The use of classification and regression algorithms using the random forests method with presence-only data to model species' distribution

Assessing the global distribution and risk of harmful microalgae: A focus on three toxic Alexandrium dinoflagellates

Harmful dinoflagellates and their resulting blooms pose a threat to marine life and human health. However, to date, global maps of marine life often overlook harmful microorganisms. As harmful algal blooms (HABs) increase in frequency, severity, and extent, understanding the distribution of harmful dinoflagellates and their drivers is crucial for their management. We used MaxEnt, random forest, and ensemble models to map the habitats of the representative HABs species in the genus Alexandrium, including A. catenella, A. minutum, and A. pacificum. Since species occurrence records used in previous studies were solely morphology-based, potentially leading to misidentifications, we corrected these species' distribution records using molecular criteria. The results showed that the key environmental drivers included the distance to the coastline, bathymetry, sea surface temperature (SST), and dissolved oxygen. Alexandrium catenella thrives in temperate to cold zones and is driven by low SST and high oxygen levels. Alexandrium pacificum mainly inhabits the Temperate Northern Pacific and prefers warmer SST and lower oxygen levels. Alexandrium minutum thrives universally and adapts widely to SST and oxygen. By analyzing the habitat suitability of locations with recorded HAB occurrences, we found that high habitat suitability could serve as a reference indicator for bloom risk. Therefore, we have proposed a qualitative method to spatially assess the harmful algae risk according to the habitat suitability. On the global risk map, coastal temperate seas, such as the Mediterranean, Northwest Pacific, and Southern Australia, faced higher risks. Although HABs currently have restricted geographic distributions, our study found these harmful algae possess high environmental tolerance and can thrive across diverse habitats. HAB impacts could increase if climate changes or ocean conditions became more favorable. Marine transportation may also spread the harmful algae to new unaffected ecosystems. This study has pioneered the assessment of harmful algal risk based on habitat suitability.

From plan to delivery: Machine learning based positional accuracy prediction of multi-leaf collimator and estimation of delivery effect in volumetric modulated arc therapy

PURPOSE

The positional accuracy of MLC is an important element in establishing the exact dosimetry in VMAT. We comprehensively analyzed factors that may affect MLC positional accuracy in VMAT, and constructed a model to predict MLC positional deviation and estimate planning delivery quality according to the VMAT plans before delivery.

METHODS

A total of 744 "dynalog" files for 23 VMAT plans were extracted randomly from treatment database. Multi-correlation was used to analyzed the potential influences on MLC positional accuracy, including the spatial characteristics and temporal variability of VMAT fluence, and the mechanical wear parameters of MLC. We developed a model to forecast the accuracy of MLC moving position utilizing the random forest (RF) ensemble learning method. Spearman correlation was used to further investigate the associations between MLC positional deviation and dosage deviations as well as gamma passing rates.

RESULTS

The MLC positional deviation and effective impact factors show a strong multi-correlation (R = 0.701, p-value < 0.05). This leads to the development of a highly accurate prediction model with average variables explained of 95.03% and average MSE of 0.059 in the 5-fold cross-validation, and MSE of 0.074 for the test data was obtained. The absolute dose deviations caused by MLC positional deviation ranging from 12.948 to 210.235 cGy, while the relative volume deviation remained small at 0.470%-5.161%. The average MLC positional deviation correlated substantially with gamma passing rates (with correlation coefficient of -0.506 to -0.720 and p-value < 0.05) but marginally with dosage deviations (with correlation coefficient < 0.498 and p-value > 0.05).

CONCLUSIONS

The RF predictive model provides a prior tool for VMAT quality assurance.

The use of an integrated framework combining eco-evolutionary data and species distribution models to predict range shifts of species under changing climates

Species distribution models (SDMs) are powerful tools that can predict potential distributions of species under climate change. However, traditional SDMs that rely on current species occurrences may underestimate their climatic tolerances and potential distributions. To address this limitation, we developed an integrated framework that incorporates eco-evolutionary data into SDMs. In our approach, the fundamental niches of species are constructed by their realized niches in different periods, and those fundamental niches are used to predict potential distributions of species. Our framework includes multiple phylogenetic analyses, such as niche evolution rate estimation and ancestral area reconstruction. These analyses provide deeper insights into the responses of species to climate change. We applied our approach to the Chrysanthemum zawadskii species complex to evaluate its efficacy through comprehensive performance evaluations and validation tests. Our framework can be applied broadly to species with available phylogenetic data and occurrence records, making it a valuable tool for understanding species adaptation in a rapidly changing world.•Integrating the niches of species in different periods estimates more complete climatic envelopes for them.•Combining eco-evolutionary data with SDMs predicts more comprehensive potential distributions of species under climate change.•Our framework provides a general procedure for species with phylogenetic data and occurrence records.

