Others” is 93.3% and 95.0%, respectively; meanwhile, the producer's accuracy for “rosemary” and “Others” is 93.72% and 94.65%.

Impact assessment of humanitarian demining using object-based peri-urban land cover classification and morphological building detection from VHR Worldview imagery

This study utilised very high-resolution (VHR) imagery to monitor and evaluate the impact of humanitarian demining activities in a peri-urban environment in Afghanistan. Identifying buildings and mapping the spatial distribution of different types of land cover is of great practical significance for demining organisations, such as the HALO Trust, to assess the impact of their mine-clearance activities, by quantifying change and the growth of a population in a specific area over time. This study had two main objectives: (i) to map the post-clearance land cover, and (ii) to detect and quantify the change in the number and area of buildings. Two independent workflows were implemented and evaluated. To map land cover, this study investigated the implementation of various machine learning algorithms in object-based image classification (OBIA) of VHR satellite imagery (Worldview-1,2,3). Image segmentation was carried out using the Large-Scale Mean-Shift (LSMS) algorithm, before classification was performed based on a machine learning Random Forests (RF) approach. Different parameters and spatial distribution of training samples were tested to analyse the model's performance. Further analysis determined that by using only the segments' mean value per spectral band (Red, Green, Blue), data redundancy in the training stage was eliminated. The final classified map had an overall accuracy of 90.67% and a total built-area of 643,660.28 m2 was detected in the 4.11 km2 study area. To detect and quantify buildings present in the study area, an alternative, automatic, unsupervised approach based on the morphological building index (MBI) was implemented using MATLAB. Two VHR (0.5 m) panchromatic images acquired by WorldView-1 in 2008 and 2018 were processed using a series of multi-scale and multi-directional morphological operators, before a series of post-processing thresholds were applied to refine the output. Parameters were systematically optimised for the datasets and their sensitivity analysed. By comparing the output to manually labelled reference maps, producer's and user's accuracy of 78.63/81.70% and 75.30/77.91% were attained for 2008 and 2018 imagery respectively. The built-area was found to increase from 58,705 to 611,920 m2 over the ten-year period, indicating a significant increase in the number of people resident in the area. Whilst OBIA was found to be more accurate than MBI for the 2018 pansharpened imagery, with producer's and user's accuracy of 91.00/90.60% versus 84.65/79.44% respectively, it must be considered that the MBI, an unsupervised method requiring no training, offers a fast solution if the sole objective is building detection.

Optical time series for the separation of land cover types with similar spectral signatures: cocoa agroforest and forest

ABSTRACT One of the main applications of machine learning (ML) in remote sensing (RS) is the pixel-level classification of satellite images into land cover types. Although classes with different spectral signatures can be easily separated, e.g. aquatic and terrestrial land cover types, others have similar spectral signatures and are hard to separate using only the information within a single pixel. This work focused on the separation of two cover types with similar spectral signatures, cocoa agroforest and forest, over an area in Pará, Brazil. For this, we study the training and application of several ML algorithms on datasets obtained from a single composite image, a time-series (TS) composite obtained from the same location and by preprocessing the TS composite using simple TS preprocessing techniques. As expected, when ML algorithms are applied to a dataset obtained from a composite image, the median producer’s accuracy (PA) and user’s accuracy (UA) in those two classes are significantly lower than the median overall accuracy (OA) for all classes. The second dataset allows the ML models to learn the evolution of the spectral signatures over 5 months. Compared to the first dataset, the results indicate that ML models generalize better using TS data, even if the series are short and without any preprocessing. This generalization is further improved in the last dataset. The ML models are subsequently applied to an area with different geographical bounds. These last results indicate that, out of seven classifiers, the popular random forest (RF) algorithm ranked fourth, while XGBoost (×GB) obtained the best results. The best OA, as well as the best PA/UA balance, were obtained by performing feature construction using the M3GP algorithm and then applying XGB to the new extended dataset.

