Automated Cow Body Condition Scoring Using Multiple 3D Cameras and Convolutional Neural Networks.

Body condition scoring is an objective scoring method used to evaluate the health of a cow by determining the amount of subcutaneous fat in a cow. Automated body condition scoring is becoming vital to large commercial dairy farms as it helps farmers score their cows more often and more consistently compared to manual scoring. A common approach to automated body condition scoring is to utilise a CNN-based model trained with data from a depth camera. The approaches presented in this paper make use of three depth cameras placed at different positions near the rear of a cow to train three independent CNNs. Ensemble modelling is used to combine the estimations of the three individual CNN models. The paper aims to test the performance impact of using ensemble modelling with the data from three separate depth cameras. The paper also looks at which of these three cameras and combinations thereof provide a good balance between computational cost and performance. The results of this study show that utilising the data from three depth cameras to train three separate models merged through ensemble modelling yields significantly improved automated body condition scoring accuracy compared to a single-depth camera and CNN model approach. This paper also explored the real-world performance of these models on embedded platforms by comparing the computational cost to the performance of the various models.

Predicting potential distribution and range dynamics of Aquilegia fragrans under climate change: insights from ensemble species distribution modelling.

Climate change is one of the primary causes of species redistribution and biodiversity loss, especially for threatened and endemic important plant species. Therefore, it is vital to comprehend "how" and "where" priority medicinal and aromatic plants (MAPs) might be effectively used to address conservation-related issues under rapid climate change. In the present study, an ensemble modelling approach was used to investigate the present and future distribution patterns of Aquilegia fragrans Benth. under climate change in the entire spectrum of Himalayan biodiversity hotspot. The results of the current study revealed that, under current climatic conditions, the northwest states of India (Jammu and Kashmir, Himachal Pradesh and the northern part of Uttarakhand), the eastern and southern parts of Pakistan Himalaya have highly suitable climatic conditions for the growth of A. fragrans. The ensemble model exhibited high forecast accuracy, with temperature seasonality and precipitation seasonality as the main climatic variables responsible for the distribution of the A. fragrans in the biodiversity hotspot. Furthermore, the study predicted that future climate change scenarios will diminish habitat suitability for the species by -46.9% under RCP4.5 2050 and -55.0% under RCP4.5 2070. Likewise, under RCP8.5, the habitat suitability will decrease by -51.7% in 2050 and -94.3% in 2070. The current study also revealed that the western Himalayan area will show the most habitat loss. Some currently unsuitable regions, such as the northern Himalayan regions of Pakistan, will become more suitable under climate change scenarios. Hopefully, the current approach may provide a robust technique and showcases a model with learnings for predicting cultivation hotspots and developing scientifically sound conservation plans for this endangered medicinal plant in the Himalayan biodiversity hotspot.

Decline in the suitable habitat of dominant Abies species in response to climate change in the Hindu Kush Himalayan region: insights from species distribution modelling.

Reliable predictions of future distribution ranges of ecologically important species in response to climate change are required for developing effective management strategies. Here we used an ensemble modelling approach to predict the distribution of three important species of Abies namely, Abies pindrow, Abies spectabilis and Abies densa in the Hindu Kush Himalayan region under the current and two shared socioeconomic pathways (SSP245 and SSP585) and time periods of 2050 and 2090s. A correlative ensemble model using presence/absence data of the three Abies species and 22 environmental variables, including 19 bioclimatic variables and 3 topographic variables, from known distributions was built to predict the potential current and future distribution of these species. The individual models used to build the final ensemble performed well and provided reliable results for both the current and future distribution of all three species. For A. pindrow, precipitation of the driest month (Bio14) was the most important environmental variable with 83.3% contribution to model output while temperature seasonality (Bio4) and annual mean diurnal range (Bio2) were the most important variables for A. spectabilis and A. densa with 48.4% and 46.1% contribution to final model output, respectively. Under current climatic conditions, the ensemble models projected a total suitable habitat of about 433,003 km2, 790,837 km2 and 676,918 km2 for A. pindrow, A. spectabilis and A. densa, respectively, which is approximately 10.36%, 18.91% and 16.91% of the total area of Hindu Kush Himalayan region. Projections of habitat suitability under future climate scenarios for all the shared socioeconomic pathways showed a reduction in potentially suitable habitats with a maximum overall loss of approximately 14% of the total suitable area of A. pindrow under SSP 8.5 by 2090. A decline in total suitable habitat is predicted to be 9.6% in A. spectabilis by 2090 under the SSP585 scenario while in A. densa 6.67% loss in the suitable area is expected by 2050 under the SSP585 scenario. Furthermore, there is no elevational change predicted in the case of A. pindrow while A. spectabilis is expected to show an upward shift by about 29 m per decade and A. densa is showing a downward shift at a rate of 11 m per decade. The results are interesting, and intriguing given the occurrence of these species across the Hindu Kush Himalayan region. Thus, our study underscores the need for consideration of unexpected responses of species to climate change and formulation of strategies for better forest management and conservation of important conifer species, such as A. pindrow, A. spectabilis and A. densa.

Spatial variation in lymphatic filariasis risk factors of hotspot zones in Ghana.

BACKGROUND: Lymphatic Filariasis (LF), a parasitic nematode infection, poses a huge economic burden to affected countries. LF endemicity is localized and its prevalence is spatially heterogeneous. In Ghana, there exists differences in LF prevalence and multiplicity of symptoms in the country's northern and southern parts. Species distribution models (SDMs) have been utilized to explore the suite of risk factors that influence the transmission of LF in these geographically distinct regions. METHODS: Presence-absence records of microfilaria (mf) cases were stratified into northern and southern zones and used to run SDMs, while climate, socioeconomic, and land cover variables provided explanatory information. Generalized Linear Model (GLM), Generalized Boosted Model (GBM), Artificial Neural Network (ANN), Surface Range Envelope (SRE), Multivariate Adaptive Regression Splines (MARS), and Random Forests (RF) algorithms were run for both study zones and also for the entire country for comparison. RESULTS: Best model quality was obtained with RF and GBM algorithms with the highest Area under the Curve (AUC) of 0.98 and 0.95, respectively. The models predicted high suitable environments for LF transmission in the short grass savanna (northern) and coastal (southern) areas of Ghana. Mainly, land cover and socioeconomic variables such as proximity to inland water bodies and population density uniquely influenced LF transmission in the south. At the same time, poor housing was a distinctive risk factor in the north. Precipitation, temperature, slope, and poverty were common risk factors but with subtle variations in response values, which were confirmed by the countrywide model. CONCLUSIONS: This study has demonstrated that different variable combinations influence the occurrence of lymphatic filariasis in northern and southern Ghana. Thus, an understanding of the geographic distinctness in risk factors is required to inform on the development of area-specific transmission control systems towards LF elimination in Ghana and internationally.

Modelling climate change impacts on maize yields under low nitrogen input conditions in sub-Saharan Africa.

Smallholder farmers in sub-Saharan Africa (SSA) currently grow rainfed maize with limited inputs including fertilizer. Climate change may exacerbate current production constraints. Crop models can help quantify the potential impact of climate change on maize yields, but a comprehensive multimodel assessment of simulation accuracy and uncertainty in these low-input systems is currently lacking. We evaluated the impact of varying [CO2 ], temperature and rainfall conditions on maize yield, for different nitrogen (N) inputs (0, 80, 160 kg N/ha) for five environments in SSA, including cool subhumid Ethiopia, cool semi-arid Rwanda, hot subhumid Ghana and hot semi-arid Mali and Benin using an ensemble of 25 maize models. Models were calibrated with measured grain yield, plant biomass, plant N, leaf area index, harvest index and in-season soil water content from 2-year experiments in each country to assess their ability to simulate observed yield. Simulated responses to climate change factors were explored and compared between models. Calibrated models reproduced measured grain yield variations well with average relative root mean square error of 26%, although uncertainty in model prediction was substantial (CV = 28%). Model ensembles gave greater accuracy than any model taken at random. Nitrogen fertilization controlled the response to variations in [CO2 ], temperature and rainfall. Without N fertilizer input, maize (a) benefited less from an increase in atmospheric [CO2 ]; (b) was less affected by higher temperature or decreasing rainfall; and (c) was more affected by increased rainfall because N leaching was more critical. The model intercomparison revealed that simulation of daily soil N supply and N leaching plays a crucial role in simulating climate change impacts for low-input systems. Climate change and N input interactions have strong implications for the design of robust adaptation approaches across SSA, because the impact of climate change in low input systems will be modified if farmers intensify maize production with balanced nutrient management.

