Exploring the potential for online data sources to enhance species threat mapping through the case study of global bat exploitation.

Expanding digital data sources, including social media and online news, provide a low-cost way to examine human-nature interactions, such as wildlife exploitation. However, the extent to which using such data sources can expand or bias understanding of the distribution and intensity of threats has not been comprehensively assessed. To address this gap, we quantified the geographical and temporal distribution of online sources documenting the hunting and trapping, consumption, or trade of bats (Chiroptera) and compared these with the distribution of studies obtained from a systematic literature search and species listed as threatened by exploitation on the International Union for Conservation of Nature Red List. Online records were collected using automated searches of Facebook, Twitter, Google, and Bing and were filtered using machine classification. This yielded 953 relevant social media posts and web pages, encompassing 1099 unique records of bat exploitation from 84 countries. Although the number of records per country was significantly predicted by the number of academic studies per country, online records provided additional locations and more recent records of bat exploitation, including 22 countries not present in academic literature. This demonstrates the value of online resources in providing more complete geographical representation. However, confounding variables can bias the analysis of spatiotemporal trends. Online bat exploitation records showed peaks in 2020 and 2014, after accounting for increases in internet users through time. The second of these peaks could be attributed to the COVID-19 outbreak, and speculation about the role of bats in its epidemiology, rather than to true changes in exploitation. Overall, our results showed that data from online sources provide additional knowledge on the global extent of wildlife exploitation, which could be used to identify early warnings of emerging threats and pinpoint locations for further research.

Sondeo del potencial de las fuentes virtuales de datos para mejorar el mapeo de amenazas para las especies por medio del estudio de caso de la explotación mundial de murciélagos Resumen La expansión de las fuentes virtuales, incluidas las redes sociales y las noticias en línea, proporciona una forma asequible de analizar las interacciones entre el humano y la naturaleza, como la explotación de fauna. Sin embargo, no se ha analizado por completo el rango al que dichas fuentes pueden expandir o sesgar el conocimiento de la distribución e intensidad de las amenazas. Para abordar este vacío cuantificamos la distribución geográfica y temporal de las fuentes virtuales que documentan la caza, captura, consumo o mercado de murciélagos (Chiroptera) y las comparamos con la distribución de los estudios obtenidos de una búsqueda sistemática en la literatura y con las especies catalogadas como amenazadas por la explotación según la Lista Roja de la Unión Internacional para la Conservación de la Naturaleza. Recolectamos los registros virtuales por medio de búsquedas automatizadas en Facebook, Twitter, Google y Bing y después las filtramos con clasificaciones automatizadas. Esto arrojó 953 publicaciones relevantes en redes sociales y sitios web que englobaban 1099 registros únicos de la explotación de murciélagos en 84 países. Aunque pronosticamos de forma significativa el número de registros por país con el número de estudios académicos por país, los registros virtuales proporcionaron localidades adicionales y registros más recientes de la explotación de murciélagos, incluyendo a 22 países que no se encuentran en la literatura académica. Lo anterior demuestra el valor que tienen los recursos en línea para proporcionar una representación geográfica más completa. Sin embargo, las variables confusas pueden sesgar el análisis de las tendencias espaciotemporales. Los registros virtuales de la explotación de murciélagos mostraron picos en 2020 y en 2014, esto después de considerar el incremento de usuarios de internet con el tiempo. El segundo pico podría atribuirse al brote de COVID­19 y la especulación en torno al papel que tenían los murciélagos en su epidemiología y no tanto a un verdadero cambio en la explotación. En general, nuestros resultados mostraron que los datos de las fuentes virtuales proporcionan conocimiento adicional sobre el alcance mundial de la explotación de fauna, el cual podría usarse para identificar señales tempranas de amenazas emergentes y ubicar localidades para su mayor investigación.

Identifying socioeconomic and biophysical factors driving forest loss in protected areas.

