Soil Moisture Estimation Using Sentinel-1/-2 Imagery Coupled With CycleGAN for Time-Series Gap Filing

Abstract

Fast soil moisture content (SMC) mapping is necessary to support water resource management and to understand crops' growth, quality and yield. Thereby, Earth Observation (EO) plays a key role due to its ability of almost real-time monitoring of large areas at a low cost. This study aimed to explore the possibility of taking advantage of freely available Sentinel-1 (S1) and Sentinel-2 (S2) EO data for the simultaneous prediction of SMC with cycle-consistent adversarial network (cycleGAN) for time-series gap filling. The proposed methodology, first, learns latent low-dimensional representation of the satellite images, then learns a simple machine learning model on top of these representations. To evaluate the methodology, a series of vineyards, located in South Australia's Eden valley are chosen. Specifically, we presented an efficient framework for extracting latent features from S1 and S2 imagery. We showed how one could use S1 to S2 feature translation based on Cycle-GAN using S1&S2 time series when there are missing images acquired over an area of interest. The resulting data in our study is then used to fill gaps in time series data. We used the resulting latent representations to predict SMC with various ML tools. In the experiments, cycleGAN and the autoencoders were trained with data randomly chosen around the site of interest, so we could augment the existing dataset. The best performance was demonstrated with random forest algorithm, whereas linear regression model demonstrated significant overfitting. The experiments demonstrate that the proposed methodology outperforms the compared state-of-the-art methods if there are missing optical and synthetic-aperture radar (SAR) images.

Número Completo (Mayo-junio)

“…Por otra parte, los modelos predictivos en la agricultura se han tratado en conjunto, según evidencia el segundo mayor clúster, se destaca las investigaciones relacionadas con la toma de decisiones, aprendizaje automático, árboles de decisión, sistema de aprendizajes, vectores de soporte automáticos, monitoreo, agricultura de precisión, adquisición de datos, meteorología, aprendizaje profundo, redes neuronales y pronóstico, es importante destacar que son métodos que incluyen la variable económico-financiera como principal factor (Caicedo Solano et al, 2022;Efremova et al, 2022;Jamei et al, 2022;Kassem et al, 2022;Kussul et al, 2022).…”

SAR2NDVI: a pix2pix generative adversarial network for reconstructing field-level normalized difference vegetation index time series using Sentinel-1 synthetic aperture radar data

“…Pix2pix GAN has been used to generate near-infrared band from RGB images acquired by unmanned aerial vehicle much more efficiently than the endmembers method 33 . Time series soil moisture content was estimated using Sentinel-1 images using the CycleGAN model 25 . Similarly, GANs are also used for cloud removal in optical images 34 …”

“…33 Time series soil moisture content was estimated using Sentinel-1 images using the CycleGAN model. 25 Similarly, GANs are also used for cloud removal in optical images. 34 Cloud cover is the major issue in vegetation monitoring using optical data.…”

“…With abundant open-access satellite data for training, deep learning 23 – 25 tends to give better results than conventional methods. In this paper, we propose a deep learning-based framework for generating NDVI from Sentinel-1 images using pix2pix generative adversarial network (GAN) architecture.…”

Sentinel-1 Backscatter and Interferometric Coherence for Soil Moisture Retrieval in Winter Wheat Fields Within a Semiarid South-Mediterranean Climate: Machine Learning Versus Semiempirical Models

“…However, a severe lack of optical data observations may occur in the event of prolonged cloud cover, which is quite frequent when crops, particularly cereals, are cultivated during the wet season, as in the South Mediterranean regions. To deal with this issue, [44] have proposed framework for unsupervised deep domain adaptation for radar and optical satellite imagery with cycle-consistent adversarial network (cycleGANs). In the same vein, using a large in-situ database collected from several irrigated and rainfed wheat in Morocco and Tunisia countries, [45] have developed a new approach to predict SSM, based only on two complementary and relatively independent information extracted from Sentinel-1 radar including backscattering coefficient and interferometric coherence (ρ).…”