Sentinel-2 Time Series Research Articles

A deep learning framework for crop mapping with reconstructed Sentinel-2 time series images

Timely and accurate access to regional scale crop plant area and spatial distribution is essential for regional agricultural production and food security, especially in the context of global population growth and climate change. Deep learning has become prevalent in crop mapping under complex conditions due to its powerful feature extraction and nonlinear ability. This study proposes a time-series image classification framework using Attention-based Bidirectional Gated Recurrent Unit (A-BiGRU) to map rice, maize, and soybean in Fujin, China, from reconstructed Sentinel-2 time-series images. Firstly, the reconstructed Sentinel-2 time-series images with 10 spectral dimensions and 22 temporal dimensions were obtained by linear interpolation and Savitzky-Golay (SG) filter. Then, a neural network, the A-BiGRU was developed to identify different crops by taking advantage of their unique growth patterns. The attention mechanism enables temporal neural networks to focus on the critical growth periods of crops. Additionally, the structure of GRU is simpler than that of long short-term memory (LSTM) and simple recurrent neural network (SRNN), which reduces the number of parameters and alleviates overfitting. Compared to GRU, BiGRU can fully uses the time-series information of the entire crop growth cycle. To assess the effectiveness of the proposed method, we compared two deep learning methods (LSTM and SRNN) and three widely used non-deep learning classifiers (Spectral Angle Mapping (SAM), Support Vector Machine (SVM)) and eXtreme Gradient Boosting (XGBoost). The results demonstrate that A-BiGRU achieved the highest accuracy, with an overall accuracy of 0.9804, a macro F1 score of 0.9788 and a kappa score of 0.9714. We also selected four typical regions and compared the classification results with optical images, which showed that the proposed method has a good recognition effect. Therefore, the A-BiGRU method is capable of achieving high-precision crop mapping.

Evaluation of soil moisture estimation techniques based on Sentinel-1 observations over wheat fields

Soil moisture (SM) is a key variable in agriculture and its monitoring is essential. SM determines the amount of water available to plants, having a direct impact on the development of crops, on the forecasting of crop yields and on the surveillance of food security. Microwave remote sensing offers a great potential for estimating SM because it is sensitive to the dielectric characteristics of observed surface that depend on surface soil moisture. The objective of this study is the evaluation of three change detection methodologies for SM estimation over wheat at the agricultural field scale based on Sentinel-1 time series: Short Term Change Detection (STCD), TU Wien Change Detection (TUWCD) and Multitemporal Bayesian Change Detection (MTBCD). Different methodological alternatives were proposed for the implementation of these techniques at the agricultural field scale. Soil moisture measurements from eight experimental wheat fields were used for validating the methodologies. All available Sentinel-1 acquisitions were processed and the eventual benefit of correcting for vegetation effects in backscatter time series was evaluated. The results were rather variable, with some experimental fields achieving successful performance metrics (ubRMSE ∼ 0.05 m3/m3) and some others rather poor ones (ubRMSE > 0.12 m3/m3). Evaluating median performance metrics, it was observed that both TUWCD and MTBCD methods obtained better results when run with vegetation corrected backscatter time series (ubRMSE ∼0.07 m3/m3) whereas STCD produced similar results with and without vegetation correction (ubRMSE ∼0.08 m3/m3). The soil moisture content had an influence on the accuracy of the different methodologies, with higher errors observed for drier conditions and rain-fed fields, in comparison to wetter conditions and irrigated fields. Taking into account the spatial scale of this case study, results were considered promising for the future application of these techniques in irrigation management.

Using Machine-Learning Algorithms to Predict Soil Organic Carbon Content from Combined Remote Sensing Imagery and Laboratory Vis-NIR Spectral Datasets

Understanding spatial and temporal variability in soil organic carbon (SOC) content helps simultaneously assess soil fertility and several parameters that are strongly associated with it, such as structural stability, nutrient cycling, biological activity, and soil aeration. Therefore, it appears necessary to monitor SOC regularly and investigate rapid, non-destructive, and cost-effective approaches for doing so, such as proximal and remote sensing. To increase the accuracy of predictions of SOC content, this study evaluated combining remote sensing time series with laboratory spectral measurements using machine and deep-learning algorithms. Partial least squares (PLS) regression, random forest (RF), and deep neural network (DNN) models were developed using Sentinel-2 (S2) time series of 58 sampling points of bare soil and according to three approaches. In the first approach, only S2 bands were used to calibrate and compare the performance of the models. In the second, S2 indices, Sentinel-1 (S1) indices, and S1 soil moisture were added separately during model calibration to evaluate their effects individually and then together. In the third, we added the laboratory indices incrementally and tested their influence on model accuracy. Using only S2 bands, the DNN model outperformed the PLS and RF models (ratio of performance to the interquartile distance RPIQ = 0.79, 1.36 and 1.67, respectively). Additional information improved performances only for model calibration, with S1 soil moisture yielding the most stable improvement among three iterations. Including equivalent indices of the S2 indices calculated using soil spectra obtained under laboratory conditions improved prediction of SOC, and the use of only two indices achieved good validation performances for the RF and DNN models (mean RPIQ = 2.01 and 1.77, respectively).

Understanding the spatio-temporal behaviour of the sunflower crop for subfield areas delineation using Sentinel‐2 NDVI time-series images in an organic farming system

The study investigates the suitability of time series Sentinel-2 NDVI-derived maps for the subfield detection of a sunflower crop cultivated in an organic farming system. The aim was to understand the spatio-temporal behaviour of subfield areas identified by the K-means algorithm from NDVI maps obtained from satellite images and the ground yield data variability to increase the efficiency of delimiting management zones in an organic farming system.Experiments were conducted on a surface of 29 ha. NDVI time series derived from Sentinel-2 images and k-means algorithm for rapidly delineating the sunflower subfield areas were used. The crop achene yields in the whole field ranged from 1.3 to 3.77 t ha−1 with a significant within-field spatial variability. The cluster analysis of hand-sampled data showed three subfields with achene yield mean values of 3.54 t ha−1 (cluster 1), 2.98 t ha−1 (cluster 2), and 2.07 t ha−1 (Cluster 3). In the cluster analysis of NDVI data, the k-means algorithm has early delineated the subfield crop spatial and temporal yield variability. The best period for identifying subfield areas starts from the inflorescences development stage to the development of the fruit stage. Analyzing the NDVI subfield areas and yield data, it was found that cluster 1 covers an area of 42.4% of the total surface and 50% of the total achene yield; cluster 2 covers 35% of both surface and yield. Instead, the surface of cluster 3 covers 22.2% of the total surface with 15% of achene yield. K-means algorithm derived from Sentinel-2 NDVI images delineates the sunflower subfield areas.Sentinel-2 images and k-means algorithms can improve an efficient assessment of subfield areas in sunflower crops. Identifying subfield areas can lead to site-specific long-term agronomic actions for improving the sustainable intensification of agriculture in the organic farming system.

ReCuSum: A polyvalent method to monitor tropical forest disturbances

Change detection methods based on Earth Observations are increasingly used to monitor rainforest in the intertropical band. Until recently, deforestation monitoring was mainly based on remotely sensed optical images which often face limitations in humid tropical areas due to frequent cloud coverage. This leads to late detections of disturbance events. Since the launch of Sentinel-1 acquiring images with a revisit time of 12 days and a spatial resolution of 5 × 20 m in Brazil, Synthetic Aperture Radar (SAR) images have been increasingly used to monitor deforestation. In this study, we propose a multitemporal version of the change detection method we previously applied to timeseries of Sentinel-1 SAR images, to monitor deforestation/degradation in the Congo rainforest. Our approach is based on a Cumulative Sum (CuSum) method combined with a spatial recombination of Critical Thresholds (CuSum cross-Tc). The newly developed multitemporal CuSum method (ReCuSum) was applied to a time-series of 82 dual polarization (VH, VV) Ground Range Detected (GRD) Sentinel-1 images acquired in the Para State, in the Brazilian Amazonia, from 29/09/2016 to 01/07/2019. The ReCuSum method consists of iteratively applying the CuSum cross-Tc to monitor multiple changes in a time-series by splitting the time-series at each date of detected change and by independently iterating over the time periods resulting from the splits. The number of changes in the time-series was then analysed according to the vegetation type (Forest, non-forest vegetation) determined by visual inspection of optical Sentinel-2 image and PlanetScope monthly mosaic. This showed a difference between non-forest vegetation and forested areas. A threshold based on the number of changes (Tnbc) was then developed to differentiate forest from non-forest disturbances. The ability to monitor non-forest vegetation was analysed: the CuSum cross-Tc detected up to 90% of the total non-forest vegetation area over the study region in the past period. After removing past disturbances and past non-forest vegetation, then removing the pixels covered with non-forest vegetations based on Tnbc, the application of the ReCuSum led to a precision of 81%, a recall of 68%, a kappa coefficient of 0.72 and a F1- score of 0.74 in forest disturbance monitoring. According to these results, ReCuSum applied to Sentinel-1 time-series of images can be used for efficient forest disturbance monitoring and for generating a forest / non-forest map after the application of newly developed post-processing steps. Sentinel-1 imagery can be used for both Forest / Non-forest mapping and for forest disturbance detection. ReCuSum was released as an open-source GIT project available at: https://forgemia.inra.fr/bertrand.ygorra/cusum-deforestation_monitoring.

