Validation Data Subsets Research Articles

Development of a machine learning model for systematics of Aspergillus section Nigri using synchrotron radiation-based fourier transform infrared spectroscopy

Aspergillus section Nigri (black aspergilli) fungi are economically important food spoilage agents. Some species in this section also produce harmful mycotoxins in food. However, it is remarkably difficult to identify this fungal group at the species level using morphological and chemical characteristics. The molecular approach for classification is preferable; however, it is time-consuming, making it inappropriate for rapid testing of large numbers of samples. To address this, we explored synchrotron radiation-based Fourier transform infrared microspectroscopy (SR-FTIR) as a rapid method for obtaining data suitable for species classification. SR-FTIR data were obtained from the mycelia/conidia of 22 black aspergilli species. The Convolutional Neural Network (CNN) approach, a supervised deep learning algorithm, was used with SR-FTIR data to classify black aspergilli at the species level. A subset of the data was used to train the CNN model, and the model classification performance was evaluated using the validation data subsets. The model demonstrated a 95.97% accuracy in species classification on the testing (blind) data subset. The technique presented herein could be an alternative method for identifying problematic black aspergilli in the food industry.

Development of Spatiotemporal Whole-Stem Models for Estimating End-Product-Based Fibre Attribute Determinates for Jack Pine and Red Pine

The objective of this study was to develop spatiotemporal whole-stem wood quality prediction models for a suite of end-product-based fibre attribute determinates for jack pine (Pinus banksiana Lamb.) and red pine (Pinus resinosa Aiton): specifically, for wood density (Wd), microfibril angle (Ma), modulus of elasticity (Me), fibre coarseness (Co), tracheid wall thickness (Wt), tracheid radial diameter (Dr), tracheid tangential diameter (Dt), and specific surface area (Sa). Procedurally, these attributes were determined for each annual ring within pith-to-bark xylem sequences extracted from 610 jack pine and 223 red pine cross-sectional disks positioned throughout the main stem of 61 jack pine and 54 red pine sample trees growing within even-aged monospecific stands in central Canada. Deploying a block cross-validation-like approach in order to reduce serial data dependency and enable predictive performance assessments, species-specific calibration and validation data subsets consisting of cumulative moving average values were systematically generated from the 27,820 jack pine and 11,291 red pine attribute-specific annual ring values. Graphical, correlation, regression and validation analyses were used to specify, parameterize and assess the predictive performance of tertiary-level (ring-disk-tree) hierarchical mixed-effects whole-stem equations for each attribute by species. As a result, the jack pine equations explained 46, 66, 74, 63, 59, 72, 42 and 48% of the variation in Wd, Ma, Me, Co, Wt, Dr, Dt and Sa, respectively. The red pine equations explained slightly higher levels of variation except for Me: 50, 71, 31, 83, 72, 78, 56 and 71% of the variation in Wd, Ma, Me, Co, Wt, Dr, Dt and Sa, respectively. Graphical assessments and statistical metrics related to attribute and species-specific residual error patterns and goodness-of-fit, lack-of-fit and predictive error metrics, revealed an absence of systematic bias, misspecification or aberrant predictive performance. Consequently, the resultant parameterized models were acknowledged as acceptable functional descriptors of the intrinsic spatiotemporal cumulative developmental patterns of the studied end-product fibre attribute determinates, for these two pine species. Although predicted development patterns were similar between the species with the greatest degree of nonlinearity occurring before a cambial age of approximately 30 years, irrespective of attribute, jack pine exhibited a greater degree of nonlinearity in the Wd and Dt developmental trajectories, whereas red pine exhibited a greater degree of nonlinearity in the Ma, Me, Co, Wt, Dr and Sa developmental trajectories. Potential biomechanical linkages underlying the observed attribute distribution patterns, as well as the potential utility of the models in forest management, are also discussed.

