Higher Root Mean Square Error Research Articles

Model-based characterization of total serum bilirubin dynamics in preterm infants.

This study aims to characterize the age-related natural dynamics of total serum bilirubin (TSB) in preterm infants through a mathematical model and to study the model parameters as potential biomarkers for detecting associated morbidities. We proposed an exponential decay model and applied it to each infant. Patient-specific parameters were obtained by minimizing the error between measured TSB and model output. Modeling evaluation was based on root-mean-square error (RMSE). The occurrence of high-risk clinical events was analyzed based on RMSE. In a subset of the CARESS-Premi study involving 373 preterm infants (24-32 weeks' gestation), 72 patient-specific models were fitted. RMSE ranged from 1.20 to 40.25 µmol/L, with a median [IQR] of 8.74 [4.89, 14.25] µmol/L. Our model effectively characterized TSB dynamics for 72 patients, providing valuable insights from model parameters and fitting errors. To our knowledge, this is the first long-term mathematical description of natural TSB decay in preterm infants. Furthermore, the model was able to estimate the occurrence of clinical events such as necrotizing enterocolitis, as reflected by the relatively high RMSE. Future implications include the development of model-based clinical decision support systems for optimizing NICU monitoring and detecting high-risk events. The study characterizes the natural dynamics of total serum bilirubin in preterm infants (24-32 weeks' gestation) using a patient-specific exponential decay model. The model describes patient-specific patterns of TSB evolution from day three to the first weeks, providing a median [IQR] root-mean-squared error of 8.74 [4.89, 14.25] µmol/L. Complementary to previous studies focusing on the first 72-96 h, our study emphasizes the later decay course, contributing to a comprehensive long-term characterization of the natural TSB dynamics in preterm infants. The proposed model holds potential for clinical decision support systems for the optimization of NICU monitoring and high-risk event detection.

Monitoring and forecasting desertification and land degradation using remote sensing and machine learning techniques in Sistan plain, Iran

Monitoring and predicting desertification in arid regions are crucial for addressing environmental and societal challenges. Remote sensing is vital for tracking land surfaces and ecosystems changes. The study aims to use remote sensing-based data to monitor and predict desertification in the Sistan Plain through a data screening approach. The study's satellite data consisted of Landsat 5 and 8 images taken in June each year over 10 years (1990–2020). Remote sensing-based indices, including land use and land cover (LULC) map, normalized differential vegetation index (NDVI), improved vegetation index (EVI), vegetation condition index (VCI), surface temperature condition index (TCI), modified normalized differential water level index (MNDWI) and salinity index (SI) were used in the study. In addition to satellite data, environmental indices, including standardized precipitation index (SPI) and streamflow drought index (SDI), were used. The study employed the random forest (RF) method and the mixed model of automated cells and Markov chain (CA-Markov) to monitor desertification and quantitatively predict its condition in 2030. Root-mean-square error (RMSE) and mean-square error (MSE) indicators were used to evaluate the error. Based on the findings, the RF correlation coefficient (R2) and RMSE were obtained about 0.97 and 0.08, respectively. High coefficient values and low RMSE values indicate that the random forest model is highly efficient in assessing desertification for the study period from 1990 to 2020. The change detection method revealed that desertification increased from 1990 to 2010 but decreased from 2010 to 2020. The decreasing trend is expected to continue until 2030. The Kappa coefficient for the prediction of desertification in 2030 was found to be 0.94, which indicates a correct classification based on the collected samples. In addition, the study identified the SI and SDI as effective indices in the desertification process in the study area. Overall, this study provides valuable insights into monitoring and predicting desertification, which could help develop appropriate strategies for managing and controlling desertification in the Sistan Plain through remote sensing and machine learning techniques.

Forecasting CO2 emissions of fuel vehicles for an ecological world using ensemble learning, machine learning, and deep learning models.