Cloud detection for satellite cloud images based on fused FCN features

ABSTRACT Cloud detection for satellite cloud images is a challenging image processing task owing to the blurring of cloud boundaries, multiplicity, and complexity of cloud types. Currently, the commonly used cloud detection methods include original full convolutional neural network (original FCN), FCN with an 8-pixel stride (FCN- 8s), FCN with a 2-pixel stride (FCN-2s), and so on. However, the aforementioned methods exclusively rely on a single network layer the final layer feature map; thus, shallow cloud image information, such as cloud profile information may not be captured. In this letter, a cloud detection method for satellite cloud images based on fused FCN features is proposed. The proposed method effectively fuses spatial and high-level semantic information, and a voting ensemble strategy is used to improve the accuracy and robustness of cloud detection. Finally, the experimental results demonstrate that the average overall accuracy (OA), average producer’ accuracy (PA), and average user’ accuracy (UA) of the proposed method for multiple training sample sizes and image sizes of the collected Fengyun satellite (FY-2 G) cloud image database increased by 7.15%, 9.04%, and 8.46%, respectively, relative to the average accuracies of the original FCN, FCN-8s, FCN-2s, SegNet, and DeepLabV3 methods.

High-resolution map of sugarcane cultivation in Brazil using a phenology-based method

Abstract. Sugarcane is the most important source of sugar, and its cultivation area has undergone rapid expansion, replacing other crops, pastures, and forests. Brazil is the world's largest sugarcane producer and contributed to approximately 38.6 % of the world's total production in 2019. Sugarcane in Brazil can be harvested from April to December in the south-central area and from September to April in the northeast area. The flexible phenology and harvest conditions of sugarcane in Brazil make it difficult to identify the harvest area at state to country scales. In this study, we developed a phenology-based method to identify the harvest area of sugarcane in Brazil by incorporating the multiple phenology conditions into a time-weighted dynamic time warping method (TWDTW). Then, we produced annual 30 m spatial resolution sugarcane harvest maps (2016–2019) for 14 states in Brazil (over 98 % of the harvest area) based on the proposed method using Landsat-7, Landsat-8, and Sentinel-2 optical data. The proposed method performed well in identifying sugarcane harvest area with limited training sample data. Validations for the 2018 harvest year displayed high accuracy, with the user's, producer's, and overall accuracies of 94.35 %, 87.04 %, and 91.47 % in Brazil, respectively. In addition, the identified harvest area of sugarcane exhibited good correlations with the agricultural statistical data provided by the Brazilian Institute of Geography and Statistics (IBGE) at the municipality, microregion, and mesoregion levels. The 30 m Brazil sugarcane harvest maps can be obtained at https://doi.org/10.6084/m9.figshare.14213909 (Zheng et al., 2021).

Detection and characterization of coastal tidal wetland change in the northeastern US using Landsat time series

Coastal tidal wetlands are highly altered ecosystems exposed to substantial risk due to widespread and frequent land-use change coupled with sea-level rise, leading to disrupted hydrologic and ecologic functions and ultimately, significant reduction in climate resiliency. Knowing where and when the changes have occurred, and the nature of those changes, is important for coastal communities and natural resource management. Large-scale mapping of coastal tidal wetland changes is extremely difficult due to their inherent dynamic nature. To bridge this gap, we developed an automated algorithm for DEtection and Characterization of cOastal tiDal wEtlands change (DECODE) using dense Landsat time series. DECODE consists of three elements, including spectral break detection, land cover classification and change characterization. DECODE assembles all available Landsat observations and introduces a water level regressor for each pixel to flag the spectral breaks and estimate harmonic time-series models for the divided temporal segments. Each temporal segment is classified (e.g., vegetated wetlands, open water, and others – including unvegetated areas and uplands) based on the phenological characteristics and the synthetic surface reflectance values calculated from the harmonic model coefficients, as well as a generic rule-based classification system. This harmonic model-based approach has the advantage of not needing the acquisition of satellite images at optimal conditions (i.e., low tide status) to avoid underestimating coastal vegetation caused by the tidal fluctuation. At the same time, DECODE can also characterize different kinds of changes including land cover change and condition change (i.e., land cover modification without conversion). We used DECODE to track status of coastal tidal wetlands in the northeastern United States from 1986 to 2020. The overall accuracy of land cover classification and change detection is approximately 95.8% and 99.8%, respectively. The vegetated wetlands and open water were mapped with user's accuracy of 94.6% and 99.0%, and producer's accuracy of 98.1% and 93.5%, respectively. The cover change and condition change were mapped with user's accuracy of 68.0% and 80.0%, and producer's accuracy of 80.5% and 97.1%, respectively. Approximately 3283 km2 of the coastal landscape within our study area in the northeastern United States changed at least once (12% of the study area), and condition changes were the dominant change type (84.3%). Vegetated coastal tidal wetland decreased consistently (~2.6 km2 per year) in the past 35 years, largely due to conversion to open water in the context of sea-level rise.

