Fig 6 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
Planetscope data are acquired from a constellation of low-cost satellites to provide 3 m red, green, blue, and near infrared (NIR) data with near-daily global coverage. Differences in the spectral characteristics of the different Planetscope sensor generations imply that they may not provide consistent reflectance time series needed for certain qua...
Citations
... For the U-Net Wieland algorithm, "clear" was defined as the snow/ice, water, and land classes. The TSI λ was used previously to evaluate the consistency of MODIS [102], Landsat and Sentinel-2 [103], and PlanetScope [104], reflectance time series. The TSI λ is zero valued for time series sensed without noise and over an unchanging surface and will be greater if any clouds or cloud shadows are present that failed to be detected correctly. ...
... The TSI λ is zero valued for time series sensed without noise and over an unchanging surface and will be greater if any clouds or cloud shadows are present that failed to be detected correctly. The TSI was derived considering only sequences of successive pixel observations satisfying (day i+2 -day i ) ≤ 32 to reduce the impact of land surface changes that will inflate the TSI values [104]. ...
Landsat cloud and cloud shadow detection has a long heritage based on the application of empirical spectral tests to single image pixels, including the Landsat product Fmask algorithm, which uses spectral tests applied to optical and thermal bands to detect clouds and uses the sun-sensor-cloud geometry to detect shadows. Since the Fmask was developed, convolutional neural network (CNN) algorithms, and in particular U-Net algorithms (a type of CNN with a U-shaped network structure), have been developed and are applied to pixels in square patches to take advantage of both spatial and spectral information. The purpose of this study was to develop and assess a new U-Net algorithm that classifies Landsat 8/9 Operational Land Imager (OLI) pixels with higher accuracy than the Fmask algorithm. The algorithm, termed the Learning Attention Network Algorithm (LANA), is a form of U-Net but with an additional attention mechanism (a type of network structure) that, unlike conventional U-Net, uses more spatial pixel information across each image patch. The LANA was trained using 16,861 512 × 512 30 m pixel annotated Landsat 8 OLI patches extracted from 27 images and 69 image subsets that are publicly available and have been used by others for cloud mask algorithm development and assessment. The annotated data were manually refined to improve the annotation and were supplemented with another four annotated images selected to include clear, completely cloudy, and developed land images. The LANA classifies image pixels as either clear, thin cloud, cloud, or cloud shadow. To evaluate the classification accuracy, five annotated Landsat 8 OLI images (composed of >205 million 30 m pixels) were classified, and the results compared with the Fmask and a publicly available U-Net model (U-Net Wieland). The LANA had a 78% overall classification accuracy considering cloud, thin cloud, cloud shadow, and clear classes. As the LANA, Fmask, and U-Net Wieland algorithms have different class legends, their classification results were harmonized to the same three common classes: cloud, cloud shadow, and clear. Considering these three classes, the LANA had the highest (89%) overall accuracy, followed by Fmask (86%), and then U-Net Wieland (85%). The LANA had the highest F1-scores for cloud (0.92), cloud shadow (0.57), and clear (0.89), and the other two algorithms had lower F1-scores, particularly for cloud (Fmask 0.90, U-Net Wieland 0.88) and cloud shadow (Fmask 0.45, U-Net Wieland 0.52). In addition, a time-series evaluation was undertaken to examine the prevalence of undetected clouds and cloud shadows (i.e., omission errors). The band-specific temporal smoothness index (TSIλ) was applied to a year of Landsat 8 OLI surface reflectance observations after discarding pixel observations labelled as cloud or cloud shadow. This was undertaken independently at each gridded pixel location in four 5000 × 5000 30 m pixel Landsat analysis-ready data (ARD) tiles. The TSIλ results broadly reflected the classification accuracy results and indicated that the LANA had the smallest cloud and cloud shadow omission errors, whereas the Fmask had the greatest cloud omission error and the second greatest cloud shadow omission error. Detailed visual examination, true color image examples and classification results are included and confirm these findings. The TSIλ results also highlight the need for algorithm developers to undertake product quality assessment in addition to accuracy assessment. The LANA model, training and evaluation data, and application codes are publicly available for other researchers.