The continuous increase in carbon dioxide (CO2) emissions from fuel vehicles generates a greenhouse effect in the atmosphere, which has a negative impact on global warming and climate change and raises serious concerns about environmental sustainability. Therefore, research on estimating and reducing vehicle CO2 emissions is crucial in promoting environmental sustainability and reducing greenhouse gas emissions in the atmosphere. This study performed a comparative regression analysis using 18 different regression algorithms based on machine learning, ensemble learning, and deep learning paradigms to evaluate and predict CO2 emissions from fuel vehicles. The performance of each algorithm was evaluated using metrics including R2, Adjusted R2, root mean square error (RMSE), and runtime. The findings revealed that ensemble learning methods have higher prediction accuracy and lower error rates. Ensemble learning algorithms that included Extreme Gradient Boosting (XGB), Random Forest, and Light Gradient-Boosting Machine (LGBM) demonstrated high R2 and low RMSE values. As a result, these ensemble learning-based algorithms were discovered to be the most effective methods of predicting CO2 emissions. Although deep learning models with complex structures, such as the convolutional neural network (CNN), deep neural network (DNN) and gated recurrent unit (GRU), achieved high R2 values, it was discovered that they take longer to train and require more computational resources. The methodology and findings of our research provide a number of important implications for the different stakeholders striving for environmental sustainability and an ecological world.

IoT-based monitoring system and air quality prediction using machine learning for a healthy environment in Cameroon.

This paper is aimed at developing an air quality monitoring system using machine learning (ML), Internet of Things (IoT), and other elements to predict the level of particulate matter and gases in the air based on the air quality index (AQI). It is an air quality assessor and therefore a means of achieving the Sustainable Development Goals (SDGs), in particular, SDG 3.9 (substantial reduction of the health impacts of hazardous substances) and SDG 11.6 (reduction of negative impacts on cities and populations). AQI quantifies and informs the public about air pollutants and their adverse effects on public health. The proposed air quality monitoring device is low-cost and operates in real-time. It consists of a hardware unit that detects various pollutants to assess air quality as well as other airborne particles such as carbon dioxide (CO2), methane (CH4), volatile organic compounds (VOCs), nitrogen dioxide (NO2), carbon monoxide (CO), and particulate matter with an aerodynamic diameter of 2.5 microns or less (PM2.5). To predict air quality, the device was deployed from November 1, 2022, to February 4, 2023, in certain bauxite-rich areas of Adamawa and certain volcanic sites in western Cameroon. Therefore, machine learning algorithm models, namely, multiple linear regression (MLR), support vector regression (SVR), random forest regression (RFR), XGBoost (XGB), and K-nearest neighbors (KNN) were applied to analyze the collected concentrations and predict the future state of air quality. The performance of these models was evaluated using mean absolute error (MAE), coefficient of determination (R-square), and root mean square error (RMSE). The obtained data in this study show that these pollutants are present in selected localities albeit to different extents. Moreover, the AQI values obtained range from 10 to 530, with a mean of 132.380 ± 63.705, corresponding to moderate air quality state but may induce an adverse effect on sensitive members of the population. This study revealed that XGB regression performed better in air quality forecasting with the highest R-squared (test score of 0.9991 and train score of 0.9999) and lowest RMSE (test score of 1.5748 and train score of 0. 0073) and MAE (test score of 0.0872 and train score of 0.0020), while the KNN model had the worst prediction (lowest R-squared and highest RMSE and MAE). This embryonic work is a prototype for projects in Cameroon as measurements are underway for a national spread over a longer period of time.

The Verification of Sea Level Anomaly (SLA) Versus Tide Gauge for Geoid Modelling

The increasing long-term sea level rise is of particular interest and needs a significant investment in coastal protection procedures. The coastal sea level has been verified with tide gauges and well maintained with reliable accuracy of data since a few centuries ago. Satellite altimeter has been used for sea level study and increasingly for operational purposes. However, the precision of altimetry data in coastal areas is frequently eliminated by the geographically correlated orbit errors, which induce a regional inhomogeneous mission bias. This study presents an effort to verify sea level anomaly based on altimetry data with tide gauge stations as an early procedure to provide a good sea level anomaly data for mean sea surface, gravity anomaly, and geoid determination over Malaysian seas. The altimetry data is validated by comparing the sea level anomaly with ground truth data (tidal data) near the coastal area. A total of nine selected areas were chosen to represent the ground truth data. The findings found that the altimetry data over the South China Sea (Pulau Tioman, Geting, Cendering, Bintulu, and Kota Kinabalu) provided a good pattern, high correlation, and a minimum root mean square error (RMSE) value after the verification with tidal sea level anomaly data. However, it clearly shows that Tawau station over the Celebes Sea provides a poor pattern, poor correlation, and a high RMSE value with tidal sea level anomaly data. In conclusion, further enhancements are expected from the refined processing and filtering of altimetry data for the sea level study in the coastal zone.

An insight into the Application of AI in maritime and Logistics toward Sustainable Transportation

This review article looks at the developing field of artificial intelligence and machine learning in maritime and marine environment management. The marine industry is increasingly interested in applying advanced AI and ML technologies to solve sustainability, efficiency, and regulatory compliance issues. This paper examines maritime and marine AI and ML applications using a deep literature review and case study analysis. Modeling ship fuel consumption, which impacts the environment and operating expenses, is a top responsibility. The study demonstrates that ML approaches such as Random Forest and Tweedie models can estimate ship fuel use. Statistical analysis demonstrates that the Random Forest model beats the Tweedie model regarding accuracy and consistency. For the training and testing datasets, the Random Forest model has high R2 values of 0.9997 and 0.9926, indicating a solid match. Low Root Mean Square Error (RMSE) and average absolute relative deviation (AARD) suggest that the model accurately reflects fuel use variability. While still performing well, the Tweedie model has lower R2 values and higher RMSE and AARD values, suggesting reduced accuracy and precision in fuel consumption prediction. These findings provide light on the potential applications of artificial intelligence and machine learning in maritime and marine environment management. Advanced analytics enables decision-makers to analyze fuel consumption patterns better, increase operational efficiency, and decrease environmental impact, thus improving maritime sustainability.

VALIDATION OF AN INFLIXIMAB POPULATION PHARMACOKINETIC MODEL FOR MAINTENANCE DOSING IN PEDIATRIC PATIENTS WITH INFLAMMATORY BOWEL DISEASE

Abstract INTRODUCTION The use of infliximab has revolutionized the care of pediatric patients with inflammatory bowel disease (IBD); however, infliximab may lose efficacy over time during the maintenance phase and the use of therapeutic drug monitoring has been advocated to improve response rates. A population pharmacokinetic (popPK) model using infliximab concentrations, common clinical biomarkers, and demographic information to predict future infliximab concentrations would be clinically invaluable. Previous popPK models such as the Buurman [1] and Fasanmade [2] exist, but their utility, accuracy, and predictive values have not been clinically validated. This study aims to validate a reliable set of covariates for an infliximab popPK model that will predict maintenance drug concentrations and guide treatment decisions for children with IBD. METHODS This retrospective study included 134 pediatric patients with 503 infliximab trough concentrations between 5 and 21 years of age, seen at the Division of Pediatric Gastroenterology, Hepatology, and Nutrition at UH Rainbow Babies and Children’s Hospital between January 1, 2008 and December 31, 2018. We recorded the patients' therapeutic infliximab trough concentrations and other clinical and demographic factors. In previous work, we refit published popPK models using induction phase data. Here, we refit our originally constructed one-compartment model based on induction phase data, the Cleveland Model, with maintenance phase data to yield the Cleveland Refit model. We then tested six different models: Buurman (B), Buurman Refit (BR), Fasanmade (F), Fasanmade Refit (FR), Cleveland (C), and Cleveland Refit (CR). Each model was evaluated on root mean square error (RMSE), mean percent error (MPE), and accuracy of the predicted infliximab concentration. We defined accuracy as the percentage of samples each model correctly predicted to be on the same side of 5 ug/mL, which is our clinical trough target. RESULTS On the first infliximab concentration following the induction phase, published models refit on induction phase data (BR, FR) have higher accuracy but also higher RMSE than published models. In contrast, the CR model was refit on maintenance phase data and has higher accuracy (71.1% vs 66.7%) and lower RMSE (6.8 vs 7.8 ug/mL) and MPE (-5% vs 10%) than the previous induction-only model. CONCLUSIONS The CR model improved upon the performance of the induction-only models tested on this cohort. This study suggests refit popPK models on a specific patient population work well for predicting maintenance phase infliximab concentrations, and may be useful for personalization of dose of frequency for pediatric patients with IBD. Further validation with external data in a multicenter, prospective study is needed. References 1) https://doi.org/10.1111/apt.13299 2) https://doi.org/10.1016/j.clinthera.2011.06.002

Models to predict nitrogen excretion from beef cattle fed a wide range of diets compiled from South America.

