Modeling complex species-environment relationships through spatially-varying coefficient occupancy models

arXiv (Cornell University), Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 1, 2023

Occupancy models are frequently used by ecologists to quantify spatial variation in species distributions while accounting for observational biases the collection of detection-nondetection data. However, common assumption that a single set regression coefficients can adequately explain species-environment relationships is often unrealistic, especially across large domains. Here we develop single-species (i.e., univariate) and multi-species multivariate) spatially-varying coefficient (SVC) occupancy account relationships. We employ Nearest Neighbor Gaussian Processes Polya-Gamma data augmentation hierarchical Bayesian framework yield computationally efficient Gibbs samplers, which implement spOccupancy R package. For models, use factor dimension reduction efficiently model datasets with numbers (e.g., > 10). The readily enables generation posterior predictive maps SVCs, fully propagated uncertainty. apply our SVC variability between maximum breeding season temperature occurrence probability 21 grassland bird U.S. Jointly modeling generally outperformed all revealed substantial temperatures. Our particularly relevant quantifying using from large-scale monitoring programs, becoming increasingly prevalent answering macroscale ecological questions regarding wildlife responses global change.

Language: Английский

---

Lessons from spatial transcriptomics and computational geography in mapping the transcriptome

AGILE GIScience Series, Journal Year: 2024, Volume and Issue: 5, P. 1 - 6

Published: May 30, 2024

Abstract. Spatial data, data with some form of location attached, are the norm: all spatial now. However requires consideration three critical characteristics, observation auto-correlated, process spatially non-stationarity and effect MAUP. Geographers familiar these have tools, rubrics workflows to accommodate them understand their impacts on statical inference, understanding prediction. However, increasingly researchers in non geographical domains, no experience of, or exposure quantitative geography GIScience undertaking analyses such without full any properties. This short paper describes recent interactions work research gene analysis Transcriptomics, highlight opportunities for inform steer many new users data.

Language: Английский

---

Patterns and drivers of population trends on individual Breeding Bird Survey routes using spatially explicit models and route-level covariates

Published: Oct. 27, 2023

Spatial patterns in population trends, particularly those at finer geographic scales, can help us better understand the factors driving change North American birds. The standard status and trend models for Breeding Bird Survey (BBS) were designed to estimate trends within broad strata, such as Conservation Regions, U.S. states, Canadian territories or provinces. Calculating estimates level of individual survey transects (“routes”) from BBS allows explore spatial simultaneously effects covariates, habitat-loss annual weather, on both relative abundance (changes through time). Here, we describe four related hierarchical Bayesian that routes, implemented probabilistic programing language Stan. All route-level abundances using a structure shares information among three share spatially explicit way. use either an intrinsic Conditional Autoregressive distance-based Gaussian process components. We fit all data 71 species then only two (one one non-spatial) additional 216 due computational limitations. Leave-future-out cross-validation showed outperformed non-spatial model 284 out 287 species. For tested here, best approach modeling components depended species; Process had highest predictive accuracy 2/3 here iCAR was remaining 1/3. also present examples covariate analyses focused temporal variation habitat Rufous Hummingbird (Selasphorus rufus) Horned Grebe (Podiceps auritus). Covariