Dense Time Series Research Articles

Evaluation of Five Deep Learning Models for Crop Type Mapping Using Sentinel-2 Time Series Images with Missing Information

Accurate crop type maps play an important role in food security due to their widespread applicability. Optical time series data (TSD) have proven to be significant for crop type mapping. However, filling in missing information due to clouds in optical imagery is always needed, which will increase the workload and the risk of error transmission, especially for imagery with high spatial resolution. The development of optical imagery with high temporal and spatial resolution and the emergence of deep learning algorithms provide solutions to this problem. Although the one-dimensional convolutional neural network (1D CNN), long short-term memory (LSTM), and gate recurrent unit (GRU) models have been used to classify crop types in previous studies, their ability to identify crop types using optical TSD with missing information needs to be further explored due to their different mechanisms for handling invalid values in TSD. In this research, we designed two groups of experiments to explore the performances and characteristics of the 1D CNN, LSTM, GRU, LSTM-CNN, and GRU-CNN models for crop type mapping using unfilled Sentinel-2 (Sentinel-2) TSD and to discover the differences between unfilled and filled Sentinel-2 TSD based on the same algorithm. A case study was conducted in Hengshui City, China, of which 70.3% is farmland. The results showed that the 1D CNN, LSTM-CNN, and GRU-CNN models achieved acceptable classification accuracies (above 85%) using unfilled TSD, even though the total missing rate of the sample values was 43.5%; these accuracies were higher and more stable than those obtained using filled TSD. Furthermore, the models recalled more samples on crop types with small parcels when using unfilled TSD. Although LSTM and GRU models did not attain accuracies as high as the other three models using unfilled TSD, their results were almost close to those with filled TSD. This research showed that crop types could be identified by deep learning features in Sentinel-2 dense time series images with missing information due to clouds or cloud shadows randomly, which avoided spending a lot of time on missing information reconstruction.

Fine-Resolution Mapping of Pan-Arctic Lake Ice-Off Phenology Based on Dense Sentinel-2 Time Series Data

The timing of lake ice-off regulates biotic and abiotic processes in Arctic ecosystems. Due to the coarse spatial and temporal resolution of available satellite data, previous studies mainly focused on lake-scale investigations of melting/freezing, hindering the detection of subtle patterns within heterogeneous landscapes. To fill this knowledge gap, we developed a new approach for fine-resolution mapping of Pan-Arctic lake ice-off phenology. Using the Scene Classification Layer data derived from dense Sentinel-2 time series images, we estimated the pixel-by-pixel ice break-up end date information by seeking the transition time point when the pixel is completely free of ice. Applying this approach on the Google Earth Engine platform, we mapped the spatial distribution of the break-up end date for 45,532 lakes across the entire Arctic (except for Greenland) for the year 2019. The evaluation results suggested that our estimations matched well with both in situ measurements and an existing lake ice phenology product. Based on the generated map, we estimated that the average break-up end time of Pan-Arctic lakes is 172 ± 13.4 (measured in day of year) for the year 2019. The mapped lake ice-off phenology exhibits a latitudinal gradient, with a linear slope of 1.02 days per degree from 55°N onward. We also demonstrated the importance of lake and landscape characteristics in affecting spring lake ice melting. The proposed approach offers new possibilities for monitoring the seasonal Arctic lake ice freeze–thaw cycle, benefiting the ongoing efforts of combating and adapting to climate change.

INVESTIGATION OF SENTINEL-1 TIME SERIES FOR SENSITIVITY TO FERN VEGETATION IN AN EUROPEAN TEMPERATE FOREST

Abstract. In this study, a dense Copernicus Sentinel-1 time series is analyzed to gain a better understanding of the influence of undergrowth vegetation, in particular of eagle fern (Pteridium aquilinum), on the C-band SAR signal in a temperate forest in the Free State of Thuringia, Germany. Even if signals from the ground below the canopy may not be expected at C-band, previous studies showed seasonal fluctuations of the backscatter for temperate forests without canopy closure, notably for evergreen coniferous stands. Many factors can be responsible for these observed fluctuations, but in this study, we analyze one possible factor: the presence of undergrowth vegetation, in particular, of fern. Especially, the Sentinel-1 backscatter signal is analyzed for different acquisition configurations regarding its temporal and its spatial stability at different growth stages. This time series study shows that a difference of backscattered signal of up to 0.7 dB exists between forest patches with a dense fern density in the understory and the ones with low undergrowth vegetation. This signal difference depends on the season and is remarkably strong comparing winter (no fern undergrowth) with summer (major fern undergrowth).

Recurrent-based regression of Sentinel time series for continuous vegetation monitoring

Dense time series of optical satellite imagery describing vegetation activity provide essential information for the efficient and regular monitoring of vegetation. Nevertheless, the temporal resolution of optical sensors is strongly affected by cloud cover, resulting in significant missing information. The use of complementary acquisitions, such as Synthetic Aperture Radar (SAR) data, opens the door to the development of new multi-sensor methodologies aiming at the reconstruction of missing information. However, the joint exploitation of new radar and optical missions, such as the Sentinel, raises new challenges given the different nature and response of the two data sources. In this work, the SenRVM methodology is proposed as a new multi-sensor approach to regress SAR time series towards Normalized Difference Vegetation Index (NDVI). A deep Recurrent Neural Network architecture which integrates SAR acquisitions and ancillary data is adopted. The regression task permits a continuous optical temporal resolution of 6 days. Multiple experiments are carried out to assess the SenRVM framework by studying two large-scale areas in France. Through an extensive interpretation of the results, SenRVM is evaluated on three main vegetation types (grasslands, crops, and forests). High accurate results (R2 > 0.83 and MAE < 0.05) over more than 140,000 vegetation polygons are obtained, both for multi-class and class-specific models. The importance of the SenRVM input features is discussed through an ablation study, highlighting that the relevance of the features differs for the various classes. The good performances of SenRVM are assessed by evaluating the quality of the reconstruction over short- and long-term data gaps. Results are compared to five state-of-the-art methodologies. Finally, some preliminary experiments are first carried out to show how SenRVM results could be used to improve the existing cloud & shadow masks. The recovering of time series breaks caused by vegetation cover changes is also illustrated.

Detecting subtle change from dense Landsat time series: Case studies of mountain pine beetle and spruce beetle disturbance

In contrast to abrupt changes caused by land cover conversion, subtle changes driven by a shift in the condition, structure, or other biological attributes of land often lead to minimal and slower alterations of the terrestrial surface. Accurate mapping and monitoring of subtle change are crucial for an early warning of long-term gradual change that may eventually result in land cover conversion. Freely accessible moderate-resolution datasets such as the Landsat archive have great potential to characterize subtle change by capturing low-magnitude spectral changes in long-term observations. However, past studies have reported limited success in accurately extracting subtle changes from satellite-based time series analysis. In this study, we introduce a supervised framework named ‘PIDS’ to detect subtle forest disturbance from a comprehensive Landsat data archive by leveraging disturbance-based calibration sites. PIDS consists of four components: (1) Parameter optimization; (2) Index selection; (3) Dynamic stratified monitoring; and (4) Spatial consideration. PIDS was applied to map the early stage of bark beetle infestations (i.e., a lower per-pixel fraction of trees cover that show visual signs of infestation), which are a typical example of subtle change in conifer forests. Landsat Analysis Ready Data were used as the time series inputs for mapping mountain pine beetle and spruce beetle disturbance between 2001 and 2019 in Colorado, USA. PIDS-detection map assessment showed that the overall performance of PIDS (namely ‘F1 score’) was 0.86 for mountain pine beetle and 0.73 for spruce beetle, making a substantial improvement (> 0.3) compared to other approaches/products including COntinuous monitoring of Land Disturbance, LandTrendr, and the National Land Cover Database forest disturbance product. A sub-pixel analysis of tree canopy mortality percentage was performed by linking classified high-resolution (0.3- and 1-m) aerial imagery and 30-m PIDS-detection maps. Results show that PIDS typically detects mountain pine beetle infestation when ≥56% of a Landsat pixel is occupied by red-stage canopy mortality (one year after initial infestation), and spruce beetle infestation when ≥55% is occupied by gray-stage mortality (two years after initial infestation). This study addresses an important methodological goal pertinent to the utility of event-based reference samples for detecting subtle forest change, which could be potentially applied to other types of subtle land change.

Sentinel-1 Backscatter Analysis and Radiative Transfer Modeling of Dense Winter Wheat Time Series

This study evaluates a temporally dense VV-polarized Sentinel-1 C-band backscatter time series (revisit time of 1.5 days) for wheat fields near Munich (Germany). A dense time series consisting of images from different orbits (varying acquisition) is analyzed, and Radiative Transfer (RT)-based model combinations are adapted and evaluated with the use of radar backscatter. The model shortcomings are related to scattering mechanism changes throughout the growth period with the use of polarimetric decomposition. Furthermore, changes in the RT modeled backscatter results with spatial aggregation from the pixel to field scales are quantified and related to the sensitivity of the RT models, and their soil moisture output are quantified and related to changes in backscatter. Therefore, various (sub)sets of the dense Sentinel-1 time series are analyzed to relate and quantify the impact of the abovementioned points on the modeling results. The results indicate that the incidence angle is the main driver for backscatter differences between consecutive acquisitions with various recording scenarios. The influence of changing azimuth angles was found to be negligible. Further analyses of polarimetric entropy and scattering alpha angle using a dual polarimetric eigen-based decomposition show that scattering mechanisms change over time. The patterns analyzed in the entropy-alpha space indicate that scattering mechanism changes are mainly driven by the incidence angle and not by the azimuth angle. Besides the analysis of differences within the Sentinel-1 data, we analyze the capability of RT model approaches to capture the observed Sentinel-1 backscatter changes due to various acquisition geometries. For this, the surface models “Oh92” or “IEM_B” (Baghdadi’s version of the Integral Equation Method) are coupled with the canopy model “SSRT” (Single Scattering Radiative Transfer). To resolve the shortcomings of the RT model setup in handling varying incidence angles and therefore the backscatter changes observed between consecutive time steps of a dense winter wheat time series, an empirical calibration parameter (coef) influencing the transmissivity (T) is introduced. The results show that shortcomings of simplified RT model architectures caused by handling time series consisting of images with varied incidence angles can be at least partially compensated by including a calibration coefficient to parameterize the modeled transmissivity for the varying incidence angle scenarios individually.

Evaluating the impacts of models, data density and irregularity on reconstructing and forecasting dense Landsat time series

We evaluated the performance of a variety of time series models, that include the harmonic (HR) model, autoregressive (AR) model, linear Gaussian state-space (LGSS) model, cubic spline (SP) model, double logistic (DL) model, and asymmetric Gaussian (AG) model, for reconstructing (all six models) and forecasting (HR, AR, and LGSS models) dense Landsats 5–7 time series based on 4562 samples. To remove the impact of land change and human interventions on data reconstruction and forecasting, this evaluation excluded croplands and samples changed between 2000 and 2011. Results show that the widely used HR model is not a good model for data reconstruction but outperforms other models in data forecasting. The DL and AG models have the best performance in data reconstruction but cannot forecast Landsat observations. The AR and LGSS models shared similar performance in reconstructing Landsat data but are less ideal for data forecasting, particularly for the LGSS model. Integrating the HR (for forecasting) and DL or AG (for reconstruction) is recommended to improve land change detection and land cover classification results. We also evaluated the impact of data density and irregularity on reconstructing and forecasting Landsat observations. When the data density is low (<7 clear observations per year), the increase of data density can substantially improve the performance of data reconstruction and forecasting, but when the data density is higher than 7 and less than 17, the noise in the data dominates the results, and slightly lower reconstruction and forecasting accuracy is observed. When the data density is higher than 17, model performance improves with the increase of data density again. Therefore, we recommend analyzing Landsat time series for places with data density higher than 7 clear observations per year if possible. On the other hand, data irregularity has a moderate impact on data reconstruction and forecasting. When the irregularity is less than 1, the smaller the irregularity the better the performance in both data reconstruction and forecasting, and when the irregularity is higher than 2.5, it will have a more substantial negative impact. Therefore, time series analysis using Landsat data with an irregularity less than 1 is generally recommended, and time series with an irregularity larger than 2.5 should be avoided if possible.

Spatiotemporal image fusion in Google Earth Engine for annual estimates of land surface phenology in a heterogenous landscape

Currently, quantifying phenology at landscape to regional scales is not feasible with field data or near-surface sensors. Consequently, the spatial and temporal complexity of phenology has been assessed using satellite-based estimates (land surface phenology, LSP). While estimates from Moderate Resolution Imaging Spectroradiometer (MODIS) capture intraannual patterns of phenology, they have relatively low spatial resolution. Estimates from sensors like Landsat capture finer spatial detail but are often limited by Landsat’s temporal resolution. We implemented a spatio-temporal image fusion method on the Google Earth Engine (GEE) platform and used the resulting dense time series of images to estimate intraannual LSP at 30-meter resolution. We utilized Landsat 8 surface reflectance and MODIS NBAR (Nadir BRDF-Adjusted Reflectance; MCD43A4) images from 2016 and 2017 in the interior Pacific Northwest of the United States. Images predicted from the GEE image fusion algorithm were evaluated with true Landsat observations and compared with the accuracy achieved by executing the original ESTARFM algorithm. Excluding snow and cloud obscured observations, the algorithm produced approximately 215 observations per 30-meter pixel in 2017. Root mean squared prediction error (RMSPE) of Normalized Difference Vegetation Index (NDVI) for the GEE predicted images ranged from 0.032 to 0.066, and the RMSPE for the original ESTARFM predicted images from the ranged from 0.027 to 0.064. Phenometric estimates were evaluated with near-surface sensors (PhenoCams) in shrubland, conifer, and agricultural sites and field observations of phenology in grassland, open-pine, and mixed-conifer sites. Although phenometric estimates were dissimilar at all PhenoCam sites, the general temporal pattern of the GEE image fusion and PhenoCam time series was often similar. The start of season derived from the GEE image fusion time series had closer correspondence to the PhenoCam-derived start of season at the shrubland site (13 days) than the agriculture and conifer sites. The end of season was closest at one of the conifer sites and the agriculture site (22 and 31 days, respectively). Trends of some of the field-based phenology observations aligned with phenometrics estimated from the image fusion time series. At the grassland and open-pine field sites, the phenometrics from GEE image fusion were associated with phenophase trends of dominant plant functional types. Though characterizing LSP within the interior Pacific Northwest remains a challenge, this study demonstrates that image fusion implemented in GEE can produce a densified time series capable of capturing seasonal trends in NDVI related to vegetation phenology, which can be used to estimate intraannual phenometrics.

Detecting tropical selective logging with C-band SAR data may require a time series approach

Selective logging is the primary driver of forest degradation in the tropics and reduces the capacity of forests to harbour biodiversity, maintain key ecosystem processes, sequester carbon, and support human livelihoods. While the preceding decade has seen a tremendous improvement in the ability to monitor forest disturbances from space, large-scale (spatial and temporal) forest monitoring systems have almost universally relied on optical satellite data from the Landsat program, whose effectiveness is limited in tropical regions with frequent cloud cover. Synthetic aperture radar (SAR) data can penetrate clouds and have been utilized in forest mapping applications since the early 1990s, but only recently has SAR data been widely available on a scale sufficient to facilitate pan-tropical selective logging detection systems. Here, a detailed selective logging dataset from three lowland tropical forest regions in the Brazilian Amazon was used to assess the effectiveness of SAR data from Sentinel-1, RADARSAT-2, and Advanced Land Observing Satellite-2 Phased Arrayed L-band Synthetic Aperture Radar-2 (ALOS-2 PALSAR-2) for monitoring tropical selective logging. We built Random Forests models aimed at classifying pixel-based differences between logged and unlogged areas. In addition, we used the Breaks For Additive Season and Trend (BFAST) algorithm to assess if a dense time series of Sentinel-1 imagery displayed recognizable shifts in pixel values after selective logging. In general, Random Forests classification with SAR data (Sentinel-1, RADARSAT-2, and ALOS-2 PALSAR-2) performed poorly, having high commission and omission errors for logged observations. This suggests little to no difference in pixel-based metrics between logged and unlogged areas for these sensors, particularly at lower logging intensities. In contrast, the Sentinel-1 time series analyses indicated that areas under higher intensity selective logging (> 20 m3 ha−1) show a distinct spike in the number of pixels that included a breakpoint during the logging season. BFAST detected breakpoints in 50% of logged pixels and exhibited a false alarm rate of approximately <5% in unlogged forest. Overall our results suggest that SAR data can be used in time series analyses to detect tropical selective logging at high intensity logging locations (> 20 m3 ha−1) within the Amazon.