Predicting uniaxial compressive strength from drilling variables aided by hybrid machine learning

Awareness of uniaxial compressive strength (UCS) as a key rock formation parameter for the design and development of gas and oil field plays. It plays an essential role in the selection of the drill bits and stability of the wellbore’s wall. Precise prediction of UCS before or during the drilling, especially in exploration wellbores, is necessary to improve the drilling speed and reduce the instability of the wellbore walls. UCS predictor machine-learning (ML) models are developed in this study using drilling parameters recorded during drilling using least-squares support-vector machine (LSSVM) and multi-layer extreme learning machine (MELM) algorithms hybridized with cuckoo optimization algorithm (COA), particle swarm optimization (PSO) and genetic algorithm (GA) optimizers. In addition, stand-alone LSSVM and convolutional neural network (CNN) models without optimizer enhancements are evaluated. Drilling and petrophysical data recorded for two wells (A and B) from the Rag-e-Safid oil field in southwest Iran were compiled to form the studied dataset. UCS was initially calculated numerically based on data from laboratory tests from petrophysical logs. The Well A dataset was pre-processed to remove outlying data records by applying the quantile regression algorithm. That analysis indicated that 9 data records should be removed from the Well A dataset. A decision tree model was employed for feature selection purposes to identify the more influential variables with respect to UCS. Depth, weight on the drill bit (WOB), drill-string rotation speed (RPM), rate of penetration (ROP), and torque (Trq) were the variables identified as being highly influential on UCS values. Application of the ML models on the training data subset (75% of Well A data records) revealed that the MELM-COA algorithm achieved the lowest root mean squared error (4.6945 MPa) and a higher coefficient of determination (0.9873) value than the other models when predicting UCS in the Well A training and validation data subsets. The Well-A-trained MELM-COA model confirmed its generalizability within the studied field by generating low UCS prediction errors when applied to the independent Well B testing dataset.

Metabolome-Based Classification of Snake Venoms by Bioinformatic Tools.

Snakebite is considered a neglected tropical disease, and it is one of the most intricate ones. The variability found in snake venom is what makes it immensely complex to study. These variations are present both in the big and the small molecules found in snake venom. This study focused on examining the variability found in the venom's small molecules (i.e., mass range of 100-1000 Da) between two main families of venomous snakes-Elapidae and Viperidae-managing to create a model able to classify unknown samples by means of specific features, which can be extracted from their LC-MS data and output in a comprehensive list. The developed model also allowed further insight into the composition of snake venom by highlighting the most relevant metabolites of each group by clustering similarly composed venoms. The model was created by means of support vector machines and used 20 features, which were merged into 10 principal components. All samples from the first and second validation data subsets were correctly classified. Biological hypotheses relevant to the variation regarding the metabolites that were identified are also given.

Stable downward continuation of the gravity potential field implemented using deep learning

Downward continuation (DC) of the gravity potential field is an important approach used to understand and interpret the density structure and boundary of anomalous bodies. It is widely used to delineate and highlight local and shallow anomalous sources. However, it is well known that direct DC transformation in the frequency domain is unstable and easily affected by high-frequency noise. Recent deep learning applications have led to the development of image recognition and resolution enhancement using the convolutional neural network technique. A similar deep learning architecture is also suitable for training a model for the DC problem. In this study, to solve the problems in existing DC methods, we constructed a dedicated model called DC-Net for the DC problem. We fully trained the DC-Net model on 38,400 pairs of gravity anomaly data at different altitudes using a convolutional neural network. We conducted several experiments and implemented a real-world example. The results demonstrate the following. First, several validation data subset and test data prediction results indicate that the DC-Net model was sufficiently trained. Moreover, it performed better than the traditional strategy in refining the upscaling of low-resolution images. Second, we performed tests on test datasets with changing levels of noise and demonstrated that the DC-Net model is noise-resistant and robust. Finally, we used the proposed model in a real-world example, which demonstrates that the DC-Net model is suitable for solving the DC problem and delineating the detailed gravity anomaly feature near the field source. For real data processing, noise in the gravity anomaly should be reduced in advance. Additionally, we recommend noise quantification of the gravity anomaly before network training.

Machine Learning Strategies for the Retrieval of Leaf-Chlorophyll Dynamics: Model Choice, Sequential Versus Retraining Learning, and Hyperspectral Predictors.