... The spatial resolution of these images' ranges from 3.7 to 4.1 meters, contingent upon the altitude of the satellite's orbit. The acquired data is subsequently resampled to a spatial resolution of 3 meters and distributed in the form of 16-bit GeoTIFF images 36 . ...
... These specific metrics performed significantly better than did those obtained with Sentinel-2, limiting the propagation of errors when multiple metrics were combined. Nonetheless, despite the strict selection of scenes, our simulation results showed that the variability in PlanetScope viewing angles could be a significant error source in phenological estimates [113]. The models demonstrated varying effects depending on the metric, season, and selection criteria such as species (through site). ...
While satellite time series are essential tools to derive phenometrics at unprecedented spatial and temporal scales, non-systematic acquisition or medium spatial resolution of available missions are potentially problematic. At the same time, low-cost observation networks bridge the gap between satellite and in-situ observations which considerably increases ground-based data and associated possibilities. Here we provide robust statistics about the reliability of satellite derived phenometrics of urban trees across phenophases. Environmental and acquisition factors influencing the quality of phenometrics estimates were analyzed. First, a multi-facet regression-based analysis was conducted to measure discrepancies between PlanetScope (and Sentinel-2) and ground-based measurements across phenophases. Then, we performed hierarchical partitioning to tackle the effects of biological parameters (canopy closure and colour leaf) for assessing phenometrics with satellite time series. Third, we ran Monte-Carlo simulations to propagate errors according to viewing angles in PlanetScope acquisition. Our results show that: (i) PlanetScope provide consistent phenometric estimates for different tree layouts belonging to the same species (average
$R^{2}$
= 0.50
$\pm$
0.18). Performances are higher than Sentinel-2 but duration-based phenometrics estimates were poorly reconstructed with both satellite missions, (ii) contributions of biological parameters in the vegetation signal above trees strongly vary between growth periods. While canopy closure drives the growing season signal (independent contribution
$>$
40%), colour leaf plays a major role in the senescence season, (iii) variable viewing angles in PlanetScope acquisitions showed only significant effects on duration-based metrics estimates. Our research opens new perspectives for monitoring urban trees which improves the measurement of ecosystem services for local inhabitants.
... Fifthly, PlanetScope images are acquired using many small satellites, belonging to difference generations, rather than a single large platform like Sentinel-2 and Landsat. Therefore, radiometric and geometric corrections of these products are not always compatible across generations of satellites, and does not always match the standard requirements expected by the remote sensing community [29,30]. Last but not least, similar to other optical satellite platforms, cloud cover can be a significant problem with PlanetScope satellites, especially over tropical regions where cloud contamination is normally very high during several months in a year [31]. ...
This study employed high spatial resolution PlanetScope imagery at 3 m resolution for mapping burned area and burn severity resulting from a wildfire that occurred from April 7-9, 2023 in the highland of Vietnam. The wildfire took place in a protection forest near the Prenn pass in Da Lat city, Lam Dong province, Vietnam. Pre-and post-fire Normalized Difference Vegetation Index (NDVI) maps were generated using no-cloud high-resolution images acquired on March 25 and April 23, 2023 by the PlanetScope's SuperDove satellites, respectively. The difference of NDVI (dNDVI) was then calculated, and thresholds, proposed by the author, were utilized to classify the study area into three different classes: unburned, low-to-moderate and high severity. The results showed that the total burned area was approximately 13.86 ha, with 8.19 ha classified as low-to-moderate severity, and 5.68 ha classified as high severity. Although there was no reference dataset to cross-validate the results, the estimated burned area is very close to the total affected area officially reported by the Forest Protection Department of Lam Dong province (about 13 ha). This study is one of the few that investigates the use of high-resolution PlanetScope imagery for environmental monitoring in Vietnam, and the first to focus on burned area and burn severity mapping in Vietnam. This work demonstrates the potential of PlanetScope images for mapping burned area and burn severity, particularly in small regions where other optical satellites, such as Sentinel-2 and Landsat, may not provide accurate results due to their spatial resolution limitations.