When image correlation is needed: Unravelling the complex dynamics of a slow-moving landslide in the tropics with dense radar and optical time series

Slow-moving landslides exhibit persistent but non-uniform motion at low rates which makes them exceptional natural laboratories to study the mechanisms that control the dynamics of unstable hillslopes. Here we leverage 4.5+ years of satellite-based radar and optical remote sensing data to quantify the kinematics of a slow-moving landslide in the tropical rural environment of the Kivu Rift, with unprecedented high spatial and temporal resolution. We measure landslide motion using sub-pixel image correlation methods and invert these data into dense time series that capture weekly to multi-year changes in landslide kinematics. We cross-validate and compare our satellite-based results with very-high-resolution Unoccupied Aircraft System topographic datasets, and explore how rainfall, simulated pore-water pressure, and nearby earthquakes control the overall landslide behaviour. The landslide exhibited seasonal and multi-year velocity variations that varied across the landslide kinematic units. While rainfall-induced changes in pore-water pressure exerts a primary control on the landslide motion, these alone cannot explain the observed variability in landslide behaviour. We suggest instead that the observed landslide kinematics result from internal landslide dynamics, such as extension, compression, material redistribution, and interactions within and between kinematic units. Our study provides, a rare, detailed overview of the deformation pattern of a landslide located in a tropical environment. In addition, our work highlights the viability of sub-pixel image correlation with long time series of radar-amplitude data to quantify surface deformation in tropical environments where optical data is limited by persistent cloud cover and emphasize the importance of exploiting synergies between multiple types of data to capture the complex kinematic pattern of landslides.

Modelling the productivity of Siberian larch forests from Landsat NDVI time series in fragmented forest stands of the Mongolian forest-steppe

The monitoring of the spatial and temporal dynamics of vegetation productivity is important in the context of carbon sequestration by terrestrial ecosystems from the atmosphere. The accessibility of the full archive of medium-resolution earth observation data for multiple decades dramatically improved the potential of remote sensing to support global climate change and terrestrial carbon cycle studies. We investigated a dense time series of multi-sensor Landsat Normalized Difference Vegetation Index (NDVI) data at the southern fringe of the boreal forests in the Mongolian forest-steppe with regard to the ability to capture the annual variability in radial stemwood increment and thus forest productivity. Forest productivity was assessed from dendrochronological series of Siberian larch (Larix sibirica) from 15 plots in forest patches of different ages and stand sizes. The results revealed a strong correlation between the maximum growing season NDVI of forest sites and tree ring width over an observation period of 20 years. This relationship was independent of the forest stand size and of the landscape’s forest-to-grassland ratio. We conclude from the consistent findings of our case study that the maximum growing season NDVI can be used for retrospective modelling of forest productivity over larger areas. The usefulness of grassland NDVI as a proxy for forest NDVI to monitor forest productivity in semi-arid areas could only partially be confirmed. Spatial and temporal inconsistencies between forest and grassland NDVI are a consequence of different physiological and ecological vegetation properties. Due to coarse spatial resolution of available satellite data, previous studies were not able to account for small-scaled land-cover patches like fragmented forest in the forest-steppe. Landsat satellite-time series were able to separate those effects and thus may contribute to a better understanding of the impact of global climate change on natural ecosystems.

Evaluating the effect of stand properties and site conditions on the forest reflectance from Sentinel-2 time series.

