The Intelligent Ventilator (INVENT) project: the role of mathematical models in translating physiological knowledge into clinical practice

This dissertation has addressed the broad hypothesis as to whether building mathematical models is useful as a tool for translating physiological knowledge into clinical practice. In doing so it describes work on the INtelligent VENTilator project (INVENT), the goal of which is to build, evaluate and integrate into clinical practice, a model-based decision support system for control of mechanical ventilation. The dissertation describes the mathematical models included in INVENT, i.e. a model of pulmonary gas exchange focusing on oxygen transport, and a model of the acid-base status of blood, interstitial fluid and tissues. These models have been validated, and applied in two other systems: ALPE, a system for measuring pulmonary gas exchange and ARTY, a system for arterialisation of the acid-base and oxygen status of peripheral venous blood. The major contributions of this work are as follows. A mathematical model has been developed which can describe pulmonary gas exchange more accurately that current clinical techniques. This model is parsimonious in that it can describe pulmonary gas exchange from measurements easily available in the clinic, along with a readily automatable variation in F(I)O(2). This technique and model have been developed into a research and commercial tool (ALPE), and evaluated both in the clinical setting and when compared to the reference multiple inert gas elimination technique (MIGET). Mathematical models have been developed of the acid- base chemistry of blood, interstitial fluid and tissues, with these models formulated using a mass-action mass-balance approach. The model of blood has been validated against literature data describing the addition and removal of CO(2), strong acid or base, and haemoglobin; and the effects of oxygenation or deoxygenation. The model has also been validated in new studies, and shown to simulate accurately and precisely the mixing of blood samples at different PCO(2) and PO(2) levels. This model of acid-base chemistry of blood has been applied in the ARTY system. ARTY has been shown to accurately and precisely calculate arterial values of acid-base and oxygen status in patients residing in the ICU, and in those with chronic lung disease. The INtelligent VENTilator (INVENT) system has been developed for optimization of mechanical ventilator settings using physiological models and utility/penalty functions, separating physiological knowledge from clinical preference. The models can be tuned to the individual patient via parameter estimation, providing patient specific advice. The INVENT team has shown prospectively that the system provides advice on F(I)O(2) which is as good as clinical practice, and retrospectively that the system provides reasonable suggestions of tidal volume, respiratory frequency and F(I)O(2). In general, this dissertation has illustrated a further example of the role of modeling in describing and understanding complex systems. The dissertation has shown that when dealing with complexity the goal of the model must be in focus if a correct balance is to be maintained between system complexity and model parameterization. The original goal of the INVENT team, i.e. to build, evaluate and integrate a DSS for control of mechanical ventilation has not as yet been completed. However, the broader hypothesis that building models generates new and interesting questions has been successfully demonstrated. The ALPE model and system has been applied in intensive care, post operative care and cardiology and is currently being evaluated in new clinical domains. ARTY has been shown to have potential benefit in eliminating the need for painful arterial punctures, and may also be useful as a screening tool. These systems illustrate the benefits of investing in models as a mechanism for translating physiological knowledge to clinical practice.     

A knowledge-based support system for mechanical ventilation of the lungs. The KUSIVAR concept and prototype

The KUSIVAR is an expert system for mechanical ventilation of adult patients suffering from respiratory insufficiency. Its main objective is to provide guidance in respirator management. The knowledge base includes both qualitative, rule-based knowledge and quantitative knowledge expressed in the form of mathematical models (expert control) which is used for prediction of arterial gas tensions and optimization purposes. The system is data driven and uses a forward chaining mechanism for rule invocation. The interaction with the user will be performed in advisory, critiquing, semi-automatic and automatic modes. The system is at present in an advanced prototype stage. Prototyping is performed using KEE (Knowledge Engineering Environment) on a Sperry Explorer workstation. For further development and clinical use the expert system will be downloaded to an advanced PC. The system is intended to support therapy with a Siemens-Elema Servoventilator 900 C.     