... Its improved spatial and temporal resolution over many commonly used RS imageries (e.g., Landsat, Sentinel-2, SAR, and MODIS) have boosted its popularity in multiple aspects, including surface water extent mapping, over the recent years. However, lack of on-board calibration (Huang & Roy, 2021) as well as the difference between sensors of different generations made images from PlanetScope less consistent compared to image series from other platforms such as Landsat and Sentinel-2 (Frazier & Hemingway, 2021;Huang & Roy, 2021). ...
... Its improved spatial and temporal resolution over many commonly used RS imageries (e.g., Landsat, Sentinel-2, SAR, and MODIS) have boosted its popularity in multiple aspects, including surface water extent mapping, over the recent years. However, lack of on-board calibration (Huang & Roy, 2021) as well as the difference between sensors of different generations made images from PlanetScope less consistent compared to image series from other platforms such as Landsat and Sentinel-2 (Frazier & Hemingway, 2021;Huang & Roy, 2021). ...
Remote sensing (RS) imagery has become more and more popular in surface water extent extraction applications with the help of increasing availability of RS data and advancements in image processing algorithms, software, and hardware. Many studies have demonstrated that RS imagery has the potential to work independently or along with other well-documented approaches in identifying flood extent. However, due to the insufficiency of images from single-sourced RS and independent references for validation, most existing studies either focused on mapping a single scene or failed to support their results with adequate non-RS validation when multi-temporal/multi-spatiotemporal images were involved. Because of that, hydrosimulations still dominate flood series mapping despite requiring huge data and computational resources. To close these gaps, this study investigated the efficacy of RS-based multi-spatiotemporal flood inundation mapping using multimodal RS imageries to take advantage of improved data availability and complementary image properties. This study also proposed a Quantile-based Filling & Refining (QFR) workflow to resolve the blocking effects of dense vegetation that occurs in study areas. We tested the workflow in four lock and dam sites on the Mississippi River, downstream to the Quad City area, by comparing the RS-based flood maps with HEC-RAS simulations. Compared to the original flood extent that only went through basic post-processing, QFR maps were noticeably more consistent with HEC-RAS maps. Results also showed that all steps in QFR contributed to performance improvements. Despite all being necessary in our case, some should be adjusted in different study regions, such as the levee step. Our findings showcased the efficacy of the multimodal RS flood mapping with QFR post-processing. Due to its simple structure, the proposed workflow has potential to be fully automated and can benefit near-real-time and real-time applications.
... When a horizontal plane is intersected with a surface, plan curvature and profile curvature are defined as the curvature of the contour line produced (Fig. 3iii). NDVI is defined as (Near Infrared -RED)/(Near Infrared þ RED) (Zhang and Roy, 2016;Zhang et al., 2018;Chen et al., 2019a;Huang and Roy, 2021;Tao et al., 2021). The convergence or divergence of water during downhill flow is the mechanism through which plan curvature influences slope erosion processes (Oh and Pradhan, 2011). ...
... The NDVI (Huang and Roy, 2021;Tao et al., 2021) is one of the most important vegetation indices and is applied to a wide range for calculating the amount of vegetation cover with acceptable precision. The amount of red to a near-infrared thermal band reflected by plants throughout the day is usually approximate. ...
