DATimeS: A machine learning time series GUI toolbox for gap-filling and vegetation phenology trends detection

Belda, Santiago; Pipia, Luca; Morcillo-Pallarés, Pablo; Rivera-Caicedo, Juan Pablo; Amin, Eatidal; Grave, Charlotte De; Verrelst, Jochem

doi:10.1016/j.envsoft.2020.104666

Cited by 60 publications

(56 citation statements)

References 78 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Among the multiple MLRA approaches currently available, the kernel-based methods developed in a Bayesian framework deserve special attention, such as GPR [13]. Recent studies demonstrated the effectiveness of GPR for gap-filling of biophysical parameter time series [18][19][20] because the hyperparameters can be optimally set for each time series (one for each pixel in the area) with a single optimization procedure. Despite its clear strategic advantage, the most important shortcomings of this technique are their (1) high computational cost and (2) memory requirements [49] for hyperparameter optimization, which grows cubically and quadratically with the number of training points, respectively [50,51].…”

Section: Gapfilling Modelmentioning

confidence: 99%

“…Trained GPRs are usually highly flexible and accurate for prediction over new inputs closer to training data points, whereas their uncertainty increases when the new inputs are further away from the available training information. With these appealing properties, apart from retrieval applications, recent studies have demonstrated the effectiveness of GPR for time series gap-filling applications [18][19][20].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Green LAI Mapping and Cloud Gap-Filling Using Gaussian Process Regression in Google Earth Engine

et al. 2021

Self Cite

View full text Add to dashboard Cite

For the last decade, Gaussian process regression (GPR) proved to be a competitive machine learning regression algorithm for Earth observation applications, with attractive unique properties such as band relevance ranking and uncertainty estimates. More recently, GPR also proved to be a proficient time series processor to fill up gaps in optical imagery, typically due to cloud cover. This makes GPR perfectly suited for large-scale spatiotemporal processing of satellite imageries into cloud-free products of biophysical variables. With the advent of the Google Earth Engine (GEE) cloud platform, new opportunities emerged to process local-to-planetary scale satellite data using advanced machine learning techniques and convert them into gap-filled vegetation properties products. However, GPR is not yet part of the GEE ecosystem. To circumvent this limitation, this work proposes a general adaptation of GPR formulation to parallel processing framework and its integration into GEE. To demonstrate the functioning and utility of the developed workflow, a GPR model predicting green leaf area index (LAIG) from Sentinel-2 imagery was imported. Although by running this GPR model into GEE any corner of the world can be mapped into LAIG at a resolution of 20 m, here we show some demonstration cases over western Europe with zoom-ins over Spain. Thanks to the computational power of GEE, the mapping takes place on-the-fly. Additionally, a GPR-based gap filling strategy based on pre-optimized kernel hyperparameters is also put forward for the generation of multi-orbit cloud-free LAIG maps with an unprecedented level of detail, and the extraction of regularly-sampled LAIG time series at a pixel level. The ability to plugin a locally-trained GPR model into the GEE framework and its instant processing opens up a new paradigm of remote sensing image processing.

show abstract

Section: Gapfilling Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Green LAI Mapping and Cloud Gap-Filling Using Gaussian Process Regression in Google Earth Engine

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…The wide applications of machine learning approaches open a new path to the development of satellite-based phenology models. Recently, Belda et al (2020) blended machine learning algorithms (e.g., Gaussian Process Regression, GPR) using leaf area index (LAI) curve reconstruction in satellitebased phenology methods, the new methods successfully reconstructed LAI curves and retrieved reliable phenological indicators. Leaf falling phase is a transient process, the curve fitting method with higher resolution can better capture the phenological information (Hermance, 2007).…”

Section: Discussionmentioning

confidence: 99%

“…Moreover, Alberton et al (2019) deployed Generalized Additive Mixed Models (GAMMs) to estimate the timing and length of growing season based on phenocameras networks observations. Recently, Belda et al (2020) applied advanced machine learning algorithms for VI curves reconstruction and estimated phenological indicators for specific crop types.…”