Planar Translational Cable‐Direct‐Driven Robots

We present the simulated dynamics and control of a planar, translational cable‐directdriven robot (CDDR). The motivation behind this work is to improve the serious cable interference problem with existing CDDRs and to avoid configurations where negative cable tensions are required to exert general forces and moments on the environment and during dynamic motions. Generally for CDDRs the commanded rotations are more demanding than commanded translations in terms of slack cable conditions. Therefore we propose a translational CDDR whose end‐effector may be fitted with a traditional serial wrist mechanism to provide rotational freedom (assuming proper design to resist the moments). Only the translational CDDR is considered in this article, including kinematics and statics modeling, statics workspace (wherein all possible Cartesian forces and moments may be exerted with only positive cable tensions), plus a dynamics model and simulated control for planar CDDRs. Here we focus only on planar CDDRs, to clearly demonstrate our dynamics and control work; we will extend this work to spatial CDDRs in the future. Examples are presented to demonstrate simulated control including feedback linearization of the four‐cable CDDR (with one degree of actuation redundancy) performing a Cartesian task. An on‐line dynamic minimum torque estimation algorithm is introduced to ensure all cable tensions remain positive for all motion; otherwise slack cables can result from CDDR dynamics and control is lost. © 2003 Wiley Periodicals, Inc.     

Translational Planar Cable-Direct-Driven Robots

A planar cable-direct-driven robot (CDDR) architecture is introduced with only translational freedoms. The motivation behind this work is to improve the serious cable interference problem with existing CDDRs and to avoid configurations where negative cable tensions are required to exert general forces on the environment and during dynamic motions. These problems generally arise for rotational CDDR motions. Thus, we propose a class of purely translational CDDRs; of course, these are not general but may only perform tasks where no rotational motion or resistance of moments is required at the end-effector. This article includes kinematics and statics modeling, determination of the statics workspace (the space wherein all possible Cartesian forces may be exerted with only positive cable tensions), plus a dynamics model and simulated control for planar translational CDDRs. Examples are presented to demonstrate simulated control including feedback linearization of the 4-cable CDDR (with two degrees of actuation redundancy) performing a Cartesian task. We introduce an on-line dynamic minimum torque estimation algorithm to ensure all cable tensions remain positive for all motion; otherwise slack cables result from the CDDR dynamics and control is lost.     

Ensemble data mining modeling in corrosion of concrete sewer: A comparative study of network-based (MLPNN & RBFNN) and tree-based (RF,CHAID, & CART) models

This research aims to evaluate ensemble learning (bagging, boosting, and modified bagging) potential in predicting microbially induced concrete corrosion in sewer systems from the data mining (DM) perspective. Particular focus is laid on ensemble techniques for network-based DM methods, including multi-layer perceptron neural network (MLPNN) and radial basis function neural network (RBFNN) as well as tree-based DM methods, such as chi-square automatic interaction detector (CHAID), classification and regression tree (CART), and random forests (RF). Hence, an interdisciplinary approach is presented by combining findings from material sciences and hydrochemistry as well as data mining analyses to predict concrete corrosion. The effective factors on concrete corrosion such as time, gas temperature, gas-phase H₂S concentration, relative humidity, pH, and exposure phase are considered as the models’ inputs. All 433 datasets are randomly selected to construct an individual model and twenty component models of boosting, bagging, and modified bagging based on training, validating, and testing for each DM base learners. Considering some model performance indices, (e.g., Root mean square error, RMSE; mean absolute percentage error, MAPE; correlation coefficient, r) the best ensemble predictive models are selected. The results obtained indicate that the prediction ability of the random forests DM model is superior to the other ensemble learners, followed by the ensemble Bag-CHAID method. On average, the ensemble tree-based models acted better than the ensemble network-based models; nevertheless, it was also found that taking the advantages of ensemble learning would enhance the general performance of individual DM models by more than 10%.     

A multiple measurements case-based reasoning method for predicting recurrent status of liver cancer patients

In general, the studies introducing the medical predictive models which frequently handle time series data by direct matching between pairs of features within sequences during calculation of similarity may have following limitations: (1) direct matching may not be a suitable matching because these paired cases by a fixed order may not be with the most similar temporal information, and (2) when two patients have different numbers of multiple cases, some cases may be ignored. For example, one patient with four cases and another one with five cases, only first four cases of these two patients are paired and the left one case may be ignored. In this paper, in order to dynamically determine matching pairs among cases and pair all cases between two patients, we propose a multiple measurements case-based reasoning (MMCBR) to be used for building liver cancer recurrence predictive models. MMCBR and single measurement case-based reasoning (SingleCBR) are evaluated and compared. According to experiment results in this study, the performance of MMCBR models is better than that of SingleCBR models. Multiple measurements accumulated during a period of time do have benefits for building predictive models with improved performance based on this proposed MMCBR method.     