Protected areas (PAs) are a commonly used strategy to confront forest conversion and biodiversity loss. Although determining drivers of forest loss is central to conservation success, understanding of them is limited by conventional modeling assumptions. We used random forest regression to evaluate potential drivers of deforestation in PAs in Mexico, while accounting for nonlinear relationships and higher order interactions underlying deforestation processes. Socioeconomic drivers (e.g., road density, human population density) and underlying biophysical conditions (e.g., precipitation, distance to water, elevation, slope) were stronger predictors of forest loss than PA characteristics, such as age, type, and management effectiveness. Within PA characteristics, variables reflecting collaborative and equitable management and PA size were the strongest predictors of forest loss, albeit with less explanatory power than socioeconomic and biophysical variables. In contrast to previously used methods, which typically have been based on the assumption of linear relationships, we found that the associations between most predictors and forest loss are nonlinear. Our results can inform decisions on the allocation of PA resources by strengthening management in PAs with the highest risk of deforestation and help preemptively protect key biodiversity areas that may be vulnerable to deforestation in the future.

Identificación de los factores biofísicos y socioeconómicos que impulsan la pérdida de bosques en las áreas protegidas Resumen Las áreas protegidas son una estrategia de uso común para hacer frente a la conversión forestal y la pérdida de biodiversidad. Aunque determinar los factores que impulsan la pérdida de bosques es fundamental para el éxito de la conservación, su comprensión se ve limitada por los supuestos de modelación convencionales. Utilizamos la regresión de bosques aleatorios para evaluar los posibles impulsores de la deforestación en las áreas protegidas de México, considerando las relaciones no lineales y las interacciones de orden superior que subyacen a los procesos de deforestación. Los impulsores socioeconómicos (densidad de carreteras, densidad de población humana) y las condiciones biofísicas subyacentes (precipitaciones, distancia al agua, elevación, pendiente) fueron predictores más fuertes de la pérdida de bosques que las características de las áreas protegidas, como la edad, el tipo y la efectividad de la gestión. Dentro de las características de las áreas protegidas, las variables que reflejan una gestión colaborativa y equitativa y el tamaño del área protegida fueron los predictores más potentes de la pérdida de bosques, aunque con menor poder explicativo que las variables socioeconómicas y biofísicas. A diferencia de los métodos utilizados anteriormente, que suelen basarse en el supuesto de relaciones lineales, observamos que las asociaciones entre la mayoría de los predictores y la pérdida de bosques no son lineales. Nuestros resultados pueden servir de base para la toma de decisiones sobre la asignación de los recursos para las áreas protegidas, reforzando la gestión en las zonas protegidas con mayor riesgo de deforestación y ayudando a proteger de forma preventiva zonas clave para la biodiversidad que pueden ser vulnerables a la deforestación en el futuro.

Implications of the scale of detection for inferring co-occurrence patterns from paired camera traps and acoustic recorders.

Multifunctional landscapes that support economic activities and conservation of biological diversity (e.g., cattle ranches with native forest) are becoming increasingly important because small remnants of native forest may comprise the only habitat left for some wildlife species. Understanding the co-occurrence between wildlife and disturbance factors, such as poaching activity and domesticated ungulates, is key to successful management of multifunctional landscapes. Tools to measure co-occurrence between wildlife and disturbance factors include camera traps and autonomous acoustic recording units. We paired 52 camera-trap stations with acoustic recorders to investigate the association between 2 measures of disturbance (poaching and cattle) and wild ungulates present in multifunctional landscapes of the Colombian Orinoquía. We used joint species distribution models to investigate species-habitat associations and species-disturbance correlations. One model was fitted using camera-trap data to detect wild ungulates and disturbance factors, and a second model was fitted after replacing camera-trap detections of disturbance factors with their corresponding acoustic detections. The direction, significance, and precision of the effect of covariates depended on the sampling method used for disturbance factors. Acoustic monitoring typically resulted in more precise estimates of the effects of covariates and of species-disturbance correlations. Association patterns between wildlife and disturbance factors were found only when disturbance was detected by acoustic recorders. Camera traps allowed us to detect nonvocalizing species, whereas audio recording devices increased detection of disturbance factors leading to more precise estimates of co-occurrence patterns. The collared peccary (Pecari tajacu), lowland tapir (Tapirus terrestris), and white-tailed deer (Odocoileus virginianus) co-occurred with disturbance factors and are conservation priorities due to the greater risk of poaching or disease transmission from cattle.