Annual Field-Scale Maps of Tall and Short Crops at the Global Scale Using GEDI and Sentinel-2

Crop type maps are critical for tracking agricultural land use and estimating crop production. Remote sensing has proven an efficient and reliable tool for creating these maps in regions with abundant ground labels for model training, yet these labels remain difficult to obtain for many regions and years. NASA’s Global Ecosystem Dynamics Investigation (GEDI) spaceborne LiDAR instrument, originally designed for forest monitoring, has shown promise for distinguishing tall and short crops. In the current study, we leverage GEDI to develop wall-to-wall maps of short vs. tall crops on a global scale at 10 m resolution for 2019–2021. Specifically, we show that (i) GEDI returns can reliably be classified into tall and short crops after removing shots with extreme view angles or topographic slope, (ii) the frequency of tall crops over time can be used to identify months when tall crops are at their peak height, and (iii) GEDI shots in these months can then be used to train random forest models that use Sentinel-2 time series to accurately predict short vs. tall crops. Independent reference data from around the world are then used to evaluate these GEDI-S2 maps. We find that GEDI-S2 performed nearly as well as models trained on thousands of local reference training points, with accuracies of at least 87% and often above 90% throughout the Americas, Europe, and East Asia. A systematic underestimation of tall crop area was observed in regions where crops frequently exhibit low biomass, namely Africa and South Asia, and further work is needed in these systems. Although the GEDI-S2 approach only differentiates tall from short crops, in many landscapes this distinction is sufficient to map individual crop types (e.g., maize vs. soy, sugarcane vs. rice). The combination of GEDI and Sentinel-2 thus presents a very promising path towards global crop mapping with minimal reliance on ground data.

Investigating mangrove canopy phenology in coastal areas of China using time series Sentinel-1/2 images

Mangrove forests with high vegetation productivity play crucial roles in the coastal blue carbon ecosystem. Accurate monitoring of mangrove canopy phenology is essential to improve the survival rate of restoration plantings and maintain the health and sustainability of mangrove ecosystems. Remote sensing technology has been commonly used for monitoring vegetation phenology, while there remain knowledge gaps in exploring the remotely-sensed phenology of mangrove forests in coastal China. In this study, we investigated the phenological characteristics of two typical mangrove sites in coastal China using remote sensing indices based on Sentinel-1 and Sentinel-2 time series during 2016–2020, compared the performances of mangrove phenology detection across different remote sensing indices and spatial scales, and explored the influences of environmental factors based on meteorological data. The results demonstrated that enhanced vegetation index (EVI), red-edge band index (NDRE2), phenology index (NDPI), and radar vegetation index (RVI) were efficient in characterizing mangrove phenological trajectories. Sentinel-2 data can precisely describe the phenological characteristics of mangroves and has the highest correlation with ground-observed phenology data (r = −0.581), when compared to Landsat-8 and MODIS data. The meteorological factors of precipitation, humidity, and wind speed mainly led to the differences in mangrove phenology across the two study sites. This finding can improve our understanding of the phenological characteristics of mangrove forests in coastal China, which facilitates local governments to develop appropriate mangrove restoration and management policies.