Artificial intelligence based ensemble model for prediction of vehicular traffic noise.

Vehicular traffic noise is the main source of noise pollution in major cities around the globe. A reliable and accurate method for the estimation of vehicular traffic noise is therefore essential for creating a healthy noise-free environment. In this study, 2 linear (simple average and weighted average) and 2-nonlinear (neural network and neuro-fuzzy) ensemble models were developed by combining the outputs of three Artificial Intelligence (AI) based non-linear models; Adaptive Neuro Fuzzy Inference System (ANFIS), Feed Forward Neural Network (FFNN), Support Vector Regression (SVR) and one Multilinear regression (MLR) model to enhance the performance of the single black box models in predicting vehicular traffic noise of Nicosia city, North Cyprus. In this way, first a nonlinear sensitivity analysis was applied to select the most relevant and dominant input parameters of the traffic data obtained from 12 observation points in the study area. The most dominant parameters in order of their importance were determined to be number of cars, number of van/pickups, number of trucks, average speed and number of buses. Classifying the number of vehicles into five categories before feeding the traffic data into the AI models was observed to improve performance of the single models up to 29% in the verification phase. Out of the four ensembles models developed, the nonlinear ANFIS ensemble was found to be the most robust by improving the performance of ANFIS, FFNN, SVR and MLR models in the verification stage by 11%, 19%, 21% and 31%, respectively.

Evaluation of the impacts of climate change on disease vectors through ecological niche modelling.

Vector-borne diseases are exceptionally sensitive to climate change. Predicting vector occurrence in specific regions is a challenge that disease control programs must meet in order to plan and execute control interventions and climate change adaptation measures. Recently, an increasing number of scientific articles have applied ecological niche modelling (ENM) to study medically important insects and ticks. With a myriad of available methods, it is challenging to interpret their results. Here we review the future projections of disease vectors produced by ENM, and assess their trends and limitations. Tropical regions are currently occupied by many vector species; but future projections indicate poleward expansions of suitable climates for their occurrence and, therefore, entomological surveillance must be continuously done in areas projected to become suitable. The most commonly applied methods were the maximum entropy algorithm, generalized linear models, the genetic algorithm for rule set prediction, and discriminant analysis. Lack of consideration of the full-known current distribution of the target species on models with future projections has led to questionable predictions. We conclude that there is no ideal 'gold standard' method to model vector distributions; researchers are encouraged to test different methods for the same data. Such practice is becoming common in the field of ENM, but still lags behind in studies of disease vectors.

Use of crop simulation modelling to aid ideotype design of future cereal cultivars.

A major challenge of the 21st century is to achieve food supply security under a changing climate and roughly a doubling in food demand by 2050 compared to present, the majority of which needs to be met by the cereals wheat, rice, maize, and barley. Future harvests are expected to be especially threatened through increased frequency and severity of extreme events, such as heat waves and drought, that pose particular challenges to plant breeders and crop scientists. Process-based crop models developed for simulating interactions between genotype, environment, and management are widely applied to assess impacts of environmental change on crop yield potentials, phenology, water use, etc. During the last decades, crop simulation has become important for supporting plant breeding, in particular in designing ideotypes, i.e. 'model plants', for different crops and cultivation environments. In this review we (i) examine the main limitations of crop simulation modelling for supporting ideotype breeding, (ii) describe developments in cultivar traits in response to climate variations, and (iii) present examples of how crop simulation has supported evaluation and design of cereal cultivars for future conditions. An early success story for rice demonstrates the potential of crop simulation modelling for ideotype breeding. Combining conventional crop simulation with new breeding methods and genetic modelling holds promise to accelerate delivery of future cereal cultivars for different environments. Robustness of model-aided ideotype design can further be enhanced through continued improvements of simulation models to better capture effects of extremes and the use of multi-model ensembles.

Predicting post-treatment symptom severity for adults receiving psychological therapy in routine care for generalised anxiety disorder: a machine learning approach.

Approximately half of generalised anxiety disorder (GAD) patients do not recover from first-line treatments, and no validated prediction models exist to inform individuals or clinicians of potential treatment benefits. This study aimed to develop and validate an accurate and explainable prediction model of post-treatment GAD symptom severity. Data from adults receiving treatment for GAD in eight Improving Access to Psychological Therapies (IAPT) services (n=15,859) were separated into training, validation and holdout datasets. Thirteen machine learning algorithms were compared using 10-fold cross-validation, against two simple clinically relevant comparison models. The best-performing model was tested on the holdout dataset and model-specific explainability measures identified the most important predictors. A Bayesian Additive Regression Trees model out-performed all comparison models (MSE=16.54 [95 % CI=15.58; 17.51]; MAE=3.19; R²=0.33, including a single predictor linear regression model: MSE=20.70 [95 % CI=19.58; 21.82]; MAE=3.94; R²=0.14). The five most important predictors were: PHQ-9 anhedonia, GAD-7 annoyance/irritability, restlessness and fear items, then the referral-assessment waiting time. The best-performing model accurately predicted post-treatment GAD symptom severity using only pre-treatment data, outperforming comparison models that approximated clinical judgement and remaining within the GAD-7 error of measurement and minimal clinically important differences. This model could inform treatment decision-making and provide desired information to clinicians and patients receiving treatment for GAD.

Climate change impact on sub-tropical lakes - Lake Kinneret as a case study.

Climate change is anticipated to alter lake ecosystems by affecting water quality, potentially resulting in loss of ecosystem services. Subtropical lakes have high temperatures to begin with and are expected to exhibit higher temperatures all year round which might affect the thermal structure and ecological processes in a different manner than lakes in temperate zones. In this study the ecosystem response of the sub-tropical Lake Kinneret to climate change was explored using lake ecosystem models. Projection reliability was increased by using a weather generator and ensemble modelling, confronting uncertainty of both climate projections and lake models. The study included running two 1D hydrodynamic-biogeochemical models over one thousand realizations of two gradual temperature increase scenarios that span over 49 years. Our predictions show that an increase in air temperature would have subtle effects on stratification properties but may result in considerable changes to biogeochemical processes. Water temperature rise would cause a reduction in dissolved oxygen. Both of these changes would produce elevated phosphate and lowered ammonium concentrations. In turn, these changes are predicted to modify the phytoplankton community, expressed chiefly in increased cyanobacteria blooms at the expense of green phytoplankton and dinoflagellates; these changes may culminate in overall reduction of primary production. Identification of these trends would not be possible without the use of many realizations of climate scenarios. The use of ensemble modelling increased prediction reliability and highlighted elements of uncertainty. Though we use Lake Kinneret, the patterns identified most likely indicate processes that are expected in sub-tropical lakes in general.

Combining models to generate consensus medium-term projections of hospital admissions, occupancy and deaths relating to COVID-19 in England.

Mathematical modelling has played an important role in offering informed advice during the COVID-19 pandemic. In England, a cross government and academia collaboration generated medium-term projections (MTPs) of possible epidemic trajectories over the future 4-6 weeks from a collection of epidemiological models. In this article, we outline this collaborative modelling approach and evaluate the accuracy of the combined and individual model projections against the data over the period November 2021-December 2022 when various Omicron subvariants were spreading across England. Using a number of statistical methods, we quantify the predictive performance of the model projections for both the combined and individual MTPs, by evaluating the point and probabilistic accuracy. Our results illustrate that the combined MTPs, produced from an ensemble of heterogeneous epidemiological models, were a closer fit to the data than the individual models during the periods of epidemic growth or decline, with the 90% confidence intervals widest around the epidemic peaks. We also show that the combined MTPs increase the robustness and reduce the biases associated with a single model projection. Learning from our experience of ensemble modelling during the COVID-19 epidemic, our findings highlight the importance of developing cross-institutional multi-model infectious disease hubs for future outbreak control.

Unravelling the impact of climate change on honey bees: An ensemble modelling approach to predict shifts in habitat suitability in Queensland, Australia.

Honey bees play a vital role in providing essential ecosystem services and contributing to global agriculture. However, the potential effect of climate change on honey bee distribution is still not well understood. This study aims to identify the most influential bioclimatic and environmental variables, assess their impact on honey bee distribution, and predict future distribution. An ensemble modelling approach using the biomod2 package in R was employed to develop three models: a climate-only model, an environment-only model, and a combined climate and environment model. By utilising bioclimatic data (radiation of the wettest and driest quarters and temperature seasonality) from 1990 to 2009, combined with observed honey bee presence and pseudo absence data, this model predicted suitable locations for honey bee apiaries for two future time spans: 2020-2039 and 2060-2079. The climate-only model exhibited a true skill statistic (TSS) value of 0.85, underscoring the pivotal role of radiation and temperature seasonality in shaping honey bee distribution. The environment-only model, incorporating proximity to floral resources, foliage projective cover, and elevation, demonstrated strong predictive performance, with a TSS of 0.88, emphasising the significance of environmental variables in determining habitat suitability for honey bees. The combined model had a higher TSS of 0.96, indicating that the combination of climate and environmental variables enhances the model's performance. By the 2020-2039 period, approximately 88% of highly suitable habitats for honey bees are projected to transition from their current state to become moderate (14.84%) to marginally suitable (13.46%) areas. Predictions for the 2060-2079 period reveal a concerning trend: 100% of highly suitable land transitions into moderately (0.54%), marginally (17.56%), or not suitable areas (81.9%) for honey bees. These results emphasise the critical need for targeted conservation efforts and the implementation of policies aimed at safeguarding honey bees and the vital apiary industry.

Rapid characterization of physical properties for the pharmaceutical pellet cores based on NIR spectroscopy and ensemble learning.

During the development of sustained-release pellets, the physical characteristics of the pellet cores can affect drug release in the preparation. The method based on near-infrared (NIR) spectroscopy and ensemble learning was proposed to swiftly assess the physical properties of the pellet cores. In the research, the potential of three algorithms, direct standardization (DS), partial least squares regression (PLSR) and generalized regression neural network (GRNN), was investigated and compared. The performance of the DS, PLSR and GRNN models were improved after applying bootstrap aggregating (Bagging) ensemble learning. And the Bagging-GRNN model showed the best predictive capacity. Except for inter-particle porosity, the mean absolute deviations of other 11 physical parameters were less than 1.0. Furthermore, the cosine coefficient values between the actual and predicted physical fingerprints was higher than 0.98 for 15 out of the 16 validation samples when using the Bagging-GRNN model. To reduce the model complexity, the 60 variables significantly correlated with angle of repose, particle size (D50) and roundness were utilized to develop the simplified Bagging-GRNN model. And the simplified model showed satisfactory predictive capacity. In summary, the developed ensemble modelling strategy based NIR spectra is a promising approach to rapidly characterize the physical properties of the pellet cores.

Development of artificial neural network models to predict the PAMPA effective permeability of new, orally administered drugs active against the coronavirus SARS-CoV-2.