Monitoring leaf Chlorophyll (Chl) in-situ is labor-intensive, limiting representative sampling for detailed mapping of Chl variability at field scales across time. Unmanned aeria-l vehicles (UAV) and hyperspectral cameras provide flexible platforms for observing agricultural systems, overcoming this spatio-temporal sampling constraint. Here, we evaluate a customized machine learning (ML) workflow to retrieve multi-temporal leaf-Chl levels, combining sub-centimeter resolution UAV-hyperspectral imagery (400–1,000 nm) with leaf-level reflectance spectra and SPAD measurements, capturing temporal correlations, selecting relevant predictors, and retrieving accurate results under different conditions. The study is performed within a phenotyping experiment to monitor wild tomato plants’ development. Several analyses were conducted to evaluate multiple ML strategies, including: (1) exploring sequential versus retraining learning; (2) comparing insights gained from using 272 spectral bands versus 60 pigment-based vegetation indices (VIs); and (3) assessing six regression methods (linear, partial-least-square regression; PLSR, decision trees, support vector, ensemble trees, and Gaussian process; GPR). Goodness-of-fit (R2) and accuracy metrics (MAE, RMSE) were determined using training/testing and validation data subsets to assess the models’ performance. Overall, while equally good performance was obtained using either PLSR, GPR, or random forest, results show: (1) the retraining strategy improved the ability of most of the approaches to model SPAD-based Chl dynamics; (2) comparative analysis between retrievals and validation data distributions informed the models’ ability to capture Chl dynamics through SPAD levels; (3) VI predictors slightly improved R2 (e.g., from 0.59 to 0.74 units for GPR) and accuracy (e.g., MAE and RMSE differences of up to 2 SPAD units) in specific algorithms; (4) feature importance examined through these methods, revealed strong overlaps between relevant bands and VI predictors, highlighting a few decisive spectral ranges and indices useful for retrieving leaf-Chl levels. The proposed ML framework allows the retrieval of high-quality spatially distributed and multi-temporal SPAD-based chlorophyll maps at an ultra-high pixel resolution (e.g., 7 mm).

Modelling Stand Variables of Beech Coppice Forest Using Spectral Sentinel-2A Data and the Machine Learning Approach

Background and Purpose: Coppice forests have a particular socio-economic and ecological role in forestry and environmental management. Their production sustainability and spatial stability become imperative for forestry sector as well as for local and global communities. Recently, integrated forest inventory and remotely sensed data analysed with non-parametrical statistical methods have enabled more detailed insight into forest structural characteristics. The aim of this research was to estimate forest attributes of beech coppice forest stands in the Sarajevo Canton through the integration of inventory and Sentinel S2A satellite data using machine learning methods. Materials and Methods: Basal area, mean stand diameter, growing stock and total volume data were determined from the forest inventory designed for represented stands of coppice forests. Spectral data were collected from bands of Sentinel S2A satellite image, vegetation indices (difference, normalized difference and ratio vegetation index) and biophysical variables (fraction of absorbed photosynthetically active radiation, leaf area index, fraction of vegetation cover, chlorophyll content in the leaf and canopy water content). Machine learning rule-based M5 model tree (M5P) and random forest (RF) methods were used for forest attribute estimation. Predictor subset selection was based on wrapping assuming M5P and RF learning schemes. Models were developed on training data subsets (402 sample plots) and evaluations were performed on validation data subsets (207 sample plots). Performance of the models was evaluated by the percentage of the root mean squared error over the mean value (rRMSE) and the square of the correlation coefficient between the observed and estimated stand variables. Results and Conclusions: Predictor subset selection resulted in a varied number of predictors for forest attributes and methods with their larger contribution in RF (between 8 and 11). Spectral biophysical variables dominated in subsets. The RF resulted in smaller errors for training sets for all attributes than M5P, while both methods delivered very high errors for validation sets (rRMSE above 50%). The lowest rRMSE of 50% was obtained for stand basal area. The observed variability explained by the M5P and RF models in training subsets was about 30% and 95% respectively, but those values were lower in test subsets (below 12%) but still significant. Differences of the sample and modelled forest attribute means were not significant, while modelled variability for all forest attributes was significantly lower (p<0.01). It seems that additional information is needed to increase prediction accuracy, so stand information (management classes, site class, soil type, canopy closure and others), new sampling strategy and new spectral products could be integrated and examined in further more complex modelling of forest attributes.

Age estimation in a long-lived seabird (Ardenna tenuirostris) using DNA methylation-based biomarkers.

