Search

Purpose.: Augmented Reality (AR) in Laparoscopic Liver Resection requires anatomical landmarks and the silhouette to be found on the laparoscopic image. They are used to register the preoperative 3D model obtained from CT segmentation. The existing AR systems rely on the surgeon to 1) annotate the landmarks and silhouette and 2) provide an initial registration. These non-trivial tasks require surgeon attention which may perturb the procedure. We propose methods to solve both tasks, hence registration, automatically. Methods.: The landmarks are the lower ridge and the falciform ligament. We solve 1) by training a U-Net from a new dataset of 1415 labelled images extracted from 68 procedures. We solve 2) by a novel automatic coarse-to-fine pose estimation method, including visibility-reasoning within an iterative robust process. In addition, we propose to divide the ridge into six anatomical sub-parts, making its annotation and use in registration more accurate. Results.: Our method detects the silhouette with an error equivalent to an experienced surgeon. It detects the ridge and ligament with higher errors owing to under-detection. Nonetheless, our method successfully initialises the registration with tumour target registration errors of 22.4, 14.8 and 7.2 mm for 3 clinical procedures. In comparison, the errors from manual initialisation are 30.5, 15.1 and 16.3 mm. Conclusion.: Our results are promising, suggesting that we have found an appropriate methodological approach. [ABSTRACT FROM AUTHOR]

Carbon (C) emissions from anthropogenic land use have accelerated climate change. To reduce C emissions, dynamic models can be used to assess the impact of human drivers on terrestrial C sequestration. Model accuracy requires correct initialisation, since incorrect initialisation can influence the results obtained. Therefore, we sought to improve the initialisation of a process-based SOC model, RothC, which can estimate the effect of climate and land-use change on SOC. The most common initialisation involves running the model until equilibrium ('spin-up run'), when the SOC pools stabilise (method 1). However, this method does not always produce realistic results. At our experimental sites, the observed SOC was not at equilibrium after 10 years, suggesting that the commonly used spin-up initialisation method assuming equilibrium might be improved. In addition to method 1, we tested two alternative initialisations for RothC that involved adjusting the total or individual SOC pool equilibrium values by regulating the C input during the entire spin-up initialisation period (method 2) and initialising each SOC pool with recently measured SOC values obtained by SOC fractionation (method 3). Analysis of the simulation accuracy for each model initialisation, quantified using the root mean square error (RMSE), indicated that a variant of method 2 that involved adjusting the equilibrium total SOC to observed values (method 2-T) generally showed less variation in the individual SOC pools and total SOC. Furthermore, as total SOC is the sum of all SOC pools, and because total SOC data are more readily available than the individual SOC pool data, we conclude that method 2-T is best for initialising RothC. [ABSTRACT FROM AUTHOR]

Mechanistic modelling is an important tool for understanding the impacts of climate change and pollutants on forest growth. One of the common practical limitations of these models is a lack of specific information regarding management activities such as thinning or harvesting, which can have a very strong influence on the accuracy of results. The use of inventory data for model parameterization and calibration is also problematic, as inventories are designed to have large volumes of data amalgamated to give accurate mean results across large areas. The precision of single point estimates is often quite low.This study uses BIOME-BGC to model forest growth on 1133 sites of the Austrian National Forest Inventory, and develops a method to estimate timber removal patterns prior to the commencement of record keeping on the sites. Recognizing the poor precision of individual point estimates in the data, we do not seek to precisely calibrate the model to the data on each point. Rather, we assume that the point-wise inventory estimates will be normally distributed around the true values. We then model each site assuming no management interventions, and compare this with inventory results. Plotting the “error” between model results and NFI data shows a strong right-skew, reflecting the modelled lack of timber removals. A Box-Cox transformation of the error plot, centred on zero, would represent an unbiased model estimate of the data, thus we can determine the historic timber removals as the difference between the original error curve and its Box-Cox transformation. Calibrating the model with this information allow us to represent forest volume with greater accuracy than would otherwise be possible.