Forest stand reflectance at the canopy level results from various factors, such as vegetation chemical properties, leaf morphology, canopy structure, and tree sizes. These factors are dependent on the species, age, and health statuses of trees, as well as the site conditions. Sentinel-2 imagery with the high spatial, spectral, and temporal resolution, has enabled analysis of the relationships between vegetation properties and their spectral responses at large spatial scales. A comprehensive study of these relationships is needed to understand the drivers of vegetation spectral patterns and is essential from the point of view of remote sensing data interpretation. Our study aimed to quantify the site and forest parameters affecting forest stands reflectance. The analysis was conducted for common beech-, silver fir- and Scots pine-dominated stands in a mountainous area of the Polish Carpathians. The effect of stands and site properties on reflectance in different parts of the growing season was captured using the dense time series provided by Sentinel-2 from 2018-2019. The results indicate that the reflectance of common beech stands is mainly influenced by elevation, particularly during spring and autumn. Other factors influencing beech stand reflectance include the share of the broadleaved understory, aspect, and, during summer, the age of stands. The reflectance of coniferous species, i.e., Scots pine and silver fir, is mainly influenced by the age and stand properties, namely the crown closure and stand density. The age is a primary driver for silver fir stands reflectance changes, while the stand properties have a large impact on Scots pine stands reflectance. Also, the understory influences Scots pine stands reflectance, while there appears to be no impact on silver fir stands. The influence of the abovementioned factors is highly diverse, depending on the used band and time of the season.

Variation in aboveground biomass in forests and woodlands in Tanzania along gradients in environmental conditions and human use

Disturbed African tropical forests and woodlands have the potential to contribute to climate change mitigation. Therefore, there is a need to understand how carbon stocks of disturbed and recovering tropical forests are determined by environmental conditions and human use. In this case study, we explore how gradients in environmental conditions and human use determine aboveground biomass (AGB) in 1958 national forest inventory (NFI) plots located in forests and woodlands in mainland Tanzania. Plots were divided into recovering forests (areas recovering from deforestation for <25years) and established forests (areas consistently defined as forests for ⩾25 years). This division, as well as the detection of year of forest establishment, was obtained through the use of dense satellite time series of forest cover probability. In decreasing order of importance, AGB in recovering forests unexpectedly decreased with water availability, increased with surrounding tree cover and time since establishment, and decreased with elevation, distance to roads, and soil phosphorus content. AGB in established forests unexpectedly decreased with water availability, increased with surrounding tree cover, and soil nitrogen content, and decreased with elevation. AGB in recovering forests increased by 0.4 Mg ha−1yr−1 during the first 20 years following establishment. Our results can serve as the basis of carbon sink estimates in African recovering tropical forests and woodlands, and aid in forest landscape restoration planning.

Analysis of the Spatio-Temporal Evolution of Dredging from Satellite Images: A Case Study in the Principality of Asturias (Spain)

One of the fundamental tasks in the maintenance of port operations is periodic dredging. These dredging operations facilitate the elimination of sediments that the coastal dynamics introduce. Dredging operations are increasingly restrictive and costly due to environmental requirements. Understanding the condition of the seabed before and after dredging is essential. In addition, determining how the seabed has behaved in recent years is important to consider when planning future dredging operations. In order to analyze the behavior of sediment transport and the changes to the seabed due to sedimentation, studies of littoral dynamics are conducted to model the deposition of sediments. Another methodology that could be used to analyze the real behavior of sediments would be to study and compare port bathymetries collected periodically. The problem with this methodology is that it requires numerous bathymetric surveys to produce a sufficiently significant analysis. This study provides an effective solution for obtaining a dense time series of bathymetry mapping using satellite data, and enables the past behavior of the seabed to be examined. The methodology proposed in this work uses Sentinel-2A (10 m resolution) satellite images to obtain historical bathymetric series by the development of a random forest algorithm. From these historical bathymetric series, it is possible to determine how the seabed has behaved and how the entry of sediments into the study area occurs. This methodology is applied in the Port of Luarca (Principality of Asturias), obtaining satellite images and extracting successive bathymetry mapping utilizing the random forest algorithm. This work reveals how once the dock was dredged, the sediments were redeposited and the seabed recovered its level prior to dredging in less than 2 months.

Remotely sensed birch forest resilience against climate change in the northern China forest-steppe ecotone