The objective of this meta-analysis was to develop and evaluate models for predicting nitrogen (N) excretion in feces, urine, and manure in beef cattle in South America. The study incorporated a total of 1,116 individual observations of N excretion in feces and 939 individual observations of N excretion in feces and in urine (g/d), representing a diverse range of diets, animal genotypes, and management conditions in South America. The datasetalso included data on dry matter intake (DMI; kg/d) and nitrogen intake (NI; g/d), concentrations of dietary components, as well as average daily gain (ADG; g/d) and average body weight (BW; kg). Models were derived using linear mixed-effects regression with a random intercept for the study. Fecal N excretion was positively associated with DMI, NI, nonfibrous carbohydrates, average BW, and ADG and negatively associated with EE and CP concentration in the diet. The univariate model predicting fecal N excretion based on DMI (model 1) performed slightly better than the univariate model, which used NI as a predictor variable (model 2) with a root mean square error (RMSE) of 38.0 vs. 39.2%, the RMSE-observations SD ratio (RSR) of 0.81 vs. 0.84, and concordance correlation coefficient (CCC) of 0.53 vs. 0.50, respectively. Models predicting urinary N excretion were less accurate than those derived to predict fecal N excretion, with an average RMSE of 43.7% vs. 37.0%, respectively. Urinary and manure N excretion were positively associated with DMI, NI, CP, average BW, and ADG and negatively associated with neutral detergent fiber concentration in the diet. As opposed to fecal N excretion, the univariate model predicting urinary N excretion using NI (model 10) performed slightly better than the univariate model using DMI (model 9) as predictor variable with an RMSE of 36.0% vs. 39.7%, RSR 0.85 vs. 0.93, and CCC of 0.43 vs. 0.29, respectively. The models developed in this study are applicable for predicting N excretion in beef cattle across a broad spectrum of dietary compositions and animal genotypes in South America. The univariate model using DMI as a predictor is recommended for fecal N prediction, while the univariate model using NI is recommended for predicting urinary and manure N excretion because the use of more complex models resulted in little to no benefits. However, it may be more useful to consider more complex models that incorporate nutrient intakes and diet composition for decision-making when N excretion is a factor to be considered. Three extant equations evaluated in this study have the potential to be used in tropical conditions typical of South America to predict fecal N excretion with good precision and accuracy. However, none of the extant equations are recommended for predicting urine or manure N excretion because of their high RMSE, and low precision and accuracy.

Predicted vs. measured paraspinal muscle activity in adolescent idiopathic scoliosis patients: EMG validation of optimization-based musculoskeletal simulations

Musculoskeletal (MSK) models offer great potential for predicting the muscle forces required to inform more detailed simulations of vertebral endplate loading in adolescent idiopathic scoliosis (AIS). In this work, simulations based on static optimization were compared with in vivo measurements in two AIS patients to determine whether computational approaches alone are sufficient for accurate prediction of paraspinal muscle activity during functional activities.We used biplanar radiographs and marker-based motion capture, ground reaction force, and electromyography (EMG) data from two patients with mild and moderate thoracolumbar AIS (Cobb angles: 21° and 45°, respectively) during standing while holding two weights in front (reference position), walking, running, and object lifting. Using a fully automated approach, 3D spinal shape was extracted from the radiographs. Geometrically personalized OpenSim-based MSK models were created by deforming the spine of pre-scaled full-body models of children/adolescents. Simulations were performed using an experimentally controlled backward approach. Differences between model predictions and EMG measurements of paraspinal muscle activity (both expressed as a percentage of the reference position values) at three different locations around the scoliotic main curve were quantified by root mean square error (RMSE) and cross-correlation (XCorr).Predicted and measured muscle activity correlated best for mild AIS during object lifting (XCorr’s ≥ 0.97), with relatively low RMSE values. For moderate AIS as well as the walking and running activities, agreement was lower, with XCorr reaching values of 0.51 and comparably high RMSE values.This study demonstrates that static optimization alone seems not appropriate for predicting muscle activity in AIS patients, particularly in those with more than mild deformations as well as when performing upright activities such as walking and running.

Potential root mean square error skill score

Consistency, in a narrow sense, denotes the alignment between the forecast-optimization strategy and the verification directive. The current recommended deterministic solar forecast verification practice is to report the skill score based on root mean square error (RMSE), which would violate the notion of consistency if the forecasts are optimized under another strategy such as minimizing the mean absolute error (MAE). This paper overcomes such difficulty by proposing a so-called “potential RMSE skill score,” which depends only on (1) the cross-correlation between forecasts and observations and (2) the autocorrelation of observations. While greatly simplifying the calculation, the new skill score does not discriminate inconsistent forecasts as much, e.g., even MAE-optimized forecasts can attain a high RMSE skill score.