Highlights: [•] We derive a method for the spin-up of first-order-based soil organic matter models. [•] The method explicitly relates soil organic matter pools with input data. [•] The method is probably the most efficient method to solve the spin-up problem. [Copyright &y& Elsevier]

Abstract: The study analyses the added value of data assimilation for short-term air quality forecasting by means of three modelling experiments with sulphur oxides. Two ways of utilising the observations are considered: determination of the optimal model initial state and adjustment of the emission fluxes of atmospheric pollutants. It is demonstrated that the influence of the initial conditions on the predicted SOx concentrations disappears within less than a day in European-scale applications. Adjusting the emission fluxes had a longer lasting impact on the model results, frequently covering the whole forecast window. The two-week long data assimilation exercise for Southern Europe showed that the largest improvement of the model score with regard to individual monitoring sites was obtained for the stations with the worst initial model-measurement agreement. With the emission adjustment, a major improvement was achieved for the stations near the Etna volcano, the strongest source in the area, where the SO2 emission was reduced by almost 50% as a result of the data assimilation. [ABSTRACT FROM AUTHOR]

Daisy is a semi-mechanistic model that simulates crop production, soil water and C and N dynamics in agro-ecosystems. In this study, the soil organic matter (SOM) submodel of Daisy was calibrated and validated with data from a series of long-term (30–100 yr) experiments dominated by cereal cropping systems. The experiments were conducted under temperate conditions at the Askov Experimental Station, Denmark. The recalibration of the Daisy model improved the model simulations considerably. The RMSE between simulations and observations of the experiments used for calibration improved from 33 to 19, whereas this measure improved from 32 to 9.0 for the experiments used for validation. The original calibration of the model appears to be based upon too low an input of organic matter (especially because rhizodeposition is ignored), too low an efficiency with which added organic matter is converted into more stable forms of SOM, and too low an overall decomposition rate. Our study suggests that the fraction of SOM involved in medium-term turnover is substantially larger than previously thought and inferred by most SOM turnover models. This may warrant a recalibration of current simulation models, and stresses the importance of using long-term experiments with widely differing treatments for model evaluation. Long-term treatments with widely contrasting initial C contents and annual C inputs, especially bare fallow treatments, appear to be valuable resources in this respect. [Copyright &y& Elsevier]

Like most other contemporary models of long-term soil organic matter (SOM) dynamics, Daisy partitions the refractory component of the SOM into two or more pools. In long-term simulations, the initial distribution of SOM between these two pools influences the simulations. However, as these pools do not correspond to any measurable entities the distribution of SOM cannot be initialised by a simple measurement. Daisy has usually been initialised by using a standard distribution, which almost corresponds to equilibrium. In applications of other models, the initial distribution between the different pools of organic matter has been chosen by assuming equilibrium at the beginning of the simulation, by using standard values or by calibrating the initial distribution to optimise simulation performance. However, as we show here the initial distribution of SOM between the different pools influences simulations notably, and the appropriate distribution is dependent on changes in management and climate at the site before the onset of a simulated experiment. This is done by simulating a scenario and initialising the Daisy model in to different ways: (1) by assuming equilibrium at the beginning of the simulation; and (2) by simulating the preexperimental management history of the site. This shows that it can be important to use plausible assumptions to initialise SOM models, and that often the only way to initialise the model is to simulate the preexperimental period of the site. [Copyright &y& Elsevier]

The transport and diffusion processes of a tracer gas released near the ground in the Rhine valley region, in Central Europe, during the 1992 TRACT field experiment, are simulated by a computational model system for complex terrain. This system (RMS) is composed of the prognostic mesoscale model RAMS, the Lagrangian stochastic dispersion model SPRAY and the interface code MIRS, which links RAMS to SPRAY. Three flow simulations were performed, with different initialisations and the one showing the best agreement with the measured flow was selected for the simulation of the TRACT tracer experiment. Tracer concentrations measured by an array of samplers at ground level and by an airplane aloft, are used to evaluate the 3-D concentration field simulated by the model system. The analysis of the simulation results generated by RMS shows that our model system very well reproduces the general behaviour of the contaminant plume, the temporal and spatial distribution of the concentration and the location of the concentration maxima. [Copyright &y& Elsevier]