A robust approach for large-scale cropping intensity mapping in smallholder farms from vegetation, brownness indices and SAR time series

Multiple cropping is a widespread agricultural intensification for increasing food production. National-scale Cropping Intensity (CI) mapping is important for achieving sustainable development goals. However, previous studies have largely relied on vegetation indices (VI) waves for detecting valid cropping cycles, which is challenged by the complexity of agricultural systems. i.e. winter crops with multiple VI waves and double rice with less distinctive waves. This study proposed a robust framework for Mapping cropping Intensity through better-characterizing crop Life cycles based on combined considerations of vegetative and productive Stages (MILS). The number of cropping cycles was estimated by the frequency of valid coupled patterns of vegetation and brownness indices from Sentinel-2 (S2) MultiSpectral Instrument (MSI) time series and further improved through fusing Sentinel-1 (S1) SAR and S2 data to reduce the omission errors of double rice. The MILS algorithm was implemented using the Google Earth Engine platform and applied to China, which is dominated by smallholder farms with diverse cropping patterns. This study produced the first 10 m updated CI map over conterminous China (CIChina10m) in 2020 by fusing S1 and S2 time series. The CIChina10m had an overall accuracy of 93.93% when validated with 14,468 widely spread reference sites. The cropping intensity was 1.2769 on the national scale, which illustrated higher values in the Middle and lower reaches of the Yangtze River plain (CI = 1.5879) and South China (CI = 1.5503). There were 1086,620, 412,620 and 1,441 km2 areas cultivated by single, double and triple cropping in China, which accounted for 72.4%, 27.5% and 0.1% of cropland, respectively. The proposed MILS algorithm showed good performances in detecting complex agricultural systems, which can be further applied to generate continental or global 10-m CI products with good quality. Codes of the MILS algorithm are publicly available at https://code.earthengine.google.com/a7f24f76291bf901ee25a130025a7ce6, and the first 10m national CI data products in China with good accuracy (ChinaCI10m) are also publicly accessed at https://doi.org/10.6084/m9.figshare.23939505.

R-Unet: A Deep Learning Model for Rice Extraction in Rio Grande do Sul, Brazil

Rice is one of the world’s three major food crops, second only to sugarcane and corn in output. Timely and accurate rice extraction plays a vital role in ensuring food security. In this study, R-Unet for rice extraction was proposed based on Sentinel-2 and time-series Sentinel-1, including an attention-residual module and a multi-scale feature fusion (MFF) module. The attention-residual module deepened the network depth of the encoder and prevented information loss. The MFF module fused the high-level and low-level rice features at channel and spatial scales. After training, validation, and testing on seven datasets, R-Unet performed best on the test samples of Dataset 07, which contained optical and synthetic aperture radar (SAR) features. Precision, intersection, and union (IOU), F1-score, and Matthews correlation coefficient (MCC) were 0.948, 0.853, 0.921, and 0.888, respectively, outperforming the baseline models. Finally, the comparative analysis between R-Unet and classic models was completed in Dataset 07. The results showed that R-Unet had the best rice extraction effect, and the highest scores of precision, IOU, MCC, and F1-score were increased by 5.2%, 14.6%, 11.8%, and 9.3%, respectively. Therefore, the R-Unet proposed in this study can combine open-source sentinel images to extract rice timely and accurately, providing important information for governments to implement decisions on agricultural management.

Burned area detection and mapping using time series Sentinel-2 multispectral images

Accurate mapping of burned areas (BAs) is essential for post-disaster reconstruction, air quality assessment, and estimation of fire emissions. Remote sensing has become the most practical means to map global/regional BAs. Space-borne coarse-resolution (250–1000 m) BA products play an essential role in fire-related science and applications, but small BAs are typically poorly represented due to resolution limitations. Moderate-resolution (10–30 m) imagery provides a more detailed spatial extent of BAs, but the various disturbances (e.g., plant phenology, cloud shading, flooding, harvesting, grazing, pests, and diseases) make application on a global scale challenging. Here, we propose an automatic BA detection algorithm based on time series images acquired by the Sentinel-2 MultiSpectral Instrument (MSI) and the Google Earth Engine. Specifically: (i) We determined an optimal set of spectral indices for MSI images based on global sampling to mitigate the background heterogeneity and BA spectral diversity. (ii) We jointly used three temporal scales, i.e., the single-temporal, dual-temporal, and short-time series MSI images, to identify potential BA regions (i.e., BA candidates). (iii) We integrated multi-source active fire (AF) products derived from the Moderate-resolution Imaging Spectroradiometer (MODIS), Visible Infrared Imaging Radiometer Suite (VIIRS), Operational Land Imager (OLI), and MSI instead of single-source products to refine BA candidates. (iv) We constructed an interannual time series to improve the BA candidates not confirmed by AF products. The proposed BA detection method was applied to different land cover and regional scales, including forests, grasslands, savannas, shrublands, and croplands at the regional (scene) and national levels (Portugal). Validation and inter-comparison with similar products demonstrated that the proposed method is reliable: the Dice coefficient of our detection results was improved by ∼7% and ∼ 9%, respectively, compared to the moderate-resolution BA products (e.g., LBA_CU for the conterminous United States and FireCCISFD20 for sub-Saharan Africa in 2019). These results suggest that the proposed method can accurately and automatically map BAs, providing a foundation for fire monitoring programs and climate change research.