The unique characteristics of drainage conditions in the Pagla river basin cause flooding and harm the socioeconomic environment. The main purpose of this study is to investigate the comparative utility of six machine learning algorithms to improve flood susceptibility and ensemble techniques' capability to elucidate the underlying patterns of floods and make a more accurate prediction of flood susceptibilities in the Pagla river basin. In the present scenario, the frequency of flood conditions in this study area becomes high with heavy and sudden rainfall, so it is essential to study flood mitigation and measure. At first, a spatial flood database was built with 200 flood locations and sixteen flood influencing factors, and its process with the help of the Geographic Information System (GIS) environment and build up different models applying the machine learning techniques. It has found different flood susceptibility zone using machine learning-based Artificial Neural Network (ANN), Support Vector Machine (SVM), Random Forest (RF), Reduced Error Pruning Tree (REPTree), Logistic Regression (LR), and Bagging helping GIS environment and the model validation using the Receiver Operating Characteristic Curve (ROC). Afterward, ensemble all the models to gate comparative accuracy of the flood zone. The calculated areas are under the very high flood susceptibility zone 8.69%, 14.92%, 14.17%, 12.98%, 14.65%, 13.24% and 13.41% for ANN, SVM, RF, REPTree, LR and Bagging, respectively. Finally, ROC curve, the Standard Error (SE), and the Confidence Interval (CI) at 95 per cent were used to assess and compare the performance of the models. The obtained results indicate that all models are highly accepted Area Under Curve (AUC) of ROC between 0.889 (LR) to 0.926 (Ensemble). From the estimation of the accuracy of the applied methods using ROC, it is found that the Ensemble model has the higher capability compared to the other applied models in projecting flood susceptibility in the study area. It has the highest area under the ROC curve the AUC values are 0.918 and 0.926, the SE (0.023, 034), and the narrowest CI (95 per cent) (0.873–0.962, 0.859–0.993) whereas highest area under Bagging (the ROC) curve (AUC) value (0.914, 0.919), for both the training and validation datasets. After ensembling, the result shows that the result is a highly flood susceptible area located at the lower part of the study area. In this area, the very high flood susceptibility zone values lie between 4.46 and 6.00 in the ensemble result. The areas comprise the low height and belong to Murarai I, Murarai II, Suti I and Suti II C.D. block of West Bengal. The current study will help the policymakers and the researcher determine the flood conditioning problems for prospects.
... These results indicate the high accuracy of PlanetScope images in predicting soil moisture content in rainfed rice fields in the toposequence costs [3]. A more modern way of using PlanetScope imagery will cost far less [4]. According to [5], PlanetScope imagery yielded a significant correlation with soil moisture in [6] showed his Research on mapping striga weed in corn fields that the accuracy of PlanetScope imagery is higher (92%) compared to sentinel-2 (88%). ...
Information on soil moisture is very important in activities cultivation of crops, especially in rainfed land which is very vulnerable to climate change. The need for a method for estimating soil water content is very urgent because of the amount of time and energy required by the gravimetric method. One way that can be applied is with Planetscope Imagery. Planetscope sensors are inexpensive so they are increasingly being used for science and environmental applications, including for land cover classification. The research was carried out on rainfed land, the slopes of Mount Lawu - Central Java, at different altitudes, in the highlands, the middle plains, and the lowlands, with descriptive explorative research method, using purposive sampling to obtain an image to determine the Color Digital Number value. The results of the color digital number are used for correlation and regression analysis to determine the relationship between the actual soil moisture content and the digital number on the Planetscope image. The T-test was used to determine whether there was a significant difference between the results of the actual moisture content and the predicted moisture content. The results showed that the accuracy of Planetscope imagery in predicting soil moisture content in the highlands, middle plains, and lowlands was 86.16%; 89.07%, and 95%. These results indicate the high accuracy of PlanetScope images in predicting soil moisture content in rainfed rice fields in the toposequence of Mount Lawu, Central Java.
... Many studies use free satellite imageries for land-use monitoring and change detection (Al-Juboury & Al-Rubaye, 2021; Chen & Wang, 2010;Chughtai et al., 2021;Fonji & Taff, 2014), crop identification and mapping (Belgiu & Csillik, 2018;Xun et al., 2021;Yan et al., 2021), phenology mapping using time series Schreier et al., 2021;Zhao et al., 2021) and other applications. In addition to access to free satellite imagery, many private satellite companies, like Planet Lab, provide high spatial and temporal resolution time series imagery for a fee (Huang & Roy, 2021). The type and use of satellite images mainly depend on the research objectives. ...
... NDVI change layers were created and defined as the differences of the respective year to pre-event conditions by ∆NDV I Yn−Y −1 = NDV I Yn − NDV I Y −1 . As radiometric inconsistencies between Plan-etScope sensors reduced the quality of time series data [71], this study restricts its use to 1-sigma quantile classification in bi-temporal (pre-and post-droughts, Y −1 and Y 2 ) differences and visualizes damage patterns spatially. The a-posteriori Canopy Development (CD) tree sample sets were defined by: ...