Mapping live fuel moisture with MODIS data: A multiple regression approach

Live fuel moisture (LFM) is an important factor for ascertaining fire risk in shrublands located in Mediterranean climate regions. We examined empirical relationships between LFM and numerous vegetation indices calculated from MODIS composite data for two southern California shrub functional types, chaparral (evergreen) and coastal sage scrub (CSS, drought-deciduous). These relationships were assessed during the annual March–September dry down period for both individual sites, and sites pooled by functional type. The visible atmospherically resistant index (VARI) consistently had the strongest relationships for individual site regressions. An independent method of accuracy assessment, cross validation, was used to determine model robustness for pooled site regressions. Regression models were developed with n ? 1 datasets and tested on the dataset that was withheld. Additional variables were included in the regression models to account for site-specific and interannual differences in vegetation amount and condition. This allowed a single equation to be used for a given functional type. Multiple linear regression models based on pooled sites had slightly lower adjusted R² values compared with simple linear regression models for individual sites. The best regression models for chaparral and CSS were inverted, and LFM was mapped across Los Angeles County, California (LAC). The methods used in this research show promise for monitoring LFM in chaparral and may be applicable to other Mediterranean shrubland communities.     

Evaluating the effects of task-individual-technology fit in multi-DSS models context: A two-phase view

We investigate the effects of individual difference with the framework of task-individual-technology fit under a multi-DSS models context using a two-phase view. Our research question is: in addition to task-technology fit, does individual-technology fit or individual-task fit matter in users' attitude and performance in the multi-tasks and multi-DSS models context? We first divide the concept of task-individual-technology fit into three dimensions - task-technology fit (TTF), individual-technology fit (IT_eF), and task-individual fit (T_aIF) - so that we can explore mechanisms and effects of interaction among these factors (i.e., task, individual difference, and technology). We then propose a two-phase view of task-individual-technology fit (i.e., pre-paradigm phase and paradigm phase) based on Kuhn's concept of revolutionary science. We conducted a controlled laboratory experiment with multiple DSS models and decision-making tasks to test our hypotheses. Results confirmed our arguments that in the paradigm phase, the effects of individual differences on user attitudes toward DSS models can be ignored and that in the pre-paradigm phases individual differences play an important role. In addition, we found that effects of individual difference can be a two-blade sword: IT_eF can enhance but T_aIF can diminish users' attitude to DSS model (i.e., technology). Our results also suggested that different dimensions of fit may affect performance directly or indirectly.     

The development of risk assessment models for carpal tunnel syndrome: a case-referent study

The present study developed risk assessment models for carpal tunnel syndrome (CTS) which can provide information of the likelihood of developing CTS for an individual having certain personal characteristics and occupational risks. A case-referent study was conducted consisting of two case groups and one referent group: (1) 22 work-related CTS patients (W-CTS), (2) 25 non-work related CTS patients (NW-CTS), and (3) 50 healthy workers (HEALTHY) having had no CTS history. The classification of CTS patients into one of the case groups was determined according to the type of insurance covering their medical costs. Personal characteristics, psychosocial stresses at work, and physical work conditions were surveyed by using a questionnaire tailor-designed to CTS (reliability of each scale > or = 0.7). By contrasting the risk information of each case group to that of the referent group, three logistic regression models were developed: W-CTS/HEALTHY, NW-CTS/HEALTHY, and C-CTS/HEALTHY (C-CTS, the combined group of W-CTS and NW-CTS). ROC analysis indicated that the models have satisfactory discriminability (d' = 1.91 to 2.51) and high classification accuracy (overall accuracy = 83-89%). Both W-CTS/HEALTHY and C-CTS/HEALTHY include personal and physical factors, while NW-CTS/HEALTHY involves only personal factors. This suggests that the injury causation of NW-CTS patients should be attributable mainly to their 'high' personal susceptibility to the disorder rather than exposure to adverse work conditions, while that of W-CTS patients be attributable to improper work conditions and CTS-prone personal characteristics in combination.     