Sentinel-2 time series: a promising tool in monitoring temperate species spring phenology

Abstract The changes in vegetation phenology observed in recent decades may have significant implications for carbon, water, and energy fluxes as well as ecosystem productivity. Phenology is an important indicator of a changing environment, and studying species-specific phenology can help understand their responses to varying climate conditions. While satellite-based phenology provides long-term information for large areas, it describes different traits than ground-based phenology. Nonetheless, emerging research on remote sensing-based phenology has shown its usefulness in determining inter- and intra-specific differences, as well as interannual changes. In this study, we examined the potential of the Sentinel-2 time series from 2018 to 2022 in assessing the spring phenology of seven temperate tree species. Three vegetation indices were calculated, namely, the Medium Resolution Imaging Spectrometer (MERIS) Terrestrial Chlorophyll Index, the Normalized Difference Vegetation Index, and the Enhanced Vegetation Index. We examined interspecific variations in the start of the season (SOS) and interannual variability. Additionally, we explored intraspecific variations in spring phenology among common beech stands across an elevation gradient. Our findings reveal substantial differences in Sentinel-2-derived phenology across species, with larch, birch, and hornbeam displaying an earlier SOS compared with oak and alder. The latest SOS dates were observed in 2021, whereas the earliest occurred in 2018 and the difference exceeded 20 days. It corresponded to the diverse spring meteorological conditions in the examined years, with the highest heat accumulation in 2018 and the lowest in 2021. Additionally, the three examined indices showed different phenological curves, with MERIS Terrestrial Chlorophyll Index exhibiting distinct patterns with a delayed start of the spring increase compared with the other two indices. The relationship between elevation and SOS dates was generally positive, but there was high variability in some years. The earliest SOS dates occurred at ~600 m a.s.l. and larger interannual variations were observed for stands located on the highest elevations. These results underscore the potential of Sentinel-2 in monitoring the spring phenology of temperate species, offering insights into inter- and intra-specific variations, as well as year-to-year changes. In the future, longer Sentinel-2 time series will improve estimates and facilitate the tracking of long-term changes. Such advancements hold relevance for forest management, carbon budget predictions, and understanding ecosystem dynamics.

Detection of Forest Fires through Deep Unsupervised Learning Modeling of Sentinel-1 Time Series

With an increase in the amount of natural disasters, the combined use of cloud-penetrating Synthetic Aperture Radar and deep learning becomes unavoidable for their monitoring. This article proposes a methodology for forest fire detection using unsupervised location-expert autoencoders and Sentinel-1 SAR time series. The models are trained on SAR multitemporal images over a specific area using a reference period and extract any deviating time series over that same area for the test period. We present three variations of the autoencoder, incorporating either temporal features or spatiotemporal features, and we compare it against a state-of-the-art supervised autoencoder. Despite their limitations, we show that unsupervised approaches are on par with supervised techniques, performance-wise. A specific architecture, the fully temporal autoencoder, stands out as the best-performing unsupervised approach by leveraging temporal information of Sentinel-1 time series using one-dimensional convolutional layers. The approach is generic and can be applied to many applications, though we focus here on forest fire detection in Canadian boreal forests as a successful use case.

Hierarchical extraction of cropland boundaries using Sentinel-2 time-series data in fragmented agricultural landscapes

Accurately extracting cropland field parcels from satellite data in fragmented agricultural landscapes presents huge challenges due to small field sizes and irregular field shapes. Here, we developed a hierarchical framework for agricultural field boundary extraction from Sentinel-2 satellites based on the concept of the degree in field fragmentation. Our framework focused on three tasks, including a core task for agricultural field extraction and two auxiliary tasks to address special situations in two different fragmented regions, namely the Plains-Basin scenes with relatively regular shapes and diverse crop types, and the Plateau-Hilly scenes, with irregular shapes and small field sizes. First, the field boundaries were delineated using a modified Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net), and post-processed using morphological filtering and the Douglas-Peucker algorithm (DPA). A rule-based decision fusion strategy assisted with a classification method was developed to address the under-segmentation problem in the Plains-Basin scenes, and an adaptive generalized buffer algorithm (AGBA) was used to address the incomplete-segmentation problem in the Plateau-Hilly scenes. We tested the proposed methods in Yangming County, Heilongjiang Province of China, using time series Sentinel-2 imageries with 10-m spatial resolution. Compared with the results from using R2U-Net or a single-temporal image alone, the fragmented field boundaries show an outstanding performance, with an overall accuracy of 86.42% and 84.15% and F1 score of 0.85 and 0.83 in Plains-Basin scenes and Plateau-Hilly scenes, respectively. Our proposed methods provide a feasible solution to finely extract field boundaries in highly fragmented and heterogeneous agricultural landscapes.