Responding to the pandemic caused by SARS-CoV-2, the scientific community intensified efforts to provide drugs effective against the virus. To strengthen these efforts, the "COVID Moonshot" project has been accepting public suggestions for computationally triaged, synthesized, and tested molecules. The project aimed to identify molecules of low molecular weight with activity against the virus, for oral treatment. The ability of a drug to cross the intestinal cell membranes and enter circulation decisively influences its bioavailability, and hence the need to optimize permeability in the early stages of drug discovery. In our present work, as a contribution to the ongoing scientific efforts, we employed artificial neural network algorithms to develop QSAR tools for modelling the PAMPA effective permeability (passive diffusion) of orally administered drugs. We identified a set of 61 features most relevant in explaining drug cell permeability and used them to develop a stacked regression ensemble model, subsequently used to predict the permeability of molecules included in datasets made available through the COVID Moonshot project. Our model was shown to be robust and may provide a promising framework for predicting the potential permeability of molecules not yet synthesized, thus guiding the process of drug design. Supplementary Information: The online version contains supplementary material available at 10.1007/s13721-023-00410-9.

AlphaFold, small-angle X-ray scattering and ensemble modelling: a winning combination for intrinsically disordered proteins.

A recent article by Brookes, Rocco, Vachette & Trewhella [J. Appl. Cryst. (2023), 56, 910-926] on improving the accuracy of AlphaFold structural predictions for disordered proteins is discussed.

AlphaFold-predicted protein structures and small-angle X-ray scattering: insights from an extended examination of selected data in the Small-Angle Scattering Biological Data Bank.

By providing predicted protein structures from nearly all known protein sequences, the artificial intelligence program AlphaFold (AF) is having a major impact on structural biology. While a stunning accuracy has been achieved for many folding units, predicted unstructured regions and the arrangement of potentially flexible linkers connecting structured domains present challenges. Focusing on single-chain structures without prosthetic groups, an earlier comparison of features derived from small-angle X-ray scattering (SAXS) data taken from the Small-Angle Scattering Biological Data Bank (SASBDB) is extended to those calculated using the corresponding AF-predicted structures. Selected SASBDB entries were carefully examined to ensure that they represented data from monodisperse protein solutions and had sufficient statistical precision and q resolution for reliable structural evaluation. Three examples were identified where there is clear evidence that the single AF-predicted structure cannot account for the experimental SAXS data. Instead, excellent agreement is found with ensemble models generated by allowing for flexible linkers between high-confidence predicted structured domains. A pool of representative structures was generated using a Monte Carlo method that adjusts backbone dihedral allowed angles along potentially flexible regions. A fast ensemble modelling method was employed that optimizes the fit of pair distance distribution functions [P(r) versus r] and intensity profiles [I(q) versus q] computed from the pool to their experimental counterparts. These results highlight the complementarity between AF prediction, solution SAXS and molecular dynamics/conformational sampling for structural modelling of proteins having both structured and flexible regions.

Habitat Suitability and Conflict Zone Mapping for the Blue Bull (Boselaphus tragocamelus) across Nepal.

Rapidly changing environmental conditions (bioclimatic, anthropogenic, topographic, and vegetation-related variables) are likely to alter the spatial distribution of flora and fauna. To understand the influence of environmental variables on the Blue bull's distribution and to identify potential conflict zones, the habitat suitability analysis of the Blue bull was performed using ensemble modeling. We modelled the distribution of the Blue bull using an extensive database on the current distribution of the Blue bull and selected 15 ecologically significant environmental variables. We used ten species distribution modeling algorithms available in the BIOMOD2 R package. Among the ten algorithms, the Random Forest, Maxent, and Generalized linear model had the highest mean true skill statistics scores, ensuring better model performance, and were considered for further analysis. We found that 22,462.57 km2 (15.26%) of Nepal is suitable for the Blue bull. Slope, precipitation seasonality, and distance to the road are the environmental variables contributing the most to the distribution of Blue bull. Of the total predicted suitable habitats, 86% lies outside protected areas and 55% overlaps with agricultural land. Thus, we recommend that the future conservation initiatives including appropriate conflict mitigation measures should be prioritized equally in both protected areas and outside protected areas to ensure the species' survival in the region.

Enhancing urban flood forecasting in drainage systems using dynamic ensemble-based data mining.