Understanding forest decline under drought pressure is receiving research attention due to the increasing frequency of large-scale heat waves and massive tree mortality events. However, since assessing mortality on the ground is challenging and costly, this study explores the capability of satellite-borne Copernicus Sentinel-1 (S-1) C-band radar data for monitoring drought-induced tree canopy damage. As droughts cause water deficits in trees and eventually lead to early foliage loss, the S-1 radiometric signal and polarimetric indices are tested regarding their sensitivities to these effects, exemplified in a deciduous broadleaf forest. Due to the scattered nature of mortality in the study site, we employed a temporal-only time series filtering scheme that provides very high spatial resolution (10 m ×10 m) for measuring at the scale of single trees. Finally, the anomaly between heavily damaged and non-damaged tree canopy samples (n = 146 per class) was used to quantify the level of damage. With a maximum anomaly of −0.50 dB ± 1.38 for S-1 Span (VV+VH), a significant decline in hydrostructural scattering (moisture and geometry of scatterers as seen by SAR) was found in the second year after drought onset. By contrast, S-1 polarimetric indices (cross-ratio, RVI, Hα) showed limited capability in detecting drought effects. From our time series evaluation, we infer that damaged canopies exhibit both lower leaf-on and leaf-off backscatters compared to unaffected canopies. We further introduce an NDVI/Span hysteresis showing a lagged signal anomaly of Span behind NDVI (by ca. one year). This time-lagged correlation implies that SAR is able to add complementary information to optical remote sensing data for detecting drought damage due to its sensitivity to physiological and hydraulic tree canopy damage. Our study lays out the promising potential of SAR remote sensing information for drought impact assessment in deciduous broadleaf forests.
... For instance, the Water Detect algorithm was initially proposed to be applied to Sentinel-2 images with SWIR bands. Planetscope sensors also have no onboard calibration devices, making it hard to systematically implement corrections to the multiple generations of this satellite constellation, resulting in spectrally-variable output data (Huang and Roy 2021). While initially problematic, this difference in spectral quality and number of bands among Planetscope images may be overcome by spectral data sharpening and band synthesizing. ...
Mapping surface water using remotely sensed optical imagery is a particular challenge in intermittent rivers because water contracts down to narrow linear features and isolated pools, which require accurate water detection methods and reliable image datasets. Of the many methods that use optical sensors to identify water, the Water Detect algorithm stands out as one of the best options due to its classification accuracy, open-source code, and because it does not require ancillary data. However, in the original study, the Water Detect algorithm was only tested with Sentinel-2 imagery. High-resolution and high-frequency imagery, such as Planetscope, combined with sharpening and band synthesizing techniques have the potential to improve the accuracy of surface water mapping, but their benefit to the Water Detect algorithm remains unknown. Uncertainty also exists about the extent to which different input parameters (i.e. maximum clustering and regularization) influence the accuracy of Water Detect. Practitioners seeking to map surface water in intermittent rivers need guidance on a best-practice approach to improve the accuracy of Water Detect. To meet this need, we automated an existing method for sharpening and synthesizing bands and applied it to a series of multispectral Sentinel-2 and Planetscope images. We then developed a sensitivity analysis algorithm that compared the accuracy for all possible combinations of input parameters in a given range for the water detection process – enabling optimal parameters to be identified. We applied this workflow to an 81 km stretch of the lower Fitzroy River (Western Australia) to periods when spatial water extent varied markedly, i.e. mid-wet (February), early-dry (June), and late-dry season (October), across three years with variable wet season flow. We found that the ability to accurately detect surface water using multispectral imagery was increased by using input parameters identified by the sensitivity analysis and using Visible + Near-infrared (VNIR) bands, with relatively little gained by image sharpening unless the area of interest was burnt or experienced considerable shading. Also, the regularization parameter exerted less influence on results than maximum clustering. Importantly, the accuracy of the Water Detect algorithm can vary drastically if input parameters are not calibrated to local conditions. Results also revealed that our approach was adept at detecting linear features in intermittent rivers. We recommend that practitioners using Water Detect to identify surface water undertake a workflow similar to that described here to improve the accuracy of the Water Detect algorithm. The automated routines provided by this study will significantly assist practitioners in doing so. Increasing the accuracy with which we detect and map water in intermittent rivers will improve our understanding and management of these important systems which are under increasing threat.