Fine-grained agricultural and pastoral information extraction using Sentinel-1 and Sentinel-2 intra-year time series in Jingyang District, Deyang City

Accurate and effective access to agricultural and pastoral information is of great practical importance for scientific planning in agriculture and pastoral husbandry and food security reserves. This paper utilised intra-annual monthly synthetic data from Sentinel-1 and Sentinel-2, combined with band information and index information to simultaneously obtain intra-annual phenological change information and extracted different types of information for random forest classification, thus obtaining the fine-grained agricultural and pastoral map of Jingyang District. On this basis, the effects of different data sources and the importance of different features on the classification results were further analysed. The results show that in the classification results of the Sentinel-1 data method, March feature is the most important and May feature has the least influence on the classification results. In the classification results of the Sentinel-2 data method and the combined Sentinel-1 data and Sentinel-2 data, the December feature has the greatest importance and the February feature has the least impact on the classification results. The NDVI index has the highest importance and the EVI index has the lowest importance score. The importance of band features varies in the classification results of different data sources. The accuracy of the agricultural and pastoral information in Jingyang District is 83.31% and the kappa is 0.76. Rice is the largest crop with an area of 239,830,800 m2, mainly located in the northwest and mountainous areas of Jingyang District. It is followed by green manure with an area of 60,274,000 m2, mainly in the mountainous part. This study demonstrates the feasibility of extracting agricultural and pastoral information from remote sensing images, and also provides basic data for future agricultural and pastoral planning in Jingyang District.

Can land surface phenology from Sentinel-2 time-series be used as an indicator of Macaronesian ecosystem dynamics?

Land surface phenology (LSP), the study of phenological patterns of vegetation using vegetation index (VI) time-series derived from multi-spectral satellite imagery, has helped to improve the understanding of the seasonal dynamics of terrestrial ecosystems from local to global scale. High spatial resolution satellite observations have emerged as a new possibility for monitoring the seasonal dynamics of heterogeneous ecosystems. Under-studied Macaronesian ecosystems have specific characteristics that make LSP estimation a complex task (e. g., frequent cloud cover, frequent presence of atmospheric aerosols, different degrees of vegetation density, diversity of plant species, landscape heterogeneity). Thus, this study aims to analyse the potential of LSP based on Sentinel-2 data to monitor the phenological dynamics of vegetation in the Macaronesian ecosystems of Canary Islands (Spain). NDVI time-series were generated using Sentinel-2 data from January 2018 to December 2021. NDVI time-series were smoothed using double logistic function. Three phenometrics, such as the start of the growing season (SOS), the end of the growing season (EOS) and the length of the growing season (LOS), were extracted using a threshold-based method (20%) only for pixels where the mean NDVI value of the smoothed four-year time-series was higher than 0.2. Growing season of the major of Macaronesian vegetation started between late summer and late autumn (start of the wet season) and ended between early spring and early summer (end of the wet season). The most representative Macaronesian tree species had a clear and marked seasonality, except for the laurel forest species. SOS of Juniperus phoenicea subsp. turbinata was slightly later than for Olea europaea var. sylvestris and Pinus canariensis, while the mean EOS was slightly later for Olea europaea var. sylvestris. The intra-specific variability of SOS and EOS in the laurel forests was very high. LSP derived from Sentinel-2 data contributed to understand the seasonal dynamics of the main Macaronesian ecosystems. However, the heterogeneity of laurel forests made it difficult the LSP estimation in laurel forests. Thus, complementary phenological approaches are suggested to improve the knowledge of the seasonal dynamics of this Macaronesian ecosystem.