Classification of underwater photogrammetry data for temperate benthic rocky reef mapping

The fine characterization of the substrate is a baseline to thoroughly investigate the relations between organisms and their biotopes. Cutting edge spatial technologies now provide access to accurate information on biotopes and biocenoses both in terrestrial and in marine environments. Photogrammetry is one of them and has recently been applied in submarine environments especially in shallow clear water. In this study, we investigated the potential of photogrammetry to characterise benthic habitats in turbid environments. Although more challenging, turbid environments are more frequent in temperate marine coastal areas. We selected two rocky sites in the bay of Saint-Malo (Brittany, France), differentiated by their level of turbidity, one being a marine site exposed to natural tides (Buharats), while the other (Bizeux) is subjected to both natural tides and artificial currents created by the functioning of a hydroelectric dam. The different substrates observed were classified into eight classes at a centimetre resolution using photogrammetry-based spatial and multispectral predictors. The spatial benthic terrain predictors were derived from a digital surface model (DSM) at various spatial scales, and the multispectral predictors were retrieved from the red-green-blue (RGB, natural colours) orthomosaic imagery. An overall classification was computed for Buharats and Bizeux, with accuracies of 84.76% and 79.54% respectively, revealing a good quality of the substrate classification. The combination of RGB, DSM, and several spatial benthic terrain variables, with a pixel resolution of 5 and 10 mm, and a kernel size of 30, 60 and 90 pixels leads to the best benthic substrate classification (highest overall accuracy). At the class scale, producer's (PA) and user's (UA) accuracy showed that big boulders and field material were correctly distinguished. Small boulders and cobbles, having similar sizes, showed the lowest classification performances. This classification methodology provides new perspectives for mesoscale (100 m2 to 1 km2) semi-automatic mapping of the fine resolution (1 cm) relationship between benthic organisms and their substrate.

Deforestation detection using scattering power decomposition and optimal averaging of volume scattering power in tropical rainforest regions

