Marko Laine

DRAM: Efficient adaptive MCMC

We propose to combine two quite powerful ideas that have recently appeared in the Markov chain Monte Carlo literature: adaptive Metropolis samplers and delayed rejection. The ergodicity of the resulting non-Markovian sampler is proved, and the efficiency of the combination is demonstrated with various examples. We present situations where the combination outperforms the original methods: adaptation clearly enhances efficiency of the delayed rejection algorithm in cases where good proposal distributions are not available. Similarly, delayed rejection provides a systematic remedy when the adaptation process has a slow start.

Large historical growth in global terrestrial gross primary production

Growth in terrestrial gross primary production (GPP)—the amount of carbon dioxide that is ‘fixed’ into organic material through the photosynthesis of land plants—may provide a negative feedback for climate change. It remains uncertain, however, to what extent biogeochemical processes can suppress global GPP growth. As a consequence, modelling estimates of terrestrial carbon storage, and of feedbacks between the carbon cycle and climate, remain poorly constrained. Here we present a global, measurement-based estimate of GPP growth during the twentieth century that is based on long-term atmospheric carbonyl sulfide (COS) records, derived from ice-core, firn and ambient air samples. We interpret these records using a model that simulates changes in COS concentration according to changes in its sources and sinks—including a large sink that is related to GPP. We find that the observation-based COS record is most consistent with simulations of climate and the carbon cycle that assume large GPP growth during the twentieth century (31% ± 5% growth; mean ± 95% confidence interval). Although this COS analysis does not directly constrain models of future GPP growth, it does provide a global-scale benchmark for historical carbon-cycle simulations.

Randomize-Then-Optimize: A Method for Sampling from Posterior Distributions in Nonlinear Inverse Problems

High-dimensional inverse problems present a challenge for Markov chain Monte Carlo (MCMC)-type sampling schemes. Typically, they rely on finding an efficient proposal distribution, which can be dif...

Reduced Models of Algae Growth

The simulation of biological systems is often plagued by a high level of noise in the data, as well as by models containing a large number of correlated parameters. As a result, the parameters are poorly identified by the data, and the reliability of the model predictions may be questionable. Bayesian sampling methods provide an avenue for proper statistical analysis in such situations. Nevertheless, simulations should employ models that, on the one hand, are reduced as much as possible, and, on the other hand, are still able to capture the essential features of the phenomena studied. Here, in the case of algae growth modeling, we show how a systematic model reduction can be done. The simplified model is analyzed from both theoretical and statistical points of view.

Changes in Rodent Abundance and Weather Conditions Potentially Drive Hemorrhagic Fever with Renal Syndrome Outbreaks in Xi’an, China, 2005–2012

Background Increased risks for hemorrhagic fever with renal syndrome (HFRS) caused by Hantaan virus have been observed since 2005, in Xi’an, China. Despite increased vigilance and preparedness, HFRS outbreaks in 2010, 2011, and 2012 were larger than ever, with a total of 3,938 confirmed HFRS cases and 88 deaths in 2010 and 2011. Methods and Findings Data on HFRS cases and weather were collected monthly from 2005 to 2012, along with active rodent monitoring. Wavelet analyses were performed to assess the temporal relationship between HFRS incidence, rodent density and climatic factors over the study period. Results showed that HFRS cases correlated to rodent density, rainfall, and temperature with 2, 3 and 4-month lags, respectively. Using a Bayesian time-series Poisson adjusted model, we fitted the HFRS outbreaks among humans for risk assessment in Xi’an. The best models included seasonality, autocorrelation, rodent density 2 months previously, and rainfall 2 to 3 months previously. Our models well reflected the epidemic characteristics by one step ahead prediction, out-of-sample. Conclusions In addition to a strong seasonal pattern, HFRS incidence was correlated with rodent density and rainfall, indicating that they potentially drive the HFRS outbreaks. Future work should aim to determine the mechanism underlying the seasonal pattern and autocorrelation. However, this model can be useful in risk management to provide early warning of potential outbreaks of this disease.

Simulation-Based Optimal Design Using a Response Variance Criterion

Classical optimal design criteria suffer from two major flaws when applied to nonlinear problems. First, they are based on linearizing the model around a point estimate of the unknown parameter and therefore depend on the uncertain value of that parameter. Second, classical design methods are unavailable in ill-posed estimation situations, where previous data lack the information needed to properly construct the design criteria. Bayesian optimal design can, in principle, solve these problems. However, Bayesian design methods are not widely applied, mainly due to the fact that standard implementations for efficient and robust routine use are not available. In this article, we point out a concrete recipe for implementing Bayesian optimal design, based on the concept of simulation-based design introduced by Muller, Sanso, and De Iorio (2004). We develop further a predictive variance criterion and introduce an importance weighting mechanism for efficient computation of the variances. The simulation-based approach allows one to start the model-based optimization of experiments at an early stage of the parameter estimation process, in situations where the classical design criteria are not available. We demonstrate that the approach can significantly reduce the number of experiments needed to obtain a desired level of accuracy in the parameter estimates. A computer code package that implements the approach in a simple case is provided as supplemental material (available online).