Assessments of forest resilience to climate change are becoming increasingly urgent with more frequent drought events. In this study, we examined warming-induced variations in canopy greenness and water content in developing a robust approach to monitor forest resilience with the dense time series of the Landsat normalized difference vegetation index (NDVI) and normalized difference infrared index (NDII). Remote sensing observations and field data from five sites in the northern China forest-steppe ecotone were selected for assessment. The results reveal that the NDVI and NDII were negatively correlated with the investigated mortality ratios (with R2 values of 0.45 and 0.27, respectively), corresponding to the canopy greenness reduction and water loss, respectively, in the context of forest mortality. We further observed four patterns of birch forest resilience based on the NDVI-NDII coordinated variation trends identified from the Mann-Kendall (MK) test. Accordingly, arid timberline forests were found to exhibit greater resilience with more significant recovery of greenness and water content even after a canopy decline period of more than 10 years. Larger standard deviation (SD) values of the NDVI residual time series (all above 0.07) and a longer time lag of the NDII variations relative to those of the NDVI were observed, indicating that the greenness changes dominated the canopy dynamics observed in birch forest. Compared to traditional field surveys, remote sensing techniques focus on the continuous and quasi-synchronous monitoring of canopy dynamics, contributing to a more accurate detection and prediction of semiarid forest resilience on large scales.

Understanding Dense Time Series of Sentinel-1 Backscatter from Rice Fields: Case Study in a Province of the Mekong Delta, Vietnam

Rice is the primary staple food of more than half of the world’s population and plays an especially important role in global economy, food security, water use, and climate change. The usefulness of Synthetic Aperture Radars (SAR) for rice mapping and monitoring has been demonstrated locally in many studies, in particular in the last five years with the availability of an unprecedented amount of free Sentinel-1 data within the Copernicus program. However, although earlier studies from the 1990s have laid the foundations of the physical understanding of the SAR response of rice fields, the more recent studies tend to overlook this aspect and to favor instead approaches driven by supervised learning which provide accurate results locally but cannot necessarily be extended to wide areas. The objective of this study is to analyze in detail the backscatter temporal variation of rice fields, using Sentinel-1 from 2015 to 2020 and in-situ data for the 5 rice seasons over 2 years 2017–2018, in order to derive robust SAR-based indicators useful for rice monitoring applications, which are essential for planning, monitoring and food security applications. The test region is the An Giang province, in the Mekong River Delta, Vietnam, one of the world’s major rice regions which presents a diversity in rice cultivation practices, in cropping density, and in crop calendar. The SAR data have been analyzed as a function of rice parameters, and the temporal and polarization behaviors of the radar backscatter of different rice varieties have been interpreted physically. New backscatter indicators for the detection of rice paddy area, the estimation of the sowing date, phenological stage and the mapping of the short cycle and long cycle rice varieties have been developed and discussed regarding the generality of the methods with respect to the rice cultural practices and the SAR data characteristics.

Mapping Forest Disturbance Due to Selective Logging in the Congo Basin with RADARSAT-2 Time Series

Dense time series of stripmap RADARSAT-2 data acquired in the Multilook Fine mode were used for detecting and mapping the extent of selective logging operations in the tropical forest area in the northern part of the Republic of the Congo. Due to limited radiometric sensitivity to forest biomass variation at C-band, basic multitemporal change detection approach was supplemented by spatial texture analysis to separate disturbed forest from intact. The developed technique primarily uses multi-temporal aggregation of orthorectified synthetic aperture radar (SAR) imagery that are acquired before and after the logging operations. The actual change analysis is based on textural features of the log-ratio image calculated using two SAR temporal composites compiled of SAR scenes acquired before and after the logging operations. Multitemporal aggregation and filtering of SAR scenes decreased speckle and made the extracted textural features more prominent. The overall detection accuracy was around 80%, with some underestimation of the area of forest disturbance compared to reference based on optical data. The user’s accuracy for disturbed forest varied from 76.7% to 94.9% depending on the accuracy assessment approach. We conclude that change detection utilizing RADARSAT-2 time series represents a useful instrument to locate areas of selective logging in tropical forests.