醇类化合物色谱保留指数与结构参数定量关系的研究

采用拓扑指数(~mQ)、定位基参数(S_(ox))与醇在6种固定相上的气相色谱保留指数值(RI)进行了相关分析,发现RI与上述参数间存在良好的相关性,其关系式可表示为：RI=d (a~0)Q-(b~1)Q cS_(ox),其中a、b、c、d为系数。相关系数均大于0．98。继以留一法(Leave-one-out,LOO)进行交互检验,相关系数RCV均大于0．97。说明所建定量结构-保留关系(QSRR)模型具有良好的稳定性和预测能力,较好地揭示了醇类物质在不同固定相上气相色谱保留指数的变化规律。     

The relationship between shoulder torques and the perception of muscular effort in loaded reaches

The objective of this study was to define the quantitative relationship between external dynamic shoulder torques and calibrated perceived muscular effort levels for load delivery tasks, for application in job analyses. Subjects performed a series of loaded reaches and, following each exertion, rated their perceived shoulder muscular effort. Motion and task physical requirements data were processed with a biomechanical upper extremity model to calculate external dynamic shoulder torques. Calculated torque values were then statistically compared to reported calibrated perceived muscular effort scores. Individual subject torque profiles were significantly positively correlated with perceived effort scores (r2 = 0.45-0.77), with good population agreement (r2 = 0.50). The accuracy of the general regression model improved (r2 = 0.72) with inclusion of factors specific to task geometry and individual subjects. This suggests two major conclusions: 1) that the perception of muscular shoulder effort integrates several factors and this interplay should be considered when evaluating tasks for their impact on the shoulder region; 2) the torque/perception relationship may be usefully leveraged in job design and analysis.     

Surface tension calculation of the Ni3S2---FeS---Cu2S mattes

The surface tension of the Ni₃S₂-FeS-Cu₂S ternary mattes have been calculated from those of the Ni₃S₂-FeS·FeS-Cu₂S and Cu₂S-Ni₃S₂ pseudo-binary boundary systems by using geometric models (Kohler and Toop model) and a general solution model (Chou model). Taking account of the experimental accuracy of ±2.5%, the respective calculated results are in agreement with the experimental data. At the same time, surface tensions of molten Ni₃S₂-FeS-Cu₂S mattes and their pseudo-binary boundary systems are predicted based on Butler's original treatment with great attention to the model parameter β. The predicted results also agree with the experimental data. Therefore, the resulting iso-surface tension curves in molten Ni₃S₂-FeS-Cu₂S mattes at 1473K, especially calculated by using Kohler, Toop and Chou models, are generally acceptable.     

Novel atom-type AI indices for QSPR studies of alcohols

The novel vertex degree v(m) for heteroatom in molecular graph is derived on the basis of the valence connectivity delta(v) of Kier-Hall. The newly proposed atom-type Al indices and previously proposed Xu index, are further modified for compounds with heteroatoms by replacing the vertex-degree of heteroatom by the proposed v(m). The multiple linear regression using the modified Xu index and Al indices can provide high-quality QSPR models for the normal boiling points (BP), molar volumes (MV), molar refractions (MR), and molecular total surface areas (TSA) of alcohols with up to 17 non-hydrogen atoms. The results imply that these physical properties may be expressed as a linear combination of the individual indices related to molecular size and atom-types. For each of the four properties, the correlation coefficient r is greater than 0.996 and particularly the decrease in the standard error is within the range of 61 83% compared with the simple linear models based on the modified Xu index, and the standard errors are 3.814, 0.939, 0.187, and 3.348 for BP, MV, MR, and TSA, respectively. The final models correspond to a fit error of 2.33, 0.70, 0.53, and 0.95% for BP, MV, MR, and TSA, respectively. The more general leave-n-out method is used to do the cross-validation. The cross-validation demonstrates the outstanding predictive power of the final models. The contributions of individual indices are used to illustrate the role of the molecular size and individual groups in molecules. The results indicate that physical properties of alcohols are dominated by the molecular size. On the other hand, although the hydrogen-bonding interactions caused by the -OH group play an important role in determining the normal BPs, the branching seems to be a more important factor influencing the MVs, MRs, and TSAs than the hydrogen-bonding interaction. The contribution of individual atom type or group to properties is not a constant and depends on its structural environment in a molecule.     

