Calibration and evaluation of individual-based models using Approximate Bayesian Computation

This paper investigates the feasibility of using approximate Bayesian computation (ABC) to calibrate and evaluate complex individual-based models (IBMs). As ABC evolves, various versions are emerging, but here we only explore the most accessible version, rejection-ABC. Rejection-ABC involves running models a large number of times, with parameters drawn randomly from their prior distributions, and then retaining the simulations closest to the observations. Although well-established in some fields, whether ABC will work with ecological IBMs is still uncertain. Rejection-ABC was applied to an existing 14-parameter earthworm energy budget IBM for which the available data consist of body mass growth and cocoon production in four experiments. ABC was able to narrow the posterior distributions of seven parameters, estimating credible intervals for each. ABC’s accepted values produced slightly better fits than literature values do. The accuracy of the analysis was assessed using cross-validation and coverage, currently the best available tests. Of the seven unnarrowed parameters, ABC revealed that three were correlated with other parameters, while the remaining four were found to be not estimable given the data available. It is often desirable to compare models to see whether all component modules are necessary. Here we used ABC model selection to compare the full model with a simplified version which removed the earthworm’s movement and much of the energy budget. We are able to show that inclusion of the energy budget is necessary for a good fit to the data. We show how our methodology can inform future modelling cycles, and briefly discuss how more advanced versions of ABC may be applicable to IBMs. We conclude that ABC has the potential to represent uncertainty in model structure, parameters and predictions, and to embed the often complex process of optimizing an IBM’s structure and parameters within an established statistical framework, thereby making the process more transparent and objective.

Improving the TanDEM-X Digital Elevation Model for flood modelling using flood extents from Synthetic Aperture Radar images

The topography of many floodplains in the developed world has now been surveyed with high resolution sensors such as airborne LiDAR (Light Detection and Ranging), giving accurate Digital Elevation Models (DEMs) that facilitate accurate flood inundation modelling. This is not always the case for remote rivers in developing countries. However, the accuracy of DEMs produced for modelling studies on such rivers should be enhanced in the near future by the high resolution TanDEM-X WorldDEM. In a parallel development, increasing use is now being made of flood extents derived from high resolution Synthetic Aperture Radar (SAR) images for calibrating, validating and assimilating observations into flood inundation models in order to improve these. This paper discusses an additional use of SAR flood extents, namely to improve the accuracy of the TanDEM-X DEM in the floodplain covered by the flood extents, thereby permanently improving this DEM for future flood modelling and other studies. The method is based on the fact that for larger rivers the water elevation generally changes only slowly along a reach, so that the boundary of the flood extent (the waterline) can be regarded locally as a quasi-contour. As a result, heights of adjacent pixels along a small section of waterline can be regarded as samples with a common population mean. The height of the central pixel in the section can be replaced with the average of these heights, leading to a more accurate estimate. While this will result in a reduction in the height errors along a waterline, the waterline is a linear feature in a two-dimensional space. However, improvements to the DEM heights between adjacent pairs of waterlines can also be made, because DEM heights enclosed by the higher waterline of a pair must be at least no higher than the corrected heights along the higher waterline, whereas DEM heights not enclosed by the lower waterline must in general be no lower than the corrected heights along the lower waterline. In addition, DEM heights between the higher and lower waterlines can also be assigned smaller errors because of the reduced errors on the corrected waterline heights. The method was tested on a section of the TanDEM-X Intermediate DEM (IDEM) covering an 11km reach of the Warwickshire Avon, England. Flood extents from four COSMO-SKyMed images were available at various stages of a flood in November 2012, and a LiDAR DEM was available for validation. In the area covered by the flood extents, the original IDEM heights had a mean difference from the corresponding LiDAR heights of 0.5 m with a standard deviation of 2.0 m, while the corrected heights had a mean difference of 0.3 m with standard deviation 1.2 m. These figures show that significant reductions in IDEM height bias and error can be made using the method, with the corrected error being only 60% of the original. Even if only a single SAR image obtained near the peak of the flood was used, the corrected error was only 66% of the original. The method should also be capable of improving the final TanDEM-X DEM and other DEMs, and may also be of use with data from the SWOT (Surface Water and Ocean Topography) satellite.