Age structure is a fundamental aspect of animal population biology. Age is strongly related to individual physiological condition, reproductive potential and mortality rate. Currently, there are no robust molecular methods for age estimation in birds. Instead, individuals must be ringed as chicks to establish known-age populations, which is a labour-intensive and expensive process. The estimation of chronological age using DNA methylation (DNAm) is emerging as a robust approach in mammals including humans, mice and some non-model species. Here, we quantified DNAm in whole blood samples from a total of 71 known-age Short-tailed shearwaters (Ardenna tenuirostris) using digital restriction enzyme analysis of methylation (DREAM). The DREAM method measures DNAm levels at thousands of CpG dinucleotides throughout the genome. We identified seven CpG sites with DNAm levels that correlated with age. A model based on these relationships estimated age with a mean difference of 2.8years to known age, based on validation estimates from models created by repeated sampling of training and validation data subsets. Longitudinal observation of individuals re-sampled over 1 or 2years generally showed an increase in estimated age (6/7 cases). For the first time, we have shown that epigenetic changes with age can be detected in a wild bird. This approach should be of broad interest to researchers studying age biomarkers in non-model species and will allow identification of markers that can be assessed using targeted techniques for accurate age estimation in large population studies.

Acoustic Velocity—Wood Fiber Attribute Relationships for Jack Pine Logs and Their Potential Utility

This study presents an acoustic-based predictive modeling framework for estimating a suite of wood fiber attributes within jack pine (Pinus banksiana Lamb.) logs for informing in-forest log-segregation decision-making. Specifically, the relationships between acoustic velocity (longitudinal stress wave velocity; vl) and the dynamic modulus of elasticity (me), wood density (wd), microfibril angle (ma), tracheid wall thickness (wt), tracheid radial and tangential diameters (dr and dt, respectively), fiber coarseness (co), and specific surface area (sa), were parameterized deploying hierarchical mixed-effects model specifications and subsequently evaluated on their resultant goodness-of-fit, lack-of-fit, and predictive precision. Procedurally, the data acquisition phase involved: (1) randomly selecting 61 semi-mature sample trees within ten variable-sized plots established in unthinned and thinned compartments of four natural-origin stands situated in the central portion the Canadian Boreal Forest Region; (2) felling and sectioning each sample tree into four equal-length logs and obtaining twice-replicate vl measurements at the bottom and top cross-sectional faces of each log (n = 4) from which a log-specific mean vl value was calculated; and (3) sectioning each log at its midpoint and obtaining a cross-sectional sample disk from which a 2 × 2 cm bark-to-pith radial xylem sample was extracted and subsequently processed via SilviScan-3 to derive annual-ring-specific attribute values. The analytical phase involved: (1) stratifying the resultant attribute—acoustic velocity observational pairs for the 243 sample logs into approximately equal-sized calibration and validation data subsets; (2) parameterizing the attribute—acoustic relationships employing mixed-effects hierarchical linear regression specifications using the calibration data subset; and (3) evaluating the resultant models using the validation data subset via the deployment of suite of statistical-based metrics pertinent to the evaluation of the underlying assumptions and predictive performance. The results indicated that apart from tracheid diameters (dr and dt), the regression models were significant (p ≤ 0.05) and unbiased predictors which adhered to the underlying parameterization assumptions. However, the relationships varied widely in terms of explanatory power (index-of-fit ranking: wt (0.53) > me > sa > co > wd >> ma (0.08)) and predictive ability (sa > wt > wd > co >> me >>> ma). Likewise, based on simulations where an acoustic-based wd estimate is used as a surrogate measure for a Silviscan-equivalent value for a newly sampled log, predictive ability also varied by attribute: 95% of all future predictions for sa, wt, co, me, and ma would be within ±12%, ±14%, ±15%, ±27%, and ±55% and of the true values, respectively. Both the limitations and potential utility of these predictive relationships for use in log-segregation decision-making, are discussed. Future research initiatives, consisting of identifying and controlling extraneous sources of variation on acoustic velocity and establishing attribute-specific end-product-based design specifications, would be conducive to advancing the acoustic approach in boreal forest management.