Understanding burn severity sensing in Arctic tundra: exploring vegetation indices,suboptimal assessment timing and the impact of increasing pixel size

Little is known about how satellite imagery can be used to describe burn severity in tundra landscapes. The Anaktuvuk River Fire (ARF) in 2007 burned over 1000 km² of tundra on the North Slope of Alaska, creating a mosaic of small (1 m²) to large (>100 m²) patches that differed in burn severity. The ARF scar provided us with an ideal landscape to determine if a single-date spectral vegetation index can be used once vegetation recovery began and to independently determine how pixel size influences burn severity assessment. We determine and explore the sensitivity of several commonly used vegetation indices to variation in burn severity across the ARF scar and the influence of pixel size on the assessment and classification of tundra burn severity. We conducted field surveys of spectral reflectance at the peak of the first growing season post-fire (extended assessment period) at 18 field sites that ranged from high to low burn severity. In comparing single-date indices, we found that the two-band enhanced vegetation index (EVI2) was highly correlated with normalized burn ratio (NBR) and better distinguished among three burn severity classes than both the NBR and the normalized difference vegetation index (NDVI). We also show clear evidence that shortwave infrared (SWIR) reflectivity does not vary as a function of burn severity. By comparing a Quickbird scene (2.4 m pixels) to simulated 30 and 250 m pixel scenes, we are able to confirm that while the moderate spatial resolution of the Landsat Thematic Mapper (TM) sensor (30 m) is sufficient for mapping tundra burn severity, the coarser resolution of the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor (250 m) is not well matched to the fine scale of spatial heterogeneity in the ARF burn scar.     

用于检测糖尿病标志物的电子鼻优化设计

人体呼气中的丙酮含量可作为糖尿病的标志物.为实现无创糖尿病诊断,设计以金属氧化物半导体气敏传感器阵列为核心的人工嗅觉系统,对完成痕量丙酮的快速检测具有重要意义.通过多个气体流量控制器MFC(Mass Flow Controller)分别配制出模拟糖尿病患者呼气样本(30×10-6丙酮)与另两种干扰气体样本(30×10-6乙醇样本、15×10-6丙酮与15×10-6乙醇混合样本)进行实验,基于BP神经网络算法对3种气体定性识别,并通过主成分分析PCA(Principal Component Analysis)算法对原始的高维特征子集进行降维优化.实验表明:PCA与BP算法相结合,可降低BP神经网络的复杂性、减少预测的误差,同时能够解决单个气体传感器交叉敏感问题,进而提高对气体的选择性.对痕量丙酮样本与另两种干扰气体样本进行分析识别,识别的结果显示:对3种样本的识别准确率为91％.该研究为准确识别糖尿病标志物实现无创诊断技术提供了理论指导.     

Re-entrainment of Body Temperature in Experimental Shift-Work Studies

In a series of experimental shift work studies six subjects worked on continuous night shift for three weeks, four subjects worked on a 1-1-1 shift system (1st day morning shift, 2nd day afternoon shift, 3rd day night shift, 4th day off), and two subjects worked on an 2-2-2 shift system (the ‘metropolitan rota’). Rectal temperature was continuously recorded in each experiment.

The greatest changes in the circadian rhythm of body temperature occurred on the second night shift day, and during the first week of continuous night shift working. The changes involved both shifts in the phasing of the rhythm, and alterations in its form. The re-entrainment of the rhythm to its normal phasing took two or more days after two or more successive night shifts, but only one day after a single night shift. Considerable individual differences in the rhythm adjustment to night work were observed.

In a field validation of these findings, 34 shiftworkers in municipal gas and water supply undertakings measured their oral temperature every 2h both on and off duty. Although only very few temperature readings were obtained during sleep periods, the results confirmed, in general, those of the experimental studies.     

Study of the ligament tension and cross-section deformation using nonlinear finite element/multibody system algorithms