Ensemble Algorithm for Simulated Corrosion Data-tentative

Abstract The application of robotics continues to advance and lends itself to various disciplines. Previous work has shown how Markov decision processes can be integrated in an algorithm to handle pipeline inspection. This has been extended to building models and addressing the errors in the model from the data collected. This paper made use of data collected by simulated robots in a pipeline environment, to build an ensemble. Five models were explored and analysed viz: Linear model, Generalised linear model, K-nearest neighbour, Neural network, and Random Forests models. The models were used to explore the responses acquired from the pipeline corrosion data. The performance of each of the ensembles are evaluated based on the Root Mean Square Error (RMSE), Adjusted R2 and Mean Absolute Error. Model evaluation show that although the averaging ensemble has a better fit (34.56%) than the stacking ensembles (17.22% and 16.98% respectively) considered, it also has the higher RMSE,(81.90) than the stacking models ( 40.91 and 38.53 respectively) forcing the decision maker to make a trade-off between fit and precision.

975. Age and Body Mass Index Impact Optimal Vancomycin Model Selection for Bayesian Dosing: Results from a Large Multi-site Retrospective Study of Adult Patients

Abstract Background Vancomycin population pharmacokinetic (popPK) models embedded in Bayesian software are crucial to informing optimal initial dosing. Since pharmacokinetics (PK) vary based on patients’ age and body mass index (BMI), pharmacists are faced to select the best model amongst many for their patients. To aid in model selection, we used a large multi-site database to determine which models were the most accurate across age and BMI categories. Methods Data on adult patients receiving intravenous vancomycin therapy during January 1, 2022 - December 31, 2022, at healthcare organizations across the U.S. were queried from a Bayesian software database. Data collected included age, weight, height, serum creatinine, sex, vancomycin doses received, and vancomycin concentration levels. The a priori root mean square error (RMSE) for six well-performing and previously validated models was calculated for each patient to measure what the accuracy would have theoretically been if each were used. The average RMSE value per model was calculated and compared between age and BMI categories shown in Table 1. Each age-BMI combination’s most accurate model was reported. Results There were 79,600 patient treatment courses included across 84 health systems. The results for the most accurate model for each age-BMI combination are reported in Table 2. The lowest RMSE was 3.91 mg/L using the modified Goti model in patients > 80 years with BMI < 18.5 kg/m2. The highest RMSE was 7.86 mg/L using the Carreno model in patients < 35 years with BMI > 30 and < 35 kg/m2. A relationship was observed where older age categories lead to better model performance. Conclusion Age and BMI categories influenced which vancomycin popPK models performed best a priori within a Bayesian software when studied within a very robust dataset. These findings may help pharmacists choose which model to use for the most accurate PK calculations and thus inform optimal initial pop-PK vancomycin dosing. It also highlights that adults under 35 years, which tend to be underrepresented in model development populations, showed higher prediction error. Further research is needed on how to best implement these findings into practice. Disclosures Maria-Stephanie Hughes, PharmD, InsightRX: Stocks/Bonds Tiffany Lee, PharmD, InsightRX: Employment

2543. Multisite Analysis of Vancomycin Predictions in a Bayesian Software in Patients with and without Acute Kidney Injury in Adult and Pediatric Patients

Abstract Background Inpatients that require vancomycin therapy often present with an acute kidney injury (AKI). Due to their unstable renal function, pharmacists hesitate to base dosing decisions on predictions provided by pharmacokinetic (PK) models within a Bayesian software. This study will therefore compare accuracy of predictions by various PK models in patients presenting with and without AKI. Methods Adult and pediatric patients receiving vancomycin therapy from 1/1/2022 - 12/31/22 that were dosed using a Bayesian software were retrospectively evaluated. Data on patient age, weight, height, serum creatinine (SCr), sex, and vancomycin doses and levels were recorded. Baseline SCr was defined as the median of SCrs before vancomycin was given. AKIs at start of therapy were defined as a baseline SCr or any SCr collected < 12 hours into therapy that was > 1.5 times higher than any subsequent SCrs. A priori accuracy was calculated by measuring root mean square error (RMSE) of 7 PK models for patients with and without AKIs at start of therapy. Statistical difference between RMSE values was determined by non-overlapping 95% confidence intervals. Results A total of 149,100 adult and 4,311 pediatric patient treatment courses were included (see Table 1). Higher RMSE, or lower accuracy, was seen in the AKI groups for all models evaluated, and all except one were statistically different (see Table 2). The RMSE in the group without an AKI at baseline ranged from 4.27 - 6.70 mg/L whereas in the AKI group it ranged from 4.58 - 7.31 mg/L. The absolute difference between groups for each model ranged from 0.03 - 2.05 mg/L and the percent difference ranged from 0.45-39%. Conclusion PK models had reduced accuracy in patients with an AKI at start of therapy compared to those that do not. However, differences were small and statistical differences were likely detected due to large sample sizes. For adults, the Colin and Thomson models showed small (< 10%) reductions in accuracy in patients with an AKI, while for pediatric patients, the Kloprogge model showed equal accuracy in all patients regardless of AKI status. These findings may help pharmacists decide whether to use PK models to inform initial vancomycin dosing. Disclosures Maria-Stephanie Hughes, PharmD, InsightRX: Stocks/Bonds Jasmine Hughes, PhD, InsightRX: Employee