Molecular Subtype Classification of Diffuse Large B-Cell Lymphoma By Immunohistochemical Algorithms and Automatic Supervised Classifiers

Diffuse large B-cell lymphoma (DLBCL) can be classified into germinal center-like (GCB) and non-germinal center-like (non-GCB) molecular subtype. These entities are driven by different intracellular oncogenic signaling pathways that lead to a distinct clinical outcome (Fang, Xu, & Li, 2010; Lenz et al., 2008). Several immunohistochemical (IHC)-based DLBCL classification algorithms have been proposed, this considers the case when genetic expression profile (GEP) studies are not available. However, there is a major discrepancy within IHC algorithms, and when they are compared to GEP (Coutinho et al., 2013).To address these inconsistencies and determine if an automatic classifier could be used to accurately categorize DLBCL subtype, we perfomed a present a performance comparison between eight reported IHC algorithms (Colomo, Hans, Hans modified [Hans*], Nyman, Choi, Choi modified [Choi*] and Visco-Young with three [VY3] and four [VY4] antibodies) against their counterparts developed by automatic classification techniques, which consider the following structures: Bayesian Classifier (B), Bayesian Simple Classifier (BS), Naïve Bayesian Classifier (BN), Artificial Neural Networks (ANN), and Support Vector Machines (SVM).The Visco-Young database (Visco et al., 2012), which contains GEP, IHC raw data corresponding to GCET1, MUM1, FOXP1, BCL6, and CD10 antibodies, and clinical information of 475 de novo DLBCL patients, was used. According to GEP, the database contained 231 GCB, and 244 non-GCB cases. Each patient in VY database was ranked by survival rate as: low survival (0 - 34 months, 237 patients), medium survival (35 - 69 months, 173 patients), or high survival (70 - 106 months, 65 patients) rate. For the implementation of automatic classifiers, the database was split into training, testing and validation data subsets (75%, 20% and 5% respectively) by random selection, but to preserve the same proportion of ranked patients, the so-called k-fold cross-validation technique was applied. The automatic classifier versions of IHC algorithms used the same raw IHC data (antibody combination) as the input, e.g. VY3 used CD10, FOXP1, and BCL6 raw IHC as well as the ANN VY3. A total of 35 automatic classifiers were trained, where Colomo and Hans use the same set of antibodies and are represented by the same automatic classifiers. The stopping criterion during the training stage for all classification algorithms was an error less than 1x10-3 or 100 training epochs, whichever was satisfied first.The performance of the eight IHC algorithms and the automatic classifiers was evaluated by computing the accuracy (Acc), specificity (Spec), and sensitivity (Sens), according to the Receiver Operating Characteristic procedure. Five classifiers obtained the highest metrics: ANN Choi, BS Choi, and BS Choi* with 94.2% Acc, 93.1% Spec, and 95.2% Sens, followed by SVM Choi and SVM VY4 with 94.2% Acc, 91.4% Spec, and 96.8% Sens. Choi was the IHC algorithm with better metrics (92.5% Acc, 84.5% Spec, and 100% Sens), which ranked 11 out of 43 models tested, followed by VY3 and VY4 (ranked 22 and 23, respectively). Survival of GCB and non-GCB groups identified by these models were compared using Kaplan-Meier curves, and the significance was calculated using log-rank test. For the best five automatic classifiers and the Choi IHC algorithm, GCB overall survival was better than non-GCB cases (p < 0.05).To statistically compare the models with GEP, all automatic classifiers and IHC algorithms results were analyzed by Cohen’s kappa (κ) for agreement analysis and Pearson’s chi-squared test. Only Choi IHC algorithm had a very good agreement when compared with GEP (κ = 0.85, p < 0.001). The best five automatic supervised classifiers provided a perfect agreement with GEP (κ = 0.88, p < 0.001). Moreover, the agreement between IHC algorithms was mainly from moderate to good (κ: 0.41 - 0.79), except for Choi having a very good agreement with both VY3 and VY4 (κ = 0.95, p < 0.001). Conversely, a very good agreement within supervised classifiers was observed (κ: 0.77 - 1.00).Harnessing all of the available immunohistochemical data in order to increase the DLBCL classification accuracy when compared with decision three pre-existing algorithms, we conclude that 4 antibody-based BS Choi* automatic classifier provided the best metrics and represents an affordable and time-saving alternative for DLBCL molecular subtype identification. DisclosuresNo relevant conflicts of interest to declare.

Genomic Prediction Accounting for Residual Heteroskedasticity.