This study presents a novel approach for urban flood forecasting in drainage systems using a dynamic ensemble-based data mining model which has yet to be utilised properly in this context. The proposed method incorporates an event identification technique and rainfall feature extraction to develop weak learner data mining models. These models are then stacked to create a time-series ensemble model using a decision tree algorithm and confusion matrix-based blending method. The proposed model was compared to other commonly used ensemble models in a real-world urban drainage system in the UK. The results show that the proposed model achieves a higher hit rate compared to other benchmark models, with a hit rate of around 85% vs 70 % for the next 3 h of forecasting. Additionally, the proposed smart model can accurately classify various timesteps of flood or non-flood events without significant lag times, resulting in fewer false alarms, reduced unnecessary risk management actions, and lower costs in real-time early warning applications. The findings also demonstrate that two features, "antecedent precipitation history" and "seasonal time occurrence of rainfall," significantly enhance the accuracy of flood forecasting with a hit rate accuracy ranging from 60 % to 10 % for a lead time of 15 min to 3 h.

A new approach to estimate soil organic carbon content targets in European croplands topsoils.

Adopting land management practices that increase the stock of soil organic carbon (SOC) in croplands is widely promoted as a win-win strategy to enhance soil health and mitigate climate change. In this context, the definition of reference SOC content and stock values is needed to provide reliable targets to farmers, policymakers, and stakeholders. In this study, we used the LUCAS dataset to compare different methods for evaluating reference SOC content and stock values in European croplands topsoils (0-20 cm depth). Methods gave generally similar estimates although being built on very different assumptions. In the absence of an objective criterion to establish which approach is the most suitable to determine SOC reference values, we propose an ensemble modelling approach that consists in extracting the estimates using different relevant methods and retaining the median value among them. Interestingly, this approach led us to select values from the three different approaches with similar frequencies. Using estimated bulk density values, we obtained a first rough estimate of 3.5 Gt C of SOC storage potential in the cropland topsoils that we interpret as a long-term aspirational target that would be reachable only under extreme changes in agricultural practices. The use of additional methods in the ensemble modelling approach and more valid statistical spatial estimates may further refine our approach designed for the estimation of SOC reference values for croplands.

Predicting priority management areas for land use/cover change in the transboundary Okavango basin based on machine learning.

Remote sensing and modelling of land use/land cover (LULC) change is useful to reveal the extent and spatial patterns of landscape changes at various environments and scales. Predicting susceptibility to LULC change is crucial for policy formulation and land management. However, the use of machine learning (ML) for modelling LULC change is limited. This study modelled LULC change susceptibility in the Okavango basin using ML techniques. Areas with high LULC change susceptibility are termed priority management areas (PMAs) in this study. Trajectories of LULC change between 1996 and 2020 are derived from existing LULC change maps of the Okavango basin. Overlay analysis is then used to detect patches of LULC change transitions. Three LULC transitional categories are adopted for modelling PMAs, namely 1) from natural to anthropogenic classes (Category A); 2) from anthropogenic to natural classes (Category B); and 3) from natural to another natural class (Category C). An ensemble of ML algorithms is calibrated with categories of LULC change and social-ecological drivers of change to produce maps showing the susceptibility of LULC change in the basin. Thereafter, thresholding is done on probability maps of susceptibility to LULC change based on the maximum sum of sensitivity and specificity (max SSS) to delineate PMAs. Results for trajectories of LULC change indicate that anthropogenic activities (croplands, built-up areas, and barelands) generally expanded, displacing natural areas (wetlands, woodlands, water, and shrubland) from 1996 to 2020. Regarding PMAs, anthropogenic-related PMAs (Category A ∼34 560 km2) covered a larger area compared to the natural ones (Categories B∼33 407 km2) and (Categories C∼15 040 km2). The findings of this study emphasize the value of ensemble ML modelling in identifying PMAs and guiding transboundary land use planning. Overall, this study highlights the role of anthropogenic activities in driving land use changes in Transboundary Drainage Basins (TDBs) and suggests a need to promote sustainable practices in predicted PMAs through comprehensive planning to ensure water availability in the Okavango basin.