Utilizing Multivariate Adaptive Regression Splines (MARS) for Precise Estimation of Soil Compaction Parameters

Traditional laboratory methods for estimating soil compaction parameters, such as the Proctor test, have been recognized as time-consuming and labor-intensive. Given the increasing need for the rapid and accurate estimation of soil compaction parameters for a range of geotechnical applications, the application of machine learning models offers a promising alternative. This study focuses on employing the multivariate adaptive regression splines (MARS) model algorithm, a machine learning method that presents a significant advantage over other models through generating human-understandable piecewise linear equations. The MARS model was trained and tested on a comprehensive dataset to predict essential soil compaction parameters, including optimum water content (wopt) and maximum dry density (ρdmax). The performance of the model was evaluated using coefficient of determination (R2) and root mean square error (RMSE) values. Remarkably, the MARS models showed excellent predictive ability with high R2 and low RMSE, MAE, and relative error values, indicating its robustness and reliability in predicting soil compaction parameters. Through rigorous five-fold cross-validation, the model’s predictions for wopt returned an RMSE of 1.948%, an R2 of 0.893, and an MAE of 1.498%. For ρdmax, the results showcased an RMSE of 0.064 Mg/m3, an R2 of 0.899, and an MAE of 0.050 Mg/m3. When evaluated on unseen data, the model’s performance for wopt prediction was marked with an MAE of 1.276%, RMSE of 1.577%, and R2 of 0.948. Similarly, for ρdmax, the predictions were characterized by an MAE of 0.047 Mg/m3, RMSE of 0.062 Mg/m3, and R2 of 0.919. The results also indicated that the MARS model outperformed previously developed machine learning models, suggesting its potential to replace conventional testing methods. The successful application of the MARS model could revolutionize the geotechnical field through providing quick and reliable predictions of soil compaction parameters, improving efficiency for construction projects. Lastly, a variable importance analysis was performed on the model to assess how input variables affect its outcomes. It was found that fine content (Cf) and plastic limit (PL) have the greatest impact on compaction parameters.

Forecasting hourly PM2.5 concentration with an optimized LSTM model

Machine learning has become a powerful tool in air quality assessment which can provide timely and predictable information, alert the public, and take timely measures to prevent deteriorating air quality. The study proposed a deep learning-based long-short term memory (LSTM) model to predict hourly PM2.5 in one of the most polluted areas in Taiwan. A series of sensitivity assessments with model settings was conducted to optimize the performance of the LSTM model. Regarding the model input parameters, aerosol optical depth, pressure, and PM2.5 concentrations from the three nearby stations were used and later showed significant improvement in the forecast results. As a result of the 1–24 h forecast in 2021, the root-mean-square error (RMSE) shows a range from 6.3 to 13.1 μg m−3, and the Pearson correlation coefficient (r) varies from 0.92 to 0.59, as compared with the observed PM2.5. The model's predictability decreases as time increases—a strong correlation (r higher than 0.7) within a 9-h PM2.5 forecast. The seasonal variation showed that the highest RMSE, about 16.2 μg m−3, was observed during the winter, which is the high-polluted season in the area. Additionally, the spatial representation of the model was examined. The model can perform an efficient and satisfied forecast in the radius of 15 km from the training station. We further compared several deep learning-based algorithms in forecasting PM2.5, and our model performs better prediction results. The deep learning–based model investigated in this study can be implemented for routine air quality monitoring in urban areas and air-quality alarms associated with public health.