Implicaciones de la escala de detección para inferir los patrones de coocurrencia a partir de fototrampas y grabaciones emparejadas Resumen Los paisajes multifuncionales que sostienen actividades económicas y la conservación de la biodiversidad (p. ej., ganadería en bosques nativos) son cada vez más importantes porque los pequeños reductos de bosque nativo podrían comprender el único hábitat disponible para algunas especies de fauna. Es importante entender la coocurrencia entre la fauna y los factores de perturbación, como la actividad furtiva y los ungulados domésticos, para tener un manejo exitoso de los paisajes multifuncionales. Las herramientas que miden esta relación incluyen las fototrampas y las unidades autónomas de grabaciones acústicas. Emparejamos 52 estaciones de fototrampas con grabadoras acústicas para investigar la asociación entre dos medidas de perturbación (actividad furtiva y ganado) y los ungulados silvestres presentes en los paisajes multifuncionales de la Orinoquía colombiana. Usamos modelos conjuntos de distribución de especies para investigar las asociaciones especie-hábitat y las correlaciones especie-perturbación. Ajustamos un modelo con datos de fototrampeo para detectar ungulados silvestres y factores de perturbación; un segundo modelo fue ajustado después de reemplazar las detecciones por fototrampas de los factores de perturbación con las detecciones acústicas correspondientes. La dirección, importancia y precisión del efecto de las covarianzas dependió del método de muestreo usado para los factores de perturbación. El monitoreo acústico casi siempre resultó en estimaciones más precisas de los efectos de las covarianzas y de las correlaciones especie-perturbación. Los patrones de asociación entre la fauna y los factores de perturbación sólo se presentaron cuando las grabadoras acústicas detectaron la perturbación. Las fototrampas nos permitieron detectar especies que no vocalizan, mientras que las grabaciones de audio incrementaron la detección de factores de perturbación, lo que resultó en estimados más precisos de los patrones de coocurrencia. El pecarí de collar (Pecari tajacu), el tapir (Tapirus terrestris) y el venado cola blanca (Odocoileus virginianus) tuvieron coocurrencia con los factores de perturbación y tienen prioridad de conservación debido al mayor riesgo de caza furtiva o transmisión de enfermedades del ganado.

Catalyzing success in community-based conservation.

Efforts to devolve rights and engage Indigenous Peoples and local communities in conservation have increased the demand for evidence of the efficacy of community-based conservation (CBC) and insights into what enables its success. We examined the human well-being and environmental outcomes of a diverse set of 128 CBC projects. Over 80% of CBC projects had some positive human well-being or environmental outcomes, although just 32% achieved positive outcomes for both (i.e., combined success). We coded 57 total national-, community-, and project-level variables and controls from this set, performed random forest classification to identify the variables most important to combined success, and calculated accumulated local effects to describe their individual influence on the probability of achieving it. The best predictors of combined success were 17 variables suggestive of various recommendations and opportunities for conservation practitioners related to national contexts, community characteristics, and the implementation of various strategies and interventions informed by existing CBC frameworks. Specifically, CBC projects had higher probabilities of combined success when they occurred in national contexts supportive of local governance, confronted challenges to collective action, promoted economic diversification, and invested in various capacity-building efforts. Our results provide important insights into how to encourage greater success in CBC.

Los esfuerzos por transferirle derechos e involucrar a los pueblos originarios y a las comunidades locales en la conservación han incrementado la demanda de evidencia sobre la eficiencia de la conservación basada en la comunidad (CBC) y de conocimiento sobre lo que posibilita su éxito. Analizamos los resultados ambientales y de bienestar humano en un conjunto diverso de 28 proyectos de CBC. Más del 80% de estos proyectos tuvieron resultados positivos para el ambiente o el bienestar humano, aunque sólo el 32% logró resultados positivos para ambos (es decir, éxito combinado). Codificamos en total 57 variables y controles a nivel nacional, comunitario y de proyecto en este conjunto, aplicamos una clasificación aleatoria de bosque para identificar las variables más importantes para el éxito combinado y calculamos los efectos locales acumulados para describir su influencia sobre la probabilidad de alcanzar el éxito combinado. Los mejores pronósticos del éxito combinado se obtuvieron con 17 variables sugerentes de varias políticas y oportunidades para los practicantes de la conservación relacionadas con los contextos nacionales, las características de la comunidad y la implementación de varias estrategias e intervenciones guiadas por los marcos existentes de CBC. Específicamente, los proyectos de CBC tuvieron mayor probabilidad de tener éxito combinado cuando se dieron dentro de contextos nacionales que respaldan la gobernanza local, enfrentan los retos de la acción colectiva, promueven la diversificación económica e invierten en varios esfuerzos por construir capacidades. Nuestros resultados proporcionan información importante sobre cómo alentar un mayor éxito en la CBC.