Combining environmental DNA and visual surveys can inform conservation planning for coral reefs

Environmental DNA (eDNA) metabarcoding has the potential to revolutionize conservation planning by providing spatially and taxonomically comprehensive data on biodiversity and ecosystem conditions, but its utility to inform the design of protected areas remains untested. Here, we quantify whether and how identifying conservation priority areas within coral reef ecosystems differs when biodiversity information is collected via eDNA analyses or traditional visual census records. We focus on 147 coral reefs in Indonesia's hyper-diverse Wallacea region and show large discrepancies in the allocation and spatial design of conservation priority areas when coral reef species were surveyed with underwater visual techniques (fishes, corals, and algae) or eDNA metabarcoding (eukaryotes and metazoans). Specifically, incidental protection occurred for 55% of eDNA species when targets were set for species detected by visual surveys and 71% vice versa. This finding is supported by generally low overlap in detection between visual census and eDNA methods at species level, with more overlap at higher taxonomic ranks. Incomplete taxonomic reference databases for the highly diverse Wallacea reefs, and the complementary detection of species by the two methods, underscore the current need to combine different biodiversity data sources to maximize species representation in conservation planning.

Distribution and co-occurrence patterns of charophytes and angiosperms in the northern Baltic Sea

The distribution data of 11 soft substrate charophyte and angiosperm species were analyzed. Our study aimed to elucidate the co-occurrence patterns among these sympatric macrophyte species and quantify their distribution areas. The central hypothesis of this study proposed that the observed co-occurrence patterns among the studied species deviate from what would be expected by random chance. Macrophyte occurrence data was derived from an extensive field sampling database. Environmental variables available as georeferenced raster layers including topographical, hydrodynamic, geological, physical, chemical, and biological variables were used as predictor variables in the random forest models to predict the spatial distribution of the species. Permutation tests revealed statistically significant deviations from random co-occurrence patterns. The analysis demonstrated that species tended to co-occur more frequently within their taxonomic groups (i.e., within charophytes and within angiosperms) than between these groups. The most extensive distribution overlap was observed between Chara aspera Willd. and Chara canescens Loisel., while Zostera marina L. exhibited the least overlap with the other species. The mean number of co-occurring species was the highest in Chara baltica (Hartman) Bruzelius while Z. marina had the largest share of single-species occurrences. Based on the distribution models, Stuckenia pectinata (L.) Börner had the largest distribution area.

Sampling from commercial vessel routes can capture marine biodiversity distributions effectively

Collecting fine-scale occurrence data for marine species across large spatial scales is logistically challenging but is important to determine species distributions and for conservation planning. Inaccurate descriptions of species ranges could result in designating protected areas with inappropriate locations or boundaries. Optimizing sampling strategies therefore is a priority for scaling up survey approaches using tools such as environmental DNA (eDNA) to capture species distributions. In a marine context, commercial vessels, such as ferries, could provide sampling platforms allowing access to undersampled areas and repeatable sampling over time to track community changes. However, sample collection from commercial vessels could be biased and may not represent biological and environmental variability. Here, we evaluate whether sampling along Mediterranean ferry routes can yield unbiased biodiversity survey outcomes, based on perfect knowledge from a stacked species distribution model (SSDM) of marine megafauna derived from online data repositories. Simulations to allocate sampling point locations were carried out representing different sampling strategies (random vs regular), frames (ferry routes vs unconstrained), and number of sampling points. SSDMs were remade from different sampling simulations and compared with the "perfect knowledge" SSDM to quantify the bias associated with different sampling strategies. Ferry routes detected more species and were able to recover known patterns in species richness at smaller sample sizes better than unconstrained sampling points. However, to minimize potential bias, ferry routes should be chosen to cover the variability in species composition and its environmental predictors in the SSDMs. The workflow presented here can be used to design effective sampling strategies using commercial vessel routes globally for eDNA and other biodiversity survey techniques. This approach has potential to provide a cost-effective method to access remote oceanic areas on a regular basis and can recover meaningful data on spatiotemporal biodiversity patterns.