Synthetic aperture radar (SAR) has been a powerful tool for deforestation detection in tropical rainforests. Polarimetric SAR (POLSAR) data are acquired in a quad-polarization mode, and L-band POLSAR data in particular are one of the few SAR data types that preserve the dielectric properties and structures of the scatterer. POLSAR data consist of 2 × 2 scattering matrices and consequently offer superior target recognition compared with dual-polarization data. In this study, we applied scattering power decomposition suitable for detecting deforestation in near real-time to POLSAR data obtained from the Earth observation sensor Phased Array type L-band Synthetic Aperture Radar-2 (PALSAR-2). The reflection symmetry condition is known to apply to natural distributed objects (i.e., the cross-correlation between co- and cross-polarization data is zero). Inspired by this, we theoretically and experimentally examined the volume scattering power component to distinguish natural forests from the surrounding area. Two important results were verified for natural forests: as the window size of the ensemble average increases, (i) the coherency matrix approaches a simple theoretical form and (ii) the volume scattering power becomes dominant among the scattering power components. Based on these results, we constructed an algorithm that applies scattering power decomposition for detecting deforestation. We produced reference data using high-resolution optical images and evaluated the performance of the derived deforestation map in the Amazon natural forest when employing various window sizes for the ensemble average. At an optimal window size of 15 × 15 pixels, the deforestation detection performance reached a user's accuracy of 94.9% ± 1.5%, a producer's accuracy of 72.3% ± 1.2%, and a kappa coefficient of 0.816 ± 0.0039. Sparse trees left after logging increased the volume scattering power and reduced the producer's accuracy. The proposed algorithm can contribute to deforestation detection with slightly lower accuracy than that of the annual map provided by Global Forest Change. Further, the proposed algorithm is robust to the seasonal variations in tropical rainforests and temporal variations in the deforestation process. Consequently, the proposed algorithm employing the six-component scattering power decomposition method can be utilized in near real-time without considering applicable areas in tropical rainforests. Subsequently, we applied our algorithm to dual-polarization data, which are acquired by PALSAR-2 much more frequently than POLSAR data. The false detection rate did not increase when using the dual-polarization data; however, the omission error increased considerably compared with that when using POLSAR data owing to the low total power obtained from the dual-polarization data.

Large‐scale and fine‐grained mapping of heathland habitats using open‐source remote sensing data

AbstractMapping natural habitats remains challenging, especially at a national scale. Although new open‐access variables for vegetation and its environment and increased spatial resolution derived from satellite remote sensing data are available at the global scale, the relevance of these new variables for fine‐grained mapping of natural habitats at a national scale remains underexplored. This study aimed to map the fine‐grained pattern of four heathland habitats throughout France (550 000 km2). Environmental (bioclimatic, soil and topographic) and spectral (vegetation) variables derived from MODerate resolution Imaging Spectroradiometer, Advanced Spaceborne Thermal Emission and Reflection Radiometer, and Sentinel‐2 satellite data were analyzed using the MaxEnt classifier. Open‐access field databases were used to calibrate and validate the classification, based on the threshold‐independent area under the curve (AUC) index and the conventional F1‐score. For each heathland habitat, potential and actual areas were mapped using environmental and spectral variables, respectively. The results showed high classification accuracy for potential (AUC 0.92–0.99) and actual (AUC 0.88–0.99) suitability maps of the four heathland habitats. Visual interpretation of maps of the probability of occurrence indicated that the fine‐grained distribution of heathland habitat was detected satisfactorily. However, although the accuracy of the crisp map of combined classifications of actual heathland habitats was high (overall accuracy 0.72), estimated producer's accuracies in terms of proportion of area were low (<0.25). This study provides the first fine‐grained pattern maps of heathland habitats at a national scale, thus highlighting the value of combining environmental and spectral variables derived from open‐remote sensing data and open‐source field databases. These suitability maps could support the identification of heathland habitats in the framework of national conservation policies.

Mapping causal agents of disturbance in boreal and arctic ecosystems of North America using time series of Landsat data

The arctic and boreal biomes are changing as temperatures increase, including changes in the type, frequency, intensity, and seasonality of disturbances. However, our understanding of the frequency, extent, and causes of disturbance events remains incomplete. Disturbances such as fire, forest harvest, drought, wind, flooding, and insects and pathogens occur at different frequencies and severities, posing challenges to characterize and assess them under a single framework. We used the Continuous Change Detection and Classification (CCDC) algorithm on all available Landsat observations from 1984 to 2014 to detect land cover and land condition change. We mapped the following causes of disturbances annually across the study domain of NASA's Arctic Boreal Vulnerability Experiment (ABoVE): fire, logging, and pest damage. Differences between Landsat Tasseled Cap (TC) values pre- and post-disturbance were used in a random forest classifier to map causal agents. For forested ecosystems, we mapped causal agents including fire, insect, and logging. In areas that were not forest before disturbance, only the fire class was mapped. The result shows that multidimensional spectral-temporal change information is useful for mapping the causes of disturbance in arctic and boreal biomes. We employed two rounds of post-processing and used the information obtained from the comparison between the map and reference data to improve the final map. The user's and producer's accuracies of an aggregated disturbance map were 94.6% (± 2.37%) and 89.3% (± 21.78%) (95% confidence intervals in parenthesis). When evaluating the causal agents, insect damage was found the most challenging to map and validate. We estimated that 10.8% of the ABoVE core domain was disturbed between 1987 and 2012, with a margin of error of 0.5% at the 95% confidence level. Rates of disturbance due to logging remained constant over time, while fires were more episodic, and insect damage was highest between 2005 and 2010. Overall, fires affected 8.8% of the study area, while logging was 1.4% and insect damage 0.6%. Our maps indicate that pest damage became a significant issue after 2000, but it was more severe for forest ecosystems in Western Canada than in Alaska.