Thermal Noise Removal From Polarimetric Sentinel-1 Data

This study proposes, for the first time, an approach to remove thermal noise from the wave coherency matrix, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathrm {C_{2}}$ </tex-math></inline-formula> , estimated from single-look complex dual-polarization Interferometric Wide Swath mode Sentinel-1 synthetic aperture radar data. The approach is straightforward; it exploits the ThermalNoiseRemoval module, provided by the European Space Agency (ESA) in its Sentinel Application Platform (SNAP) software, to remove thermal noise from the channel intensities. Then, noise correction on the complex data is applied, in order to estimate the noise-free <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathrm {C_{2}}$ </tex-math></inline-formula> matrix. As a further novelty, the proposed approach can be implemented in SNAP, through the use of a processing graph that is here provided. The method is applied on a dense time series of Sentinel-1 data, collected on an agricultural area located near Seville, Spain. The impact of thermal noise on the estimation of the eigendecomposition parameters of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathrm {C_{2}}$ </tex-math></inline-formula> , i.e., entropy ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$H_{2}$ </tex-math></inline-formula> ), average alpha angle ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\overline {\alpha _{2}}$ </tex-math></inline-formula> ), and anisotropy ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$A_{2}$ </tex-math></inline-formula> ), is assessed for different land-cover types, namely river, rice, forest, and urban areas. Monte Carlo simulations are implemented to assess the performance of the proposed approach in estimating <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$H_{2}$ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\overline {\alpha _{2}}$ </tex-math></inline-formula> , and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$A_{2}$ </tex-math></inline-formula> . Results show that the proposed noise removal method improves the estimation of these parameters for the considered land-cover classes.

Developing a new disturbance index for tracking gradual change of forest ecosystems in the hilly red soil region of southern China using dense Landsat time series

Gradual change is prevalent across the forest landscape and generates long-lasting effects for the landscape surface; thus, tracking long-term gradual change can effectively characterize forest change processes. The objective of this study was to establish a vegetation index for change monitoring so as to determine long-term gradual change processes of forest ecosystems in typical red soil areas. The study area was located in Hengyang City, Hunan Province, China. Landsat images, field survey, and auxiliary data were collected to devise a disturbance sensitive vegetation index (DSVI) as an indicator of forest change. Long-term (1985–2019) forest changes were detected using the LandTrendr algorithm on the Google Earth Engine (GEE), while three evaluation aspects of velocity, frequency, and variance were used to analyze the processes of forest gradual change in red soil regions. Results indicate that the DSVI is a suitable index for forest change detection due to its stronger sensitivity compared to other indexes. Further, it shows excellent change detection ability for different types of gradual changes, such as those caused by drought and significant soil erosion. Furthermore, 97.26% of the forests showed gradual change, and approximately 2/3 of monitored forests showed an increasing growth trend while 1/3 showed a decreasing trend. The dominant (28.33%) forest disturbance frequency indicated instability in red soil regions. Dispersion degree of forest variance was mainly low (48.46%) or medium (28.84%). This research establishes the DSVI as a promising method to track forest gradual change and contributes to a better understanding of gradual change processes of forest landscape over time.

European Wide Forest Classification Based on Sentinel-1 Data

The constellation of two Sentinel-1 satellites provides an unprecedented coverage of Synthetic Aperture Radar (SAR) data at high spatial (20 m) and temporal (2 to 6 days over Europe) resolution. The availability of dense time series enables the analysis of the SAR temporal signatures and exploitation of these signatures for classification purposes. Frequent backscatter observations allow derivation of temporally filtered time series that reinforce the effect of changes in vegetation phenology by limiting the influence of short-term changes related to environmental conditions. Recent studies have already shown the potential of multitemporal Sentinel-1 data for forest mapping, forest type classification (coniferous or broadleaved forest) as well as for derivation of phenological variables at local to national scales. In the present study, we tested the viability of a recently published multi-temporal SAR classification method for continental scale forest mapping by applying it over Europe and evaluating the derived forest type and tree cover density maps against the European-wide Copernicus High Resolution Layers (HRL) forest datasets and national-scale forest maps from twelve countries. The comparison with the Copernicus HRL datasets revealed high correspondence over the majority of the European continent with overall accuracies of 86.1% and 73.2% for the forest/non-forest and forest type maps, respectively, and a Pearson correlation coefficient of 0.83 for tree cover density map. Moreover, the evaluation of both datasets against the national forest maps showed that the obtained accuracies of Sentinel-1 forest maps are almost within range of the HRL datasets. The Sentinel-1 forest/non-forest and forest type maps obtained average overall accuracies of 88.2% and 82.7%, respectively, as compared to 90.0% and 87.2% obtained by the Copernicus HRL datasets. This result is especially promising due to the facts that these maps can be produced with a high degree of automation and that only a single year of Sentinel-1 data is required as opposed to the Copernicus HRL forest datasets that are updated every three years.