Comparison between optimized MaxEnt and random forest modeling in predicting potential distribution: A case study with Quasipaa boulengeri in China

Random forest (RF) and MaxEnt models are shallow machine learning approaches that perform well in predicting species' potential distributions. RF models can produce robust results with the default automatic configuration in most cases, but it is necessary for MaxEnt to optimize the model settings to improve the performance, and the predictive performance difference between optimized MaxEnt and RF is uncertain. To explore this issue, the potential distribution of the endangered amphibian Quasipaa boulengeri in China was predicted using optimized MaxEnt and RF models. A total of 408 occurrence data were selected, 1000 locations were generated as pseudo-absence data by the geographic distance method, and 10,000 sites were selected as background data by creating a bias file. Partial ROC at different thresholds and success rate curves were used to compare the predictive performances between optimized MaxEnt and RF. Our results showed that the RF and optimized MaxEnt models both had good performance in predicting the potential distribution of Q. boulengeri, with the RF model performing slightly better whether based on partial ROC or success rate curves. Furthermore, the core suitable habitat regions of Q. boulengeri identified by RF and MaxEnt were similar and were all located in the Sichuan, Chongqing, Hubei, Hunan, and Guizhou provinces. However, the RF model produced a habitat suitability map with higher discrimination and greater heterogeneity. Temperature annual range, mean temperature of the driest quarter, and annual precipitation were the vital environmental variables limiting the distribution of Q. boulengeri. The RF model is the stronger machine learner. We believe it may be more applicable in predicting the native potential distributions of species with sufficient occurrence data, given the additional predictive detail, the simplicity of use, the computational time involved, and the operational complexity.

Screening the Influence of Biomarkers for Metabolic Syndrome in Occupational Population Based on the Lasso Algorithm

Aim: Metabolic syndrome (MS) screening is essential for the early detection of the occupational population. This study aimed to screen out biomarkers related to MS and establish a risk assessment and prediction model for the routine physical examination of an occupational population.

Methods: The least absolute shrinkage and selection operator (Lasso) regression algorithm of machine learning was used to screen biomarkers related to MS. Then, the accuracy of the logistic regression model was further verified based on the Lasso regression algorithm. The areas under the receiving operating characteristic curves were used to evaluate the selection accuracy of biomarkers in identifying MS subjects with risk. The screened biomarkers were used to establish a logistic regression model and calculate the odds ratio (OR) of the corresponding biomarkers. A nomogram risk prediction model was established based on the selected biomarkers, and the consistency index (C-index) and calibration curve were derived.

Results: A total of 2,844 occupational workers were included, and 10 biomarkers related to MS were screened. The number of non-MS cases was 2,189 and that of MS was 655. The area under the curve (AUC) value for non-Lasso and Lasso logistic regression was 0.652 and 0.907, respectively. The established risk assessment model revealed that the main risk biomarkers were absolute basophil count (OR: 3.38, CI:1.05–6.85), platelet packed volume (OR: 2.63, CI:2.31–3.79), leukocyte count (OR: 2.01, CI:1.79–2.19), red blood cell count (OR: 1.99, CI:1.80–2.71), and alanine aminotransferase level (OR: 1.53, CI:1.12–1.98). Furthermore, favorable results with C-indexes (0.840) and calibration curves closer to ideal curves indicated the accurate predictive ability of this nomogram.

Conclusions: The risk assessment model based on the Lasso logistic regression algorithm helped identify MS with high accuracy in physically examining an occupational population.

Ensemble ecological niche modeling of West Nile virus probability in Florida

Ecological Niche Modeling is a process by which spatiotemporal, climatic, and environmental data are analyzed to predict the distribution of an organism. Using this process, an ensemble ecological niche model for West Nile virus habitat prediction in the state of Florida was developed. This model was created through the weighted averaging of three separate machine learning models—boosted regression tree, random forest, and maximum entropy—developed for this study using sentinel chicken surveillance and remote sensing data. Variable importance differed among the models. The highest variable permutation value included mean dewpoint temperature for the boosted regression tree model, mean temperature for the random forest model, and wetlands focal statistics for the maximum entropy mode. Model validation resulted in area under the receiver curve predictive values ranging from good [0.8728 (95% CI 0.8422–0.8986)] for the maximum entropy model to excellent [0.9996 (95% CI 0.9988–1.0000)] for random forest model, with the ensemble model predictive value also in the excellent range [0.9939 (95% CI 0.9800–0.9979]. This model should allow mosquito control districts to optimize West Nile virus surveillance, improving detection and allowing for a faster, targeted response to reduce West Nile virus transmission potential.