Application of Remote Sensing Tools to Assess the Land Use and Land Cover Change in Coatzacoalcos, Veracruz, Mexico

Land use and land cover (LULC) change has become an important research topic for global environmental change and sustainable development. As an important part of worldwide land conservation, sustainable development and management of water resources, developing countries must ensure the use of innovative technology and tools that support their various decision making systems. This study provides the most recent LULC change analysis for the last six years (2015–2021) of Coatzacoalcos, Veracruz, Mexico, one of the most important petrochemical cities in the world and host of the ongoing Interoceanic Corridor project. The analysis was carried out using Landsat 8 Operational Land Imager (OLI) satellite images, ancillary data and ground-based surveys and the Normalized Difference Vegetation Index (NDVI) to identify and to ameliorate the discrimination between four main macro-classes and fourteen classes. The LULC classification was performed using the maximum likelihood classifier (MLC) to produce maps for each year, as it was found to be the best approach when compared to minimum distance (MDM) and spectral angle mapping (SAM) methods. The macro-classes were water, built-up, vegetation and bare soil, whereas the classes were an improved classification within those. Our study achieved both user accuracy (UA) and producer accuracy (PA) above 90% for the proposed macro-classes and classes. The average Kappa coefficient for macro-classes was 0.93, while for classes it was 0.96, both comparable to previous studies. The results from the LULC analysis show that residential, industry and commercial areas slowed down their growth throughout the study period. These changes were associated with socio-economical drivers such as insecurity and lack of economic investments. Groves and trees presented steady behaviors, with small increments during the five-year period. Swamps, on the other hand, significantly degraded, being about 2% of the study area in 2015 and 0.93% in 2021. Dunes and medium and high vegetation densities (∼80%) transitioned mostly to low vegetation densities. This behavior is associated with rainfall below the annual reference and increments of surface runoff due to the loss of vegetation cover. Lastly, the present study seeks to highlight the importance of remote sensing for a better understanding of the dynamics between human–nature interactions and to provide information to assist planners and decision-makers for more sustainable land development.

Cross-system legacy data applied to digital soil mapping: A case study of Second National Soil Survey data in China

The accuracy of digital soil maps (DSMs) is influenced by the quality of the input data. The application of legacy data obtained from past soil surveys was considered a potential method to improve DSM input data. However, changes in soil classification systems have occurred in many countries, and legacy data are therefore not directly applicable in the generation of the latest DSMs. This research established a cross-system framework for digital soil mapping, which enabled the conversion of legacy data and application to a new classification system for digital soil mapping. Jiangxi Province, China, was considered to evaluate this framework, where the classification system transitioned from the Genetic Soil Classification of China (GSCC) to the China Soil Taxonomy (CST). Second National Soil Survey legacy point data acquired based on the GSCC were converted and updated. Moreover, boundary placement of legacy polygon maps (BP-LPMs) containing GSCC soil class maps and soil parent material maps were converted into the expert knowledge of CST soil class boundaries. Fuzzy logic inference was employed to construct DSMs. The actual sampling points were extracted and considered in simple digital soil map (S-DSM) construction. The actual sampling points, the converted legacy point, and expert knowledge of soil class boundaries were used for hybrid digital soil map (H-DSM) construction. Cross-validation was used to evaluate the accuracy of digital soil mapping. The evaluation results indicated that compared to the S-DSM, the total accuracy of the H-DSM were improved by 17.0%, 15.1%, and 9.4% at the soil Order, Suborder, and Group class levels, respectively, and the producer's accuracy was further improved for most soil classes. This study demonstrated that legacy data obtained from past soil classification systems could be applied to improve digital soil maps under new soil classification systems through updating and conversion. This cross-system framework provides important implications for many countries with similar legacy data contexts.