In this investigation, the effects of the knee-joint movements on the ligament tension and cross-section deformation are examined using large displacement nonlinear finite element/multibody system formulations. Two knee-joint models that employ different constitutive equations and significantly different deformation kinematics are developed and implemented to analyze the ligament dynamics in a computational solution procedure that integrates large displacement finite element and multibody system algorithms. The first model employs a lower fidelity large displacement cable element that does not capture the cross-section deformations and allows for using only nonlinear classical beam theory with a linear Hookean material law instead of a general continuum mechanics approach. In the second model, a higher fidelity large displacement beam model that captures more coupled deformation modes including Poisson modes as well the cross-section deformation is used. This higher fidelity model also allows for a straight forward implementation of general nonlinear constitutive models, such as Neo Hookean material laws, based on a general continuum mechanics approach. Cauchy stress tensor and Nanson’s formula are used to obtain an accurate expression for the ligament tension forces, which as shown in this investigation depend on the ligament cross section deformation. The two models are implemented in a general multibody system algorithm that allows introducing general constraint and force functions. The finite element/multibody system computational algorithm used in this investigation is based on an optimum sparse matrix structure and ensures that the kinematic constraint equations are satisfied at the position, velocity, and acceleration levels. The results obtained in this investigation show that models that ignore coupled deformation modes including some Poisson modes and the cross-section deformations can lead to inaccurate prediction of the ligament forces. These simpler models, as demonstrated in this investigation, can be used to obtain only simplified expressions for the ligament tensions. A three-dimensional knee-joint model that consists of five bodies including two flexible bodies that represent the medial collateral ligament (MCL) and lateral collateral ligament (LCL) is used in the numerical comparative study presented in this paper. The large displacement procedure presented in this investigation can be applied to other types of Ligaments, Muscles, and Soft Tissues (LMST) in biomechanics applications.     

Toward an optimal ensemble of kernel-based approximations with engineering applications

This paper presents a general approach toward the optimal selection and ensemble (weighted average) of kernel-based approximations to address the issue of model selection. That is, depending on the problem under consideration and loss function, a particular modeling scheme may outperform the others, and, in general, it is not known a priori which one should be selected. The surrogates for the ensemble are chosen based on their performance, favoring non-dominated models, while the weights are adaptive and inversely proportional to estimates of the local prediction variance of the individual surrogates. Using both well-known analytical test functions and, in the surrogate-based modeling of a field scale alkali-surfactant-polymer enhanced oil recovery process, the ensemble of surrogates, in general, outperformed the best individual surrogate and provided among the best predictions throughout the domains of interest. This work was supported in part by the Fondo Nacional de Ciencia, Tecnología e Innovación (FONACIT), Venezuela under Grant F-2005000210. N. Q. Author also acknowledges that this material is based upon work supported by National Science Foundation under Grant DDM-423280.     

Surface tension calculation of the Sn–Ga–In ternary alloy

Surface tensions of the Sn–Ga–In ternary alloy are calculated from the surface tensions of the Sn–Ga, Ga–In and In–Sn sub-binary systems by using geometric models (the Kohler model and the Toop model), and a general solution model (the Chou model). The calculated results are in excellent agreement with the experimental data (except for the Kohler model), assuming an experimental accuracy of ±2.5%. At the same time, the surface tensions of the Sn–Ga–In ternary alloy at 773 K and their sub-binary systems are predicted on the basis of Butler’s equation, in combination with thermodynamic data, and different model parameters β$\beta$ equal to the ratio of the coordination number in the surface phase to that in the bulk phase. Different values of β$\beta$ have almost no influence on the surface tensions. The predicted results agree with the experimental data. Therefore, the resulting iso-surface tension curves for the Sn–Ga–In ternary alloy at 773 K, especially those calculated by using the Toop model, are reasonable.     

Estimating forage quantity and quality under different stress and senescent biomass conditions via spectral reflectance

Assessing forage quantity and quality through remote sensing can facilitate grassland and pasture management. However, the high spatial and temporal variability of canopy conditions may limit the predictive accuracy of models based on reflectance measurements. The objective of this work was to develop this type of models, and to challenge their capacity to predict plant properties under a wide range of environmental conditions. We manipulated Paspalum dilatatum canopies through different stress treatments (flooding, drought, nutrient availability, and control) and by artificially varying the amount of senescent biomass. We measured canopy reflectance and constructed simple models, based on either normalized vegetation indices or a few selected wavebands, to estimate biomass and two variables related to forage quality: proportion of photosynthetic vegetation and biomass C:N ratio. General models satisfactorily predicted plant properties for the whole set of environmental conditions, but failed under specific conditions such as drought (for estimates of plant biomass), fertilization (for estimates of C:N ratio), and different levels of senescent tillers (for estimates of the proportion of photosynthetic vegetation). Where general models failed, specific models, based on different bands, achieved satisfactory accuracy. The general models performed better when based on a few selected bands than when based on two-band vegetation indices, having better accuracy (higher R²) and parsimony (lower BIC). However specific models performed similarly for both approaches (similar R² and BIC). These results indicate that these plant properties can be predicted from reflectance information under a broad range of conditions, but not for some particular conditions, where ancillary data or more complex models are probably needed to increase predictive accuracy.