Applying deep learning to right whale photo identification.

Photo identification is an important tool for estimating abundance and monitoring population trends over time. However, manually matching photographs to known individuals is time-consuming. Motivated by recent developments in image recognition, we hosted a data science challenge on the crowdsourcing platform Kaggle to automate the identification of endangered North Atlantic right whales (Eubalaena glacialis). The winning solution automatically identified individual whales with 87% accuracy with a series of convolutional neural networks to identify the region of interest on an image, rotate, crop, and create standardized photographs of uniform size and orientation and then identify the correct individual whale from these passport-like photographs. Recent advances in deep learning coupled with this fully automated workflow have yielded impressive results and have the potential to revolutionize traditional methods for the collection of data on the abundance and distribution of wild populations. Presenting these results to a broad audience should further bridge the gap between the data science and conservation science communities.

Aplicación del Aprendizaje Profundo a la Identificación Fotográfica de la Ballena Franca Resumen La identificación fotográfica es una herramienta importante para la estimación de la abundancia y el monitoreo de las tendencias poblacionales en el tiempo. Sin embargo, corresponder las fotografías con los individuos conocidos requiere de mucho tiempo. Motivados por los avances recientes en el reconocimiento de imágenes, decidimos acoger un reto de datos científicos en la plataforma de colaboración masiva Kaggle para automatizar la identificación de ballenas francas del Atlántico norte (Eubalaena glacialis), especie que se encuentra en peligro de extinción. La solución ganadora identificó automáticamente a las ballenas individuales con una certeza del 87% y con una serie de redes neurales convolucionales para identificar la región de interés en una imagen, rotar, recortar, y crear fotografías estandarizadas de tamaño y orientación uniforme y después identificar al individuo correcto a partir de estas fotografías tamaño pasaporte. Los avances recientes en el aprendizaje profundo acoplados a este flujo de trabajo completamente automatizado han producido resultados impresionantes y tienen el potencial para revolucionar los métodos tradicionales de recolección de datos de abundancia y distribución de las poblaciones silvestres. La presentación de estos resultados ante un público amplio debería reducir aún más el vacío que existe entre los datos científicos y las comunidades científicas para la conservación.

Nailfold capillaroscopy and deep learning in diabetes.

OBJECTIVE: To determine whether nailfold capillary images, acquired using video capillaroscopy, can provide diagnostic information about diabetes and its complications. RESEARCH DESIGN AND METHODS: Nailfold video capillaroscopy was performed in 120 adult patients with and without type 1 or type 2 diabetes, and with and without cardiovascular disease. Nailfold images were analyzed using convolutional neural networks, a deep learning technique. Cross-validation was used to develop and test the ability of models to predict five5 prespecified states (diabetes, high glycosylated hemoglobin, cardiovascular event, retinopathy, albuminuria, and hypertension). The performance of each model for a particular state was assessed by estimating areas under the receiver operating characteristics curves (AUROC) and precision recall curves (AUPR). RESULTS: A total of 5236 nailfold images were acquired from 120 participants (mean 44 images per participant) and were all available for analysis. Models were able to accurately identify the presence of diabetes, with AUROC 0.84 (95% confidence interval [CI] 0.76, 0.91) and AUPR 0.84 (95% CI 0.78, 0.93), respectively. Models were also able to predict a history of cardiovascular events in patients with diabetes, with AUROC 0.65 (95% CI 0.51, 0.78) and AUPR 0.72 (95% CI 0.62, 0.88) respectively. CONCLUSIONS: This proof-of-concept study demonstrates the potential of machine learning for identifying people with microvascular capillary changes from diabetes based on nailfold images, and for possibly identifying those most likely to have diabetes-related complications.

Transparent machine learning suggests a key driver in the decision to start insulin therapy in individuals with type 2 diabetes.

