Nonparametric statistical downscaling for the fusion of data of different spatiotemporal support

Abstract

Statistical downscaling has been developed for the fusion of data of different spatial support. However, environmental data often have different temporal support, which must also be accounted for. This paper presents a novel method of nonparametric statistical downscaling, which enables the fusion of data of different spatiotemporal support through treating the data at each location as observations of smooth functions over time. This is incorporated within a Bayesian hierarchical model with smoothly spatially varying coefficients, which provides predictions at any location or time, with associated estimates of uncertainty. The method is motivated by an application for the fusion of in situ and satellite remote sensing log(chlorophyll‐a) data from Lake Balaton, in order to improve the understanding of water quality patterns over space and time.

The Estimation of Water Quality Parameters in Lentic Environments Through Remote Sensing Technologies: a Review of the Past Two Decades

“…The increase concentration of these species is a major cause of water contamination (PEPPA; VASILAKOS; KAVROUDAKIS, 2020). High nutrient loads (nitrogen and phosphorus) accelerate the growth and biomass production of algae (PYO et al, 2018) and can be associated with phytoplankton blooms (WILKIE et al, 2019). There is real concern about cyanotoxins blooming in water bodies as these toxins pose an effective threat to public health.…”

“…Remote sensing applied to WQ has been investigated by a variety of scientific communities such as hydrology (POTES; COSTA; SALGADO, 2012;CURTARELLI et al, 2014;BONANSEA et al, 2015;HANSEN;WILLIAMS, 2018), hydrobiology (ZHANG et al, 2016;BRESCIANI et al, 2018), public health (TORBICK et al, 2014; VAN DER MERWE;PRICE, 2015;TORBICK et al, 2018), urban planning (HUO et al, 2014;LIU et al, 2015) and applied statistics (WILKIE et al, 2019;, amongst others. Whatever the application or the purpose, most authors have brought forward the relative ease and low cost of remote sensing methods over traditional in situ methods (BONANSEA et al, 2015;ABDELMALIK, 2018;AVDAN et al, 2019).…”

“…Whatever the application or the purpose, most authors have brought forward the relative ease and low cost of remote sensing methods over traditional in situ methods (BONANSEA et al, 2015;ABDELMALIK, 2018;AVDAN et al, 2019). Furthermore, it has been reported that these traditional approaches often use a relatively small number of samples and at an insufficient frequency (WILKIE et al, 2019). Within that context, remote sensing can bring a comprehensive distributed estimate of some important WQ parameters (STEFOULI; CHAROU, 2012) even considering some important limitations like the restriction to the surface layer and several other uncontrolled factors affecting the optical properties of water like the atmospheric conditions at the moment of acquisition and the interference of the air-water interface (for a review, see Mertes et al (2004)).…”

Connected Sensors, Innovative Sensor Deployment, and Intelligent Data Analysis for Online Water Quality Monitoring

“…Dealing with the data volume as well as the different data streams have also presented challenges [186]. In this space, there are new developments concerning methods to fuse and assimilate different data streams [187][188][189]. By harnessing the power of AI algorithms and big data analytics, water utilities can maximize information and data available to make better decisions while enhancing service delivery and reducing costs [190].…”

Data fusion with Gaussian processes for estimation of environmental hazard events

“…Boaz, Lawson, and Pearce (2019) develop a multivariate fusion framework to deal with air pollution prediction with partial missingness. Wilkie et al (2019) treat data for fusion as realisations of smooth temporal functions, Gilani, Berrocal, and Batterman (2019) accounts for nonstationarity by incorporating covariates, postulated to drive the nonstationary behavior, in the covariance function and Ma and Kang (2020) develop a stochastic expectation–maximization that facilitates the use of large spatiotemporal datasets. Forlani, Bhatt, Cameletti, Krainski, and Blangiardo (2020) demonstrate the added benefit of allowing multiple sources of model output in a framework that combines stochastic partial differential equations and the integrated nested Laplace approximation (Lindgren, Rue & Lindström, 2011).…”