Assessing the Effectiveness of the Use of the InVEST Annual Water Yield Model for the Rivers of Colombia: A Case Study of the Meta River Basin

This paper presents the results of one of the hydrological models, the InVEST “Annual Water Yield” (InVEST–AWY), applied to the Meta River basin in Colombia, which covers an area of 113,981 km². The study evaluates the performance of the model in different subbasins of the Meta River basin. The model’s accuracy was assessed using different statistical measures, including Nash–Sutcliffe Efficiency (NSE) coefficient, Root Mean Square Error (RMSE), correlation coefficients for the calibration (rcal) and validation (rval) periods. The overall performance of the model in the Meta River basin is relatively poor as indicated by the low NSE value of 0.07 and high RMSE value of 1071.61. In addition, the model explains only a 7% of the variance in the observed data. The sensitivity analysis revealed that a 30% reduction in crop coefficient (Kc) values would result in a 10.7% decrease in water yield. The model estimated, for example, the annual average water yield of the river in 2018 as 1.98 × 1011 m3/year or 6273.4 m3/s, which is 1.3% lower than the reported value. The upper Meta River subbasin shows the highest NSE value (0.49), indicating a good result between observed and simulated water discharge. In contrast, the South Cravo River subbasin shows a negative NSE value of −1.29, indicating poor model performance. The Yucao River subbasin and the upper Casanare River subbasin also show lower NSE values compared to the upper Meta River subbasin, indicating less accurate model performance in these subbasins. The correlation coefficients in calibration (rcal) and validation (rval) for the upper Meta River, Yucao River, South Cravo River, and upper Casanare River subbasins were 0.79 and 0.83, 0.4 and 0.22, 0.5 and −0.25, and 0 and 0.18, respectively. These results provide useful insights into the limitations for the proper use of the InVEST–AWY model in Colombia. This study is the first to use the InVEST–AWY model on a large scale in the territory of Colombia, allowing to evaluate its effectiveness in hydrological modeling for water management.

Effects of Climate on Stand-Level Biomass for Larch Plantations in Heilongjiang Province, Northeast China

Climate change affects forest resource availability, growing season length, and thus forest biomass accumulation. However, only a limited number of studies have been conducted on forest biomass management based on climate effects, particularly at the stand-level. Thus, an allometric biomass equation based on conventional and climate-based stand biomass models, was developed and compared for larch trees (Larix spp.). A total of 160 experimental plots of larch plantations have been collected in Heilongjiang Province, Northeast China. In this study, we developed four types of additive model systems for stand-level biomass: two types of the stand-level biomass basic models (M-1 and M-2) with stand variables (stand basal area (BA) and stand mean height (Hm)) as the predictors, and two types of the proposed stand-level biomass climate-based models (M-3 and M-4) with stand variables (BA and Hm) and climatic variables (mean annual temperature (MAT) and annual precipitation (AP)) as the predictors. Accordingly, this study evaluated the effects of climatic variables (MAT and AP) and stand variables (BA and Hm) on the model’s performance. Model fitting and validation results revealed that the climatic variables significantly improved the model performance of the fitted equation by increasing the coefficient of determination (R2) values and reducing the root mean square error (RMSE) values. A higher R2 and a lower RMSE were consistently generated by M-2 and M-4, whereas M-1 and M-3 consistently generated a lower R2 and a higher RMSE. We found that the proposed stand-level biomass climate-based model type 4 (M-4) performed better than the other models and slightly better than in previous studies of climate-sensitive models. This study provided an additional and beneficial method of analyzing climate effects on stand-level biomass estimation.

Intraocular Lens Power Calculations in Keratoconus Eyes Comparing Keratometry, Total Keratometry, and Newer Formulae

To compare the utility of keratometry vs total keratometry (TK) for intraocular lens power calculations in eyes with keratoconus (KCN) using KCN and non-KCN formulae. Retrospective cohort study. This study was conducted at 2 academic centers and included 87 eyes in 67 patients who underwent cataract surgery between 2019 and 2021. Biometry measurements were obtained using a swept-source optical coherence tomography biometer (IOL Master 700). Refractive prediction errors, including root mean square error (RMSE), were calculated for 13 formulae. These included 4 classical formulae (Haigis, Hoffer Q, Holladay 1 [H1], and SRK/T), 5 new formulae (NF) (Barrett Universal II [BU2], Cooke K6, EVO 2.0, Kane, and Pearl-DGS), 3 KCN formulae (BU2 KCN: M-PCA, BU2 KCN: P-PCA, and Kane KCN), and H1 with equivalent keratometry reading values (H1-EKR). Formulae were ranked by RMSE. Friedman analysis of variance with post hoc analysis and H-testing was used for statistical significance testing. KCN formulae had the lowest RMSEs in all eyes, and BU2 KCN:M-PCA performed the best among KCN formulae in all subgroups. In eyes with severe KCN, if TK values are unavailable, the BU2 KCN: P-PCA performed better than the top-ranked non-KCN formula (SRK/T). In eyes with nonsevere KCN, if TK values are unavailable, EVO 2.0 K was statistically superior to the next competitor (Kane K). H1-EKR had the highest RMSE. KCN formulae and TK are useful for intraocular lens power calculations in KCN eyes, especially in eyes with severe KCN. The BU2 KCN: M-PCA using TK values performed best for eyes with all severities of KCN. For eyes with nonsevere KCN, the EVO 2.0 TK or K can also be used.