AIMS: The objective of this study is to establish a predictive model using transparent machine learning (ML) to identify any drivers that characterize therapeutic inertia. METHODS: Data in the form of both descriptive and dynamic variables collected from electronic records of 1.5 million patients seen at clinics within the Italian Association of Medical Diabetologists between 2005-2019 were analyzed using logic learning machine (LLM), a "clear box" ML technique. Data were subjected to a first stage of modeling to allow ML to automatically select the most relevant factors related to inertia, and then four further modeling steps individuated key variables that discriminated the presence or absence of inertia. RESULTS: The LLM model revealed a key role for average glycated hemoglobin (HbA1c) threshold values correlated with the presence or absence of insulin therapeutic inertia with an accuracy of 0.79. The model indicated that a patient's dynamic rather than static glycemic profile has a greater effect on therapeutic inertia. Specifically, the difference in HbA1c between two consecutive visits, what we call the HbA1c gap, plays a crucial role. Namely, insulin therapeutic inertia is correlated with an HbA1c gap of <6.6 mmol/mol (0.6%), but not with an HbA1c gap of >11 mmol/mol (1.0%). CONCLUSIONS: The results reveal, for the first time, the interrelationship between a patient's glycemic trend defined by sequential HbA1c measurements and timely or delayed initiation of insulin therapy. The results further demonstrate that LLM can provide insight in support of evidence-based medicine using real world data.

Unravelling psychiatric heterogeneity and predicting suicide attempts in women with trauma-related dissociation using artificial intelligence.

Background: Suicide is a leading cause of death, and rates of attempted suicide have increased during the COVID-19 pandemic. The under-diagnosed psychiatric phenotype of dissociation is associated with elevated suicidal self-injury; however, it has largely been left out of attempts to predict and prevent suicide.Objective: We designed an artificial intelligence approach to identify dissociative patients and predict prior suicide attempts in an unbiased, data-driven manner.Method: Participants were 30 controls and 93 treatment-seeking female patients with posttraumatic stress disorder (PTSD) and various levels of dissociation, including some with the PTSD dissociative subtype and some with dissociative identity disorder (DID).Results: Unsupervised learning models identified patients along a spectrum of dissociation. Moreover, supervised learning models accurately predicted prior suicide attempts with an F1 score up to 0.83. DID had the highest risk of prior suicide attempts, and distinct subtypes of dissociation predicted suicide attempts in PTSD and DID.Conclusions: These findings expand our understanding of the dissociative phenotype and underscore the urgent need to assess for dissociation to identify individuals at high-risk of suicidal self-injury.

Dissociation, feelings of detachment and disruption in one's sense of self and surroundings, is associated with an elevated risk of suicidal self-injury; however, it has largely been left out of attempts to predict and prevent suicide.Using machine learning techniques, we found dissociative identity disorder had the highest risk of prior suicide attempts, and distinct subtypes of dissociation predicted suicide attempts in posttraumatic stress disorder and dissociative identity disorder.These findings underscore the urgent need to assess for dissociation to identify individuals at high-risk of suicidal self-injury.

Prediction of the development of islet autoantibodies through integration of environmental, genetic, and metabolic markers.

BACKGROUND: The Environmental Determinants of the Diabetes in the Young (TEDDY) study has prospectively followed, from birth, children at increased genetic risk of type 1 diabetes. TEDDY has collected heterogenous data longitudinally to gain insights into the environmental and biological mechanisms driving the progression to persistent islet autoantibodies. METHODS: We developed a machine learning model to predict imminent transition to the development of persistent islet autoantibodies based on time-varying metabolomics data integrated with time-invariant risk factors (eg, gestational age). The machine learning was initiated with 221 potential features (85 genetic, 5 environmental, 131 metabolomic) and an ensemble-based feature evaluation was utilized to identify a small set of predictive features that can be interrogated to better understand the pathogenesis leading up to persistent islet autoimmunity. RESULTS: The final integrative machine learning model included 42 disparate features, returning a cross-validated receiver operating characteristic area under the curve (AUC) of 0.74 and an AUC of ~0.65 on an independent validation dataset. The model identified a principal set of 20 time-invariant markers, including 18 genetic markers (16 single nucleotide polymorphisms [SNPs] and two HLA-DR genotypes) and two demographic markers (gestational age and exposure to a prebiotic formula). Integration with the metabolome identified 22 supplemental metabolites and lipids, including adipic acid and ceramide d42:0, that predicted development of islet autoantibodies. CONCLUSIONS: The majority (86%) of metabolites that predicted development of islet autoantibodies belonged to three pathways: lipid oxidation, phospholipase A2 signaling, and pentose phosphate, suggesting that these metabolic processes may play a role in triggering islet autoimmunity.