Improving digital mapping of soil organic matter in cropland by incorporating crop rotation

Despite human activities are key influencing factors for cropland soil organic matter (SOM), detailed characterization of human activities has always been limited in the digital mapping of SOM due to the lack of proper representations of human’s cropland use activities. Crop rotation is an essential human agricultural practice significantly affecting the spatial–temporal variations of SOM due to the periodically dynamic changes of crops. Thus, incorporating crop rotation in the digital soil mapping holds high potential for improving SOM prediction. Here, we applied time-series radar Sentinel-1 and optical Sentinel-2 to map crop rotation systems by a hierarchical rule-based method. Then we explored the effectiveness of incorporating such information in predicting SOM by implementing various combinations of predictive variables. We chose a typical multiple cropping region with various crop rotations in southern China. The model performance was evaluated by 10-fold cross-validation. Results showed significant differences in SOM among the crop rotation systems, and the single rice rotated with vegetables has the highest SOM followed by the high-diversity vegetables and long-term orchard systems. Adding crop rotation enhanced the predictability of SOM with a decrease in RMSE by 7% and an increase in R2 by 24%. Furthermore, the crop rotation systems appeared more important in the predictive models than the soil, topographic, and climatic variables. Our results demonstrated the effectiveness of including crop rotation in predicting SOM over complex agricultural landscapes. Our study indicated that human activities should be characterized more detailedly in cropland soil mapping, and that crop rotation containing information on the seasonal dynamics of cropland may be an option for such characterization.

Developing an operational algorithm for near-real-time monitoring of crop progress at field scales by fusing harmonized Landsat and Sentinel-2 time series with geostationary satellite observations

Crop phenology has been widely detected from multiple historical satellite observations. Conversely, Near-Real-Time (NRT) monitoring of crop progress from timely available remote sensing data is barely investigated because of the lack of high-frequency cloud-free satellite observations and future potential crop development. To address the challenge, this study proposes a novel algorithm for operational NRT monitoring of crop progress at the field scale. This algorithm first fuses the high spatial resolution (30 m) Harmonized Landsat and Sentinel-2 (HLS) data and the high temporal frequent (10 min) Advanced Baseline Imager (ABI) observations to generate cloud-free time series of HLS-ABI EVI2 (two-band Enhanced Vegetation Index) with a Spatiotemporal Shape-Matching Model (SSMM). It then predicts future potential EVI2 values at a given pixel using a reference EVI2 time series obtained from the neighboring pixels in the preceding year. Integrating the currently available HLS-ABI observations and the predicted future EVI2 values to generate annual EVI2 time series, the algorithm finally detects six crop phenometrics including greenup onset, mid-greenup phase, maturity onset, senescence onset, mid-senescence phase, and dormancy onset. The NRT monitoring, which are separated as near-real-time prediction (phenological event detected after the occurrence), real-time prediction (phenological event detected around the occurrence), and short-term prediction (phenological event detected before the occurrence), are continuously updated and improved with new HLS and ABI observations at a weekly basis throughout the growing season. We evaluate the NRT monitoring against standard phenology products, PhenoCam observations, as well as the weekly Crop Progress Reports (CPRs) released from the National Agricultural Statistics Service (NASS) of the United States Department of Agriculture (USDA) in 2020 across Iowa. The evaluation demonstrates the robustness of the developed algorithm in NRT monitoring of crop phenology. Although the uncertainties are relatively large for short-term prediction compared with standard detections, the real-time prediction shows that the Mean Absolute Difference (MAD) is <10 days for greenup and dormancy onset, and ∼ 5 days in the other four phenometrics. Further, the real-time prediction aligns well with PhenoCam observations (R2 = 0.96, P < 0.001) with a MAD of 7.8 days. Moreover, the HLS-ABI real-time prediction of crop phenometrics is capable of tracking NASS crop progress closely with small time shifts (≤ 5 days) and significant correlations (R2 > 0.85, P < 0.001) for various phenological stages of corn and soybean. These results prove that the algorithm could be implemented for NRT monitoring of various crop phenometrics from field, state, to national scales.