TOC prediction using a gradient boosting decision tree method: A case study of shale reservoirs in Qinshui Basin

Shale oil and shale gas are important unconventional resources. As the total organic carbon (TOC) of shales is related to the generation potentiality of hydrocarbons, accurate TOC prediction models can improve development efficiency and reduce cost. However, there has been little work on intelligent prediction of TOC. This study focused on hydrocarbon-rich shale from the Linfen area of Qinshui Basin in China. Multiple regression analysis (MRA), ΔlgR, back propagation (BP) neural network, gradient boosting decision tree (GBDT) were selected to predict the TOC of the studied shale reservoirs using density (DEN), uranium (U), acoustic (AC), and resistivity (RT) logs. The predicted TOC was compared with the geochemical-measured TOC. To fairly evaluate these models, the correlation of determination (R2), root-mean-square-error (RMSE) and three-scale analytic hierarchy process (TAHP) were introduced. The results show that MRA, ΔlgR, and BP methods gave low R2, high RMSE values, and low total weight, indicating these methods poorly predicted TOC. These models significantly underestimate the TOC, resulting in evident deviations between the predicted TOC and their real values. In contrast, GBDT provided accurate TOC prediction, with values of R2, RMSE, and total weight of 0.9254, 0.032 and 0.42. The GBDT model can greatly outperform other models by extracting complex relationships between well log data and TOC values. Sensitivity analysis of the GBDT model revealed that the U parameter exhibited the greatest impact on TOC prediction. The results of this work show that the GBDT model is well suited for TOC prediction of unconventional resources.

Comparison of the Applicability of Two Reanalysis Products in Estimating Tall Tower Wind Based on Multiple Linear Regression and Artificial Neural Network in South China

Climate reanalysis products have been widely used to overcome the absence of high-quality and long-term observational records for wind energy users. In this study, the applicability of two popular reanalysis datasets (ERA5 and MERRA2) in estimating wind characteristics for four tall tower observatories (TTOs) in South China was assessed. For each TTO, linear and nonlinear downscaling techniques, namely, multiple linear regression (MLR) and an artificial neural network (ANN), respectively, were adopted for the downscaling of the scalar wind speed and the corresponding U/V components. The downscaled wind speed and U/V components were subsequently compared with the TTO observations by correlation coefficient (Pearson’s r), the root mean square error (RMSE), the uncertainty analysis (U95), and the reliability analysis (RE). According to the results, ERA5 had a better applicability (higher Pearson’s r and RE, but lower RMSE and U95) in estimating TTO wind speed than MERRA2 when using both the MLR and ANN downscaling method. The average Pearson’s r, RE, RMSE, and U95 of the downscaled wind from ERA5 by the MLR (ANN) method were 0.66 (0.69), 40.8% (41.8%), 2.20 m/s (2.11 m/s), 0.181 m/s (0.179 m/s), respectively, and 0.60 (0.63), 38.0% (39.7%), 2.32 m/s (2.25 m/s), 0.189 m/s (0.187 m/s), respectively, for MERRA2. The wind components analysis showed that the better performance of ERA5 was attributed to its smaller error in estimating V component than MERRA2. For the wind direction, the two reanalysis datasets did not display distinct differences. Additionally, the misalignment of the wind direction between the reanalysis products and the TTOs was higher for the secondary predominant wind direction (SPWD) than for the predominant wind direction (PWD). Furthermore, we found that the reanalysis U wind had a higher RMSE but a lower RE and Pearson’s r than the V wind, which indicates that the misalignment in the wind direction was mainly associated with the deviation in the U component.