Infestation risk of the intermediate snail host of Schistosoma japonicum in the Yangtze River Basin: improved results by spatial reassessment and a random forest approach

BACKGROUND

Oncomelania hupensis is only intermediate snail host of Schistosoma japonicum, and distribution of O. hupensis is an important indicator for the surveillance of schistosomiasis. This study explored the feasibility of a random forest algorithm weighted by spatial distance for risk prediction of schistosomiasis distribution in the Yangtze River Basin in China, with the aim to produce an improved precision reference for the national schistosomiasis control programme by reducing the number of snail survey sites without losing predictive accuracy.

METHODS

The snail presence and absence records were collected from Anhui, Hunan, Hubei, Jiangxi and Jiangsu provinces in 2018. A machine learning of random forest algorithm based on a set of environmental and climatic variables was developed to predict the breeding sites of the O. hupensis intermediated snail host of S. japonicum. Different spatial sizes of a hexagonal grid system were compared to estimate the need for required snail sampling sites. The predictive accuracy related to geographic distances between snail sampling sites was estimated by calculating Kappa and the area under the curve (AUC).

RESULTS

The highest accuracy (AUC = 0.889 and Kappa = 0.618) was achieved at the 5 km distance weight. The five factors with the strongest correlation to O. hupensis infestation probability were: (1) distance to lake (48.9%), (2) distance to river (36.6%), (3) isothermality (29.5%), (4) mean daily difference in temperature (28.1%), and (5) altitude (26.0%). The risk map showed that areas characterized by snail infestation were mainly located along the Yangtze River, with the highest probability in the dividing, slow-flowing river arms in the middle and lower reaches of the Yangtze River in Anhui, followed by areas near the shores of China's two main lakes, the Dongting Lake in Hunan and Hubei and the Poyang Lake in Jiangxi.

CONCLUSIONS

Applying the machine learning of random forest algorithm made it feasible to precisely predict snail infestation probability, an approach that could improve the sensitivity of the Chinese schistosome surveillance system. Redesign of the snail surveillance system by spatial bias correction of O. hupensis infestation in the Yangtze River Basin to reduce the number of sites required to investigate from 2369 to 1747.

Estimation of nitrogen and phosphorus concentrations from water quality surrogates using machine learning in the Tri An Reservoir, Vietnam

Surface water eutrophication due to excessive nutrients has become a major environmental problem around the world in the past few decades. Among these nutrients, nitrogen and phosphorus are two of the most important harmful cyanobacterial bloom (HCB) drivers. A reliable prediction of these parameters, therefore, is necessary for the management of rivers, lakes, and reservoirs. The aim of this study is to test the suitability of the powerful machine learning (ML) algorithm, random forest (RF), to provide information on water quality parameters for the Tri An Reservoir (TAR). Three species of nitrogen and phosphorus, including nitrite (N-NO₂^-), nitrate (N-NO₃^-), and phosphate (P-PO₄^3-), were empirically estimated using the field observation dataset (2009-2014) of six surrogates of total suspended solids (TSS), total dissolved solids (TDS), turbidity, electrical conductivity (EC), chemical oxygen demand (COD), and biochemical oxygen demand (BOD₅). Field data measurement showed that water quality in the TAR was eutrophic with an up-trend of N-NO₃^- and P-PO₄^3- during the study period. The RF regression model was reliable for N-NO₂^-, N-NO₃^-, and P-PO₄^3- prediction with a high R² of 0.812-0.844 for the training phase (2009-2012) and 0.888-0.903 for the validation phase (2013-2014). The results of land use and land cover change (LUCC) revealed that deforestation and shifting agriculture in the upper region of the basin were the major factors increasing nutrient loading in the TAR. Among the meteorological parameters, rainfall pattern was found to be one of the most influential factors in eutrophication, followed by average sunshine hour. Our results are expected to provide an advanced assessment tool for predicting nutrient loading and for giving an early warning of HCB in the TAR.