New landscape-perspective exploration of Moso bamboo forests under on/off-year phenomena and human activities

On-year and off-year phenomena is an unique spatiotemporal characteristics of Moso bamboo forests (MBFs), the combination of multiple sources data to assess the impact of on-/off-year phenomena and human activities is a new perspective to better understand MBFs. There is an urgent need to explore how to characterize the spatial variability of MBFs, whether there is a pattern in the spatial distribution of MBFs boundary lines, and whether the causes of the boundary lines are directly related to human activities. In this paper, Anji County, with a large MBF area, was selected as the study area. Based on Sentinel-2 time-series data, topography and impermeability data, the on-year/off-year status and boundary lines of MBFs were determined, and the Integrated Moso bamboo On-off year Differentiation Index (IMODI) was proposed to evaluate spatial differentiation. Spatial heterogeneity, the shape pattern of boundary lines, and the impact of human activities on the MBFs were quantitatively analyzed, and 631.5 km2 of MBFs and 140 on-off boundaries were accurately obtained. The on-year and off-year MBFs and boundaries were mainly distributed in mountainous areas, especially at elevations of 100–400 m and on slopes of 10–35°. The patterns in elevation can be categorized as stable, single-change, ascending-descending and descending-ascending patterns. The patterns in slope can be summarized as stable, single-change, and trapezoidal patterns. The newly proposed IMODI shows the best performance in reflecting the spatial differentiation of MBFs, and the on-year and off-year phenomenon of MBFs was closely related to human activities because the boundary lines were mostly concentrated within 2 km of villages. The results of this paper provide a new perspective to understand landscape of MBFs and the integrated management of carbon sequestration and sinks.

Early-season and refined mapping of winter wheat based on phenology algorithms -a case of Shandong, China.

Winter wheat is one of the major food crops in China, and timely and effective early-season identification of winter wheat is crucial for crop yield estimation and food security. However, traditional winter wheat mapping is based on post-season identification, which has a lag and relies heavily on sample data. Early-season identification of winter wheat faces the main difficulties of weak remote sensing response of the vegetation signal at the early growth stage, difficulty of acquiring sample data on winter wheat in the current season in real time, interference of crops in the same period, and limited image resolution. In this study, an early-season refined mapping method with winter wheat phenology information as priori knowledge is developed based on the Google Earth Engine cloud platform by using Sentinel-2 time series data as the main data source; these data are automated and highly interpretable. The normalized differential phenology index (NDPI) is adopted to enhance the weak vegetation signal at the early growth stage of winter wheat, and two winter wheat phenology feature enhancement indices based on NDPI, namely, wheat phenology differential index (WPDI) and normalized differential wheat phenology index (NDWPI) are developed. To address the issue of " different objects with the same spectra characteristics" between winter wheat and garlic, a plastic mulched index (PMI) is established through quantitative spectral analysis based on the differences in early planting patterns between winter wheat and garlic. The identification accuracy of the method is 82.64% and 88.76% in the early overwintering and regreening periods, respectively, These results were consistent with official statistics (R2 = 0.96 and 0.98, respectively). Generalization analysis demonstrated the spatiotemporal transferability of the method across different years and regions. In conclusion, the proposed methodology can obtain highly precise spatial distribution and planting area information of winter wheat 4_6 months before harvest. It provides theoretical and methodological guidance for early crop identification and has good scientific research and application value.

Upscaling field-measured seasonal ground vegetation patterns with Sentinel-2 images in boreal ecosystems

ABSTRACT Aboveground biomass (AGB) and leaf area index (LAI) are key variables of ecosystem processes and functioning. Knowledge is lacking on how well the seasonal patterns of ground vegetation AGB and LAI can be detected by satellite images in boreal ecosystems. We conducted field measurements between May and September during one growing season to investigate the seasonal development of ground vegetation AGB and LAI of seven plant functional types (PFTs) across seven vegetation types (VTs) within three peatland and forest study areas in northern Finland. We upscaled field-measured AGB and LAI with Sentinel-2 (S2) imagery by applying random forest (RF) regressions. Field-measured AGB peaked around the first week of August and, in most cases, one to two weeks later than LAI. Regarding PFTs, deciduous vascular plants had clear unimodal seasonal patterns, while the AGB and LAI of evergreen vegetation and mosses remained steady over the season. Remote sensing regression models explained 24.2–50.2% of the AGB (RMSE: 78.8–198.7 g m−2) and 48.5–56.1% of the LAI (RMSE: 0.207–0.497 m2 m−2) across sites. Peatland-dominant sites and VTs had a higher prediction accuracy. S2-predicted peak dates of AGB and LAI were one to three weeks earlier than the field-based ones. Our findings suggest that boreal ground vegetation seasonality varies among PFTs and VTs and that S2 time series data can be applied to monitor its spatiotemporal patterns, especially in treeless regions.