Assessment of the Usability of SAR and Optical Satellite Data for Monitoring Spatio-Temporal Changes in Surface Water: Bodrog River Case Study

Mapping watercourses and their surroundings through remote sensing methods is a fast, continuous, and effective method and is a crucial tool for capturing change and possibly predicting hazards. Thanks to Synthetic Aperture Radar (SAR) technology and the ability of its backscattered and emitted radiation to penetrate the atmosphere under any conditions, this type of mapping of water surfaces is of particular importance. This paper presents the possibility of using SAR technology for long-term observations of changes in the behaviour of rivers and river systems, combined with optical multispectral images Sentinel-2. Additionally, it aims to demonstrate the suitability of satellite SAR and multispectral data implementation for mapping changes in watercourses, caused not only by their natural development but especially by inundation processes in their catchment area. Appropriate Sentinel-1 image processing evaluation procedures demonstrate that the usage of vertical-vertical (VV) type polarisation configuration is a suitable methodology for documenting water bodies, and a Lee filter is an acceptable tool for radar noise filtering. The extraction process of water surfaces is based on the determination of threshold values using the “Otsu” principle. Subsequently, the comparison of the results is realised by the spectral indices of water—the Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), a pair of Automated Water Extraction Index (AWEI) indices, and supervised classification method Maximum Likelihood Classification (MLC). The results are numerical and graphical evaluated. In assessing the accuracy of SAR extraction, the highest values achieved in Overall Accuracy (OA) were a maximum of 98.6%. On average, the lower values were in User Accuracy (UA) with a maximum of 93.1%, where VV polarisation also dominates. However, vertical-horizontal (VH) polarisation dominates in Producer Accuracy (PA) with a maximum of 84.9%.

Paddy lands detection using Landsat-8 satellite images and object-based classification in Rasht city, Iran

Rice is one of the most important food staples in many countries, particularly Iran. Because irrigated rice production differs from other agricultural fields, this study developed a paddy field mapping model based on phenological aspects, various satellite sensor data, and an object-based approach. This study uses the phonological features of rice plants as well as annual data on land surface temperature (LST) to produce the paddy map. The core remote sensing data consists of the yearly LST from MODIS and multi-temporal Landsat-8 satellite imagery. The detection of phenological characteristics and the selection of relevant Landsat images were made possible by analyzing the LST time series with Google Earth Engine. After that, object-based image classification and fuzzy functions are used to create flexible and comprehensible rules for discovering paddy fields in Rasht, Iran. Data such as the digital elevation model (DEM) and spectral indices including NDVI, EVI, and LSWI were employed to improve the object-based classification. Due to the unique properties of paddy lands, a DEM of 12.5 m obtained from the ALOS PALSAR sensor could help distinguish paddy lands from other vegetation. A comparison finally made between the object-based and pixel-based classification methods showed that the former is more accurate. Overall accuracy and kappa coefficient for the pixel-based classification approach were 92% and 0.89, respectively, whereas overall accuracy and kappa coefficient for the object-based classification method were 94% and 0.92, respectively. Eventually, the producer's accuracy of the paddy class has increased from 88% to 94%.