Trends in adult body-mass index in 200 countries from 1975 to 2014: a pooled analysis of 1698 population-based measurement studies with 19·2 million participants

Background Underweight and severe and morbid obesity are associated with highly elevated risks of adverse health outcomes. We estimated trends in mean body-mass index (BMI), which characterises its population distribution, and in the prevalences of a complete set of BMI categories for adults in all countries. Methods We analysed, with use of a consistent protocol, population-based studies that had measured height and weight in adults aged 18 years and older. We applied a Bayesian hierarchical model to these data to estimate trends from 1975 to 2014 in mean BMI and in the prevalences of BMI categories (<18·5 kg/m2 [underweight], 18·5 kg/m2 to <20 kg/m2, 20 kg/m2 to <25 kg/m2, 25 kg/m2 to <30 kg/m2, 30 kg/m2 to <35 kg/m2, 35 kg/m2 to <40 kg/m2, ≥40 kg/m2 [morbid obesity]), by sex in 200 countries and territories, organised in 21 regions. We calculated the posterior probability of meeting the target of halting by 2025 the rise in obesity at its 2010 levels, if post-2000 trends continue. Findings We used 1698 population-based data sources, with more than 19·2 million adult participants (9·9 million men and 9·3 million women) in 186 of 200 countries for which estimates were made. Global age-standardised mean BMI increased from 21·7 kg/m2 (95% credible interval 21·3–22·1) in 1975 to 24·2 kg/m2 (24·0–24·4) in 2014 in men, and from 22·1 kg/m2 (21·7–22·5) in 1975 to 24·4 kg/m2 (24·2–24·6) in 2014 in women. Regional mean BMIs in 2014 for men ranged from 21·4 kg/m2 in central Africa and south Asia to 29·2 kg/m2 (28·6–29·8) in Polynesia and Micronesia; for women the range was from 21·8 kg/m2 (21·4–22·3) in south Asia to 32·2 kg/m2 (31·5–32·8) in Polynesia and Micronesia. Over these four decades, age-standardised global prevalence of underweight decreased from 13·8% (10·5–17·4) to 8·8% (7·4–10·3) in men and from 14·6% (11·6–17·9) to 9·7% (8·3–11·1) in women. South Asia had the highest prevalence of underweight in 2014, 23·4% (17·8–29·2) in men and 24·0% (18·9–29·3) in women. Age-standardised prevalence of obesity increased from 3·2% (2·4–4·1) in 1975 to 10·8% (9·7–12·0) in 2014 in men, and from 6·4% (5·1–7·8) to 14·9% (13·6–16·1) in women. 2·3% (2·0–2·7) of the world's men and 5·0% (4·4–5·6) of women were severely obese (ie, have BMI ≥35 kg/m2). Globally, prevalence of morbid obesity was 0·64% (0·46–0·86) in men and 1·6% (1·3–1·9) in women. Interpretation If post-2000 trends continue, the probability of meeting the global obesity target is virtually zero. Rather, if these trends continue, by 2025, global obesity prevalence will reach 18% in men and surpass 21% in women; severe obesity will surpass 6% in men and 9% in women. Nonetheless, underweight remains prevalent in the world's poorest regions, especially in south Asia.

Bayesian analysis of uncertainty in the GlobCover 2009 land cover product at climate model grid scale

Land cover data derived from satellites are commonly used to prescribe inputs to models of the land surface. Since such data inevitably contains errors, quantifying how uncertainties in the data affect a model’s output is important. To do so, a spatial distribution of possible land cover values is required to propagate through the model’s simulation. However, at large scales, such as those required for climate models, such spatial modelling can be difficult. Also, computer models often require land cover proportions at sites larger than the original map scale as inputs, and it is the uncertainty in these proportions that this article discusses. This paper describes a Monte Carlo sampling scheme that generates realisations of land cover proportions from the posterior distribution as implied by a Bayesian analysis that combines spatial information in the land cover map and its associated confusion matrix. The technique is computationally simple and has been applied previously to the Land Cover Map 2000 for the region of England and Wales. This article demonstrates the ability of the technique to scale up to large (global) satellite derived land cover maps and reports its application to the GlobCover 2009 data product. The results show that, in general, the GlobCover data possesses only small biases, with the largest belonging to non–vegetated surfaces. In vegetated surfaces, the most prominent area of uncertainty is Southern Africa, which represents a complex heterogeneous landscape. It is also clear from this study that greater resources need to be devoted to the construction of comprehensive confusion matrices.