Whole-genome prediction (WGP) models that use single-nucleotide polymorphism marker information to predict genetic merit of animals and plants typically assume homogeneous residual variance. However, variability is often heterogeneous across agricultural production systems and may subsequently bias WGP-based inferences. This study extends classical WGP models based on normality, heavy-tailed specifications and variable selection to explicitly account for environmentally-driven residual heteroskedasticity under a hierarchical Bayesian mixed-models framework. WGP models assuming homogeneous or heterogeneous residual variances were fitted to training data generated under simulation scenarios reflecting a gradient of increasing heteroskedasticity. Model fit was based on pseudo-Bayes factors and also on prediction accuracy of genomic breeding values computed on a validation data subset one generation removed from the simulated training dataset. Homogeneous vs. heterogeneous residual variance WGP models were also fitted to two quantitative traits, namely 45-min postmortem carcass temperature and loin muscle pH, recorded in a swine resource population dataset prescreened for high and mild residual heteroskedasticity, respectively. Fit of competing WGP models was compared using pseudo-Bayes factors. Predictive ability, defined as the correlation between predicted and observed phenotypes in validation sets of a five-fold cross-validation was also computed. Heteroskedastic error WGP models showed improved model fit and enhanced prediction accuracy compared to homoskedastic error WGP models although the magnitude of the improvement was small (less than two percentage points net gain in prediction accuracy). Nevertheless, accounting for residual heteroskedasticity did improve accuracy of selection, especially on individuals of extreme genetic merit.

EuroSCORE II

To update the European System for Cardiac Operative Risk Evaluation (EuroSCORE) risk model.A dedicated website collected prospective risk and outcome data on 22,381 consecutive patients undergoing major cardiac surgery in 154 hospitals in 43 countries over a 12-week period (May-July 2010). Completeness and accuracy were validated during data collection using mandatory field entry, error and range checks and after data collection using summary feedback confirmation by responsible officers and multiple logic checks. Information was obtained on existing EuroSCORE risk factors and additional factors proven to influence risk from research conducted since the original model. The primary outcome was mortality at the base hospital. Secondary outcomes were mortality at 30 and 90 days. The data set was divided into a developmental subset for logistic regression modelling and a validation subset for model testing. A logistic risk model (EuroSCORE II) was then constructed and tested.Compared with the original 1995 EuroSCORE database (in brackets), the mean age was up at 64.7 (62.5) with 31% females (28%). More patients had New York Heart Association class IV, extracardiac arteriopathy, renal and pulmonary dysfunction. Overall mortality was 3.9% (4.6%). When applied to the current data, the old risk models overpredicted mortality (actual: 3.9%; additive predicted: 5.8%; logistic predicted: 7.57%). EuroSCORE II was well calibrated on testing in the validation data subset of 5553 patients (actual mortality: 4.18%; predicted: 3.95%). Very good discrimination was maintained with an area under the receiver operating characteristic curve of 0.8095.Cardiac surgical mortality has significantly reduced in the last 15 years despite older and sicker patients. EuroSCORE II is better calibrated than the original model yet preserves powerful discrimination. It is proposed for the future assessment of cardiac surgical risk.

Abstract 3276: Assessment of Predictive Accuracy of Centers for Medicare and Medicaid Services’ Method to Risk Adjust Patients for Interhospital Comparison of 30-day Mortality Rates

Introduction : The Centers for Medicare and Medicaid Services (CMS) will initiate public reporting of 30 day death rates for hospitals caring for patients with heart failure (HF). While hospital specific rates will be adjusted for medical comorbidities, it is not known if this method is adequate to allow direct hospital comparisons. Hypothesis : More accurate risk adjustment of patients can be accomplished by including variables that have previously been described to be predictors of mortality in HF. Methods : We assessed the CMS HF risk adjustment model in a population of 3639 patients studied in the Guidelines Applied in Practice (GAP) - HF study. Probabilties for 30 day mortality were calculated for the CMS model. Multivariable logistic regression analysis was then performed with other models including variables (no mVO 2 ) from the Heart Failure Survival Score (HFSS), ADHERE Registry model (ARM), body mass index categories (BMI), pre-admission origin of patient (PAO), and admission in the previous 6 months (AP6). Probabilities of mortality were tabulated. Backward selection was then used to derive the best model from all candidate variables from a derivation data subset, followed by testing of the model on a validation data subset. Area under the curve (AUC) was then compared for each model. Results : The AUCs and 95% CI are shown in the table below. Calculated AUCs for the CMS model and separate models were similar. The best model was defined by the following variables: admission in the previous 6 months, BUN &gt; 43, systolic blood pressure &gt; 115, 35 ≤ BMI &lt; 40, BMI ≥ 40, mean arterial pressure, serum sodium, and PAO as well as the CMS model variables. The AUCs for the derivation and validation sets were both significantly greater than that provided by the CMS model alone. Conclusions : Risk adjustment for the purpose of interhospital comparison of 30-day mortality rates is best performed with models that include clinical admission variables in addition to the medical comorbidites. Comparison of AUCs for 30 day mortality prediction