Classification of Laser Footprint Based on Random Forest in Mountainous Area Using GLAS Full-Waveform Features

Full-waveform spaceborne laser altimeter can provide more characteristic parameters of the laser footprint and rich vertical structure information on the target surface. This technology has the potential for land-cover classification, especially in hard-to-reach mountain areas. Classifying the land types based on the returned waveform can provide a convenient way for the online classification needs and assess the quality of footprint used as the ground control point in photogrammetry. This article presents a random forest (RF) classification using geoscience laser altimeter system waveform, in the west-central Yunnan Province, China. First, an improved threshold wavelet is performed to denoise the waveform, and then Gaussian decomposition is used to extract the typical characteristic features of footprint. Second, an RF algorithm is implemented to clarify the footprints into five types: flat, building, terrace, forest, and mountain. Finally, quantitative analysis is conducted with producer's accuracy (PA), user's accuracy (UA), overall accuracy (OA), precision, recall rate, F1-score, and kappa coefficient to compare the performance of RF with other classifiers, including linear support vector machine (linear-SVM), radial basis function SVM (RBF-SVM), logistic regression (LR), K-nearest neighbor (KNN), and naive Bayes (NB). The results show that all the six methods can accurately classify the flat land with 100.00% PA and UA. The RF also has the best performances in other four land types, with PA and UA of 98.14% and 100.00%, 97.24% and 95.49%, 98.64% and 96.03%, and 94.64% and 100.00%, respectively. The OA, precision, recall, F1-score, and kappa coefficient for the RF are 97.95%, 97.73%, 98.30%, 97.99%, and 0.9737, respectively; while 83.45%, 82.55%, 82.98%, 81.16%, and 0.7765 for NB, which has the worst performance. LR performs better than RBF-SVM, linear-SVM, and KNN. It also observes worse classification accuracy for all methods when the waveforms are more complex.

Quantitative Evaluation of Urban Expansion using NPP-VIIRS Nighttime Light and Landsat Spectral Data

Quantifying the evolution of urban expansion is of vital importance for optimizing land use patterns and promoting sustainable urban development. This research constructed the Enhanced Nighttime Light Urban Index (ENUI), which combines NPP-VIIRS nighttime light with Landsat spectral data, and used this index for the quantitative evaluation of urban expansion in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) urban agglomeration. The results showed that, between 2012 and 2018, (1) the extraction using the ENUI achieved an overall accuracy of 0.93–0.94, producer's accuracy in urban areas of 0.82–0.84, user's accuracy in urban areas of 0.73–0.75, and Kappa values of 0.75–0.78, showing a relatively high extraction accuracy; (2) the GBA urban expansion followed a circular and radial expansion pattern; (3) the growth trend of expansion intensity and development potential for sub-central cities was stronger than central cities; and (4) the agglomeration autocorrelation of the spatial expansion in the GBA presents a continuous "dispersion" trend. Compared with previous methods, our study improved the extraction accuracy of urban expansion and effectively reduced the deficiency of overflow effect for NPP-VIIRS nighttime light. The research can thus provide valuable datasets and reference for decision-making to adjust and optimize urban development pattern and design urban sustainable development. • NPP-VIIRS nighttime light, NDVI, NDWI and NDBI are used to quantify urban expansion. • The resulting overall accuracy and Kappa are 0.93–0.94 and 0.75–0.78, respectively. • The resulting producer's accuracy and user's accuracy are 0.82–0.84 and 0.73–0.75, respectively. • Multiple indicators are used to quantify urban expansion and analyze agglomeration. • Urban expansion rate showed different development patterns for different cities.

New Generation Hyperspectral Data From DESIS Compared to High Spatial Resolution PlanetScope Data for Crop Type Classification