Section: Introductionmentioning

confidence: 99%

Regional evaluation of satellite‐based methods for identifying end of vegetation growing season

Shen

Yuan

et al. 2021

Remote Sens Ecol Conserv

View full text Add to dashboard Cite

Autumn phenology plays an important role in regulating ecosystem carbon and water cycling, but it has received less attention than spring phenology. Satellitebased methods have been widely applied in monitoring autumn phenology at large spatial scales. However, few studies have evaluated and compared the performance of different satellite-based methods in autumn phenology identification. Here, we compared the spatiotemporal variations of end of vegetation growing season dates (EOS) as determined from eight prevailing satellite-based methods against long-term field observations at 31 sites in China. We found that field-based observations in forest and grassland sites, respectively, had rates of EOS delay of 2.11 and 3.85 days per 1°C increase in mean annual temperature (MAT) during 2001-2014. However, nearly all the eight satellite-based methods underestimated these delay rates compared with the ground observations over all sites. We also found that the eight methods weakly agreed with the field-observed interannual variations of EOS. At the regional scale, the identified average EOS differed up to 38 and 40 days among the investigated satellite-based methods in forest and grassland ecosystems respectively. The delayed rate of identified EOS with the increase of MAT ranged from 0.77 to 3.51 days°C −1 for forests and from 0.41 to 2.95 days°C −1 for grasslands. The identified EOS by most of the eight methods had delayed temporal trends in forests during 2001-2014 while we found advanced trends in grassland ecosystems. The large discrepancy in EOS identification among the prevailing satellite-based methods highlight the need for more accurate satellite-based methods in data gap-filling and phenometrics detection, and more extensive, multi-species based field observations that can be used to constrain and validate the satellite-based methods.

show abstract

“…MVI has been carried out using statistical techniques such as simple means, Multiple Linear Regressions (MLR), Logistic Regressions (LR), Random Forest Decision Trees (RFD), or Bayesian inference [10,[23][24][25][26][27]. It now benefits from the most recent developments in Machine Learning techniques such as K-Nearest Neighbour (KNN), Support Vector Machines, Artificial Neural Networks, Long Short-Term Memory algorithms [20,[28][29][30][31], and more recently Graph Neural Networks [32]. The latter are particularly interesting for missing value imputation on urban water networks whose design rules follow topological relationships both for network configuration and geometric properties.…”

Section: Introductionmentioning

confidence: 99%

Graph Convolutional Networks: Application to Database Completion of Wastewater Networks

Bel-Ghaddar

Chahinian

Seriai³

et al. 2021

Water

View full text Add to dashboard Cite

Wastewater networks are mandatory for urbanisation. Their management, including the prediction and planning of repairs and expansion operations, requires precise information on their underground components (manhole covers, equipment, nodes, and pipes). However, due to their years of service and to the increasing number of maintenance operations they may have undergone over time, the attributes and characteristics associated with the various objects constituting a network are not all available at a given time. This is partly because (i) the multiple actors that carry out repairs and extensions are not necessarily the operators who ensure the continuous functioning of the network, and (ii) the undertaken changes are not properly tracked and reported. Therefore, databases related to wastewater networks may suffer from missing data. To overcome this problem, we aim to exploit the structure of wastewater networks in the learning process of machine learning approaches, using topology and the relationship between components, to complete the missing values of pipes. Our results show that Graph Convolutional Network (GCN) models yield better results than classical methods and represent a useful tool for missing data completion.

show abstract

DATimeS: A machine learning time series GUI toolbox for gap-filling and vegetation phenology trends detection

Cited by 60 publications

References 78 publications

Green LAI Mapping and Cloud Gap-Filling Using Gaussian Process Regression in Google Earth Engine

Green LAI Mapping and Cloud Gap-Filling Using Gaussian Process Regression in Google Earth Engine

Regional evaluation of satellite‐based methods for identifying end of vegetation growing season

Graph Convolutional Networks: Application to Database Completion of Wastewater Networks

Contact Info

Product

Resources

About