Retrieval of pigment concentrations and size structure of algal populations from their absorption spectra using multilayered perceptrons

Spectral absorption coefficients of phytoplankton can now be derived, under some assumptions, from hyperspectral ocean color measurements and thus become accessible from space. In this study, multilayer perceptrons have been developed to retrieve information on the pigment composition and size structure of phytoplankton from these absorption spectra. The retrieved variables are the main pigment groups (chlorophylls a, b, c, and photosynthetic and nonphotosynthetic carotenoids) and the relative contributions of three algal size classes (pico-, nano-, and microphytoplankton) to total chlorophyll a. The networks have been trained, tested, and validated using more than 3,700 simultaneous absorption and pigment measurements collected in the world ocean. Among pigment groups, chlorophyll a is the most accurately retrieved (average relative errors of 17% and 16% for the test and validation data subsets, respectively), while the poorest performances are found for chlorophyll b (average relative errors of 51% and 40%). Relative contributions of algal size classes to total chlorophyll a are retrieved with average relative errors of 19% to 33% for the test subset and of 18% to 47% for the validation subset. The performances obtained for the validation data, showing no strong degradation with respect to test data, suggest that these neural networks might be operated with similar performances for a large variety of marine areas.

An Improved Outdoor Calibration Procedure for BroadbandUltraviolet Radiometers

This article aims at improving the broadband ultraviolet radiometer's calibration methodology. For this goal, three broadband radiometers are calibrated against a spectrophotometer of reference. Three different one-step calibration models are tested: ratio, first order and second order. The latter is proposed in order to adequately reproduce the high dependence on the solar zenith angle shown by the other two models and, therefore, to improve the calibration performance at high solar elevations. The proposed new second-order model requires no additional information and, thus, keeps the operational character of the one-step methodology. The models are compared in terms of their root mean square error and the most qualified is subsequently validated by comparing its predictions with the spectrophotometer measurements within an independent validation data subset. Results show that the best calibration is achieved by the second-order model, with a mean bias error and mean absolute bias error lower than 2.2 and 6.7%, respectively.

Estimation of sensible and latent heat flux from natural sparse vegetation surfaces using surface renewal

This paper reports a study undertaken to evaluate the feasibility of the surface renewal method to accurately estimate long-term evaporation from the playa and margins of an endorreic salty lagoon (Gallocanta lagoon, Spain) under semiarid conditions. High-frequency temperature readings were taken for two time lags ( r) and three measurement heights ( z) in order to get surface renewal sensible heat flux ( H SR ) values. These values were compared against eddy covariance sensible heat flux ( H EC ) values for a calibration period (25–30 July 2000). Error analysis statistics (index of agreement, IA; root mean square error, RMSE; and systematic mean square error, MSE s ) showed that the agreement between H SR and H EC improved as measurement height decreased and time lag increased. Calibration factors α were obtained for all analyzed cases. The best results were obtained for the z=0.9 m ( r=0.75 s) case for which α=1.0 was observed. In this case, uncertainty was about 10% in terms of relative error ( RE). Latent heat flux values were obtained by solving the energy balance equation for both the surface renewal ( LE SR ) and the eddy covariance ( LE EC ) methods, using H SR and H EC , respectively, and measurements of net radiation and soil heat flux. For the calibration period, error analysis statistics for LE SR were quite similar to those for H SR , although errors were mostly at random. LE SR uncertainty was less than 9%. Calibration factors were applied for a validation data subset (30 July–4 August 2000) for which meteorological conditions were somewhat different (higher temperatures and wind speed and lower solar and net radiation). Error analysis statistics for both H SR and LE SR were quite good for all cases showing the goodness of the calibration factors. Nevertheless, the results obtained for the z=0.9 m ( r=0.75 s) case were still the best ones.