Thoroughly investigating the characteristics of new generation hyperspectral and high spatial resolution spaceborne sensors will advance the study of agricultural crops. Therefore, we compared the performances of hyperspectral Deutsches Zentrum fur Luft - und Raumfahrt- (DLR) Earth Sensing Imaging Spectrometer (DESIS) and high spatial resolution PlanetScope in classifying eight crop types in California's Central Valley during the 2020 growing season. The DESIS sensor onboard the International Space Station collects data at 235 hyperspectral narrowbands (HNB) each with 2.55 nm bandwidth from 400-1000 nm and 30 m spatial resolution. In contrast, PlanetScope Dove-R data have 4 multispectral broadbands (MBB) with 3-4 m spatial resolution. We obtained best classification accuracies using 14 DESIS HNB from the August 2020 image, with an overall accuracy of 85% and producer's and user's accuracies of 72-100% and 75-100% respectively for the eight crops. The best classification accuracies using PlanetScope data were obtained using an image mosaic pair from June and August 2020; this resulted in an overall accuracy of 79% and producer's and user's accuracies of 56-100% and 61-100% respectively. Combining the best 14 DESIS HNB from August 2020 with the 4 PlanetScope MBB from August 2020 yielded an overall accuracy of 82% and producer's and user's accuracies of 65-100% and 60-94% respectively. On one-to-one single date comparisons of DESIS versus PlanetScope data, the hyperspectral data always outperformed high spatial resolution data in crop type classification. Nevertheless, high spatial resolution data will remain invaluable in assessing within-field variability and crop biophysical/biochemical modeling in precision agriculture.

Knowledge-Based Morphological Deep Transparent Neural Networks for Remote Sensing Image Classification

Land use/land cover classification of remote sensing images provide information to take efficient decisions related to resource monitoring. There exists several algorithms for remote sensing image classification. In the recent years, Deep learning models like convolution neural networks (CNNs) are widely used for remote sensing image classification. The learning and generalization ability of CNN, results in better performance in comparison with similar type of models. The functional behavior of CNNs is unexplainable because of its multiple layers of convolution and pooling operations. This results in black box characteristics of CNNs. Motivated with this factor, a CNN model with functional transparency is proposed in the present study. The model is named as Knowledge Based Morphological Deep Transparent Neural Networks (KBMDTNN) for remote sensing image classification. The architecture of KBMDTNN model provides functional transparency due to application of morphological operators, convolutional and pooling layers, and transparent neural network. In KBMDTNN model, the morphological operator preserve the shape/size information of the objects through efficient image segmentation. Convolution and pooling layers are used to produce minimal number of features from the image. The operational transparency of proposed model is coined based on the mathematical understanding of each layer in the model instead of randomly adding layers to the architecture of model. The transparency of proposed model is also because of assigning the initial weights of NN in output layer of model with computed values instead of random values. The proposed KBMDTNN model outperformed similar type of models as tested with multispectral and hyperspectral remote sensing images. The performance of KBMDTNN model is evaluated with the metrics like overall accuracy (OA), overall accuracy standard deviation ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$OA_{\text{STD}}$</tex-math></inline-formula> ), producer’s accuracy (PA), user’s accuracy (UA), dispersion score (DS), and kappa coefficient (KC).

Determination of Forest Burn Scar and Burn Severity from Free Satellite Images: a Comparative Evaluation of Spectral Indices and Machine Learning Classifiers

Remote sensing data indicates a considerable ability to map post-forest fire destructed areas and burned severity. In this research, the ability of spectral indices, which are difference Normalized Burned Ratio (dNBR), relative differenced Normalized Burn Ratio (RdNBR), Relativized Burn Ratio (RBR), and difference Normalized Vegetation Index (dNDVI), in mapping burn severity was investigated. The research was conducted with free access moderate to high-resolution Landsat 8 and Sentinel 2 satellite images for two forest fires cases that occurred in Izmir and Antalya provinces of Turkey. Performance of the burn severity maps from different indices were validated by use of NASA Firms active fires dataset. The results confirmed that, RdNBR showed more precise results than the other indices with an accuracy of (89%, 93%) and (84%, 79%) for Landsat 8 and Sentinel 2 satellites over Izmir and Antalya respectively. Moreover, in this research, the ability of machine learning classifiers, which are Support Vector Machine (SVM) and Random Forest (RF), in mapping burned areas were evaluated. According to the accuracy metrics that are user’s accuracy, producer's accuracy and Kappa coefficient, we concluded that both classifiers indicate reliable and accurate detection for both regions.