Anthropogenically driven environmental changes shift the ecological dynamics of hemorrhagic fever with renal syndrome

Zoonoses are increasingly recognized as an important burden on global public health in the 21st century. High-resolution, long-term field studies are critical for assessing both the baseline and future risk scenarios in a world of rapid changes. We have used a three-decade-long field study on hantavirus, a rodent-borne zoonotic pathogen distributed worldwide, coupled with epidemiological data from an endemic area of China, and show that the shift in the ecological dynamics of Hantaan virus was closely linked to environmental fluctuations at the human-wildlife interface. We reveal that environmental forcing, especially rainfall and resource availability, exert important cascading effects on intra-annual variability in the wildlife reservoir dynamics, leading to epidemics that shift between stable and chaotic regimes. Our models demonstrate that bimodal seasonal epidemics result from a powerful seasonality in transmission, generated from interlocking cycles of agricultural phenology and rodent behavior driven by the rainy seasons.

Interannual cycles of Hantaan virus outbreaks at the human–animal interface in Central China are controlled by temperature and rainfall

Significance Interannual cycles of many zoonotic diseases are considered to be driven by climate variability. However, the role of climate forcing in the modulation of zoonotic dynamics has been highly controversial, chiefly due to the difficulty in quantifying the links between climate forcing, animal population dynamics, and disease dynamics. Here, we address this issue by using a unique field surveillance dataset from Central China, covering one-half century. Our results shed light on the drivers behind interannual variability and the dynamic patterns of disease ecology, and the links between interannual climate variability and the human–animal interface, adding up to 3-mo lead time over outbreak warnings. Hantavirus, a rodent-borne zoonotic pathogen, has a global distribution with 200,000 human infections diagnosed annually. In recent decades, repeated outbreaks of hantavirus infections have been reported in Eurasia and America. These outbreaks have led to public concern and an interest in understanding the underlying biological mechanisms. Here, we propose a climate–animal–Hantaan virus (HTNV) infection model to address this issue, using a unique dataset spanning a 54-y period (1960–2013). This dataset comes from Central China, a focal point for natural HTNV infection, and includes both field surveillance and an epidemiological record. We reveal that the 8-y cycle of HTNV outbreaks is driven by the confluence of the cyclic dynamics of striped field mouse (Apodemus agrarius) populations and climate variability, at both seasonal and interannual cycles. Two climatic variables play key roles in the ecology of the HTNV system: temperature and rainfall. These variables account for the dynamics in the host reservoir system and markedly affect both the rate of transmission and the potential risk of outbreaks. Our results suggest that outbreaks of HTNV infection occur only when climatic conditions are favorable for both rodent population growth and virus transmission. These findings improve our understanding of how climate drives the periodic reemergence of zoonotic disease outbreaks over long timescales.

A dilemma of the uniqueness of weather and climate model closure parameters

Parameterisation schemes of subgrid-scale physical processes in atmospheric models contain so-called closure parameters. Their precise values are not generally known; thus, they are subject to fine-tuning for achieving optimal model performance. In this article, we show that there is a dilemma concerning the optimal parameter values: an identical prediction model formulation can have two different optimal closure parameter value settings depending on the level of approximations made in the data assimilation component of the prediction system. This result tends to indicate that the prediction model re-tuning in large-scale systems is not only needed when the prediction model undergoes a major change, but also when the data assimilation component is updated. Moreover, we advocate an accurate albeit expensive method based on so-called filter likelihood for the closure parameter estimation that is applicable in fine-tuning of both prediction model and data assimilation system parameters. In this article, we use a modified Lorenz-95 system as a prediction model and extended Kalman filter and ensemble adjustment Kalman filter for data assimilation. With this setup, we can compute the filter likelihood for the chosen parameters using the output of the two versions of the Kalman filter and apply a Markov chain Monte Carlo algorithm to explore the parameter posterior distributions.

Bayesian principal component regression model with spatial effects for forest inventory variables under small field sample size

Abstract Remote sensing observations are extensively used for analysis of environmental variables. These variables often exhibit spatial correlation, which has to be accounted for in the calibration models used in predictions, either by direct modelling of the dependencies or by allowing for spatially correlated stochastic effects. Another feature in many remote sensing instruments is that the derived predictor variables are highly correlated, which can lead to unnecessary model over-training and at worst, singularities in the estimates. Both of these affect the prediction accuracy, especially when the training set for model calibration is small. To overcome these modelling challenges, we present a general model calibration procedure for remotely sensed data and apply it to airborne laser scanning data for forest inventory. We use a linear regression model that accounts for multicollinearity in the predictors by principal components and Bayesian regularization. It has a spatial random effect component for the spatial correlations that are not explained by a simple linear model. An efficient Markov chain Monte Carlo sampling scheme is used to account for the uncertainty in all the model parameters. We tested the proposed model against several alternatives and it outperformed the other linear calibration models, especially when there were spatial effects, multicollinearity and the training set size was small.