Abstract:<p>Stochastic models are frequently applied to generate synthetic time series with unlimited length. Most of the traditional stochastic models are single-site based and not able to reconstruct the inter-site and inter-variable correlation structures, while most of the existing multi-variate multi-site stochastic models (MMSM) failed to reproduce the low frequency attributes. Some MMSMs tried to fix this problem through shuffling methods, which in turn leads to lack of stochasticity. This study proposes a new MMSM based on multi-scale quantile mapping and empirical copula, which is able to reconstruct the low-frequency attributes and inter-variable correlations and can provide with stochasticity at the meantime. An indicator named 'stochasticity error' is proposed to test whether the new MMSM is able to reproduce natural stochasticity. An application of the new MMSM is presented in the Lower Yellow River basin to generate stochastic time series of runoff and sediment transport rate. The performance is evaluated in comparison with two other MMSMs based on schaake shuffle and empirical copula. Results show that the proposed MMSM successfully reproduces the marginal distributional attributes, inter-variable and inter-site correlations, as well as the low frequency properties of sediment, but exhibits limited improvement in reconstructing the low frequency properties of runoff. On the other hand, the proposed MMSM shows acceptable performance at reproducing natural stochasticity, as the other two MMSMs are completely incapable of preserving natural stochasticity. In general, the proposed MMSM is capable of reconstructing inter-variable, inter-site correlations and low frequency attributes, as well as natural stochasticity.</p>

Wasserstein multivariate auto-regressive models for modeling distributional time series

Wasserstein autoregressive models for density time series

Autoregressive Optimal Transport Models

First-order multivariate integer-valued autoregressive model with multivariate mixture distributions

Sliced Wasserstein Regression

A Vector Autoregressive Moving Average Model for Interval-Valued Time Series Data

On distributional autoregression and iterated transportation

Sparse Interval-valued Time Series Modeling with Machine Learning

Negative Binomial Community Network Vector Autoregression for Multivariate Integer-Valued Time Series

Projected Statistical Methods for Distributional Data on the Real Line with the Wasserstein Metric

Statistical analysis of multivariate discrete-valued time series

A latent space model for multivariate count data time series analysis

A New Bivariate Random Coefficient INAR(1) Model with Applications

Periodically stationary multivariate autoregressive models

A Markov-switching Hidden Heterogeneous Network Autoregressive Model for Multivariate Time Series Data with Multimodality

Regularized Estimation in High-Dimensional Vector Auto-Regressive Models using Spatio-Temporal Information

Multivariate Distributions in Non-Stationary Complex Systems I: Random Matrix Model and Formulae for Data Analysis

Bivariate Random Coefficient Integer-Valued Autoregressive Model Based on a ρ-Thinning Operator

On Consistency and Sparsity for High-Dimensional Functional Time Series with Application to Autoregressions

A new multi-site multi-variable stochastic model with inter-site and inter-variable correlations, low frequency attributes and stochasticity: A case study in the lower Yellow River basin

Theory and inference for multivariate autoregressive binary models with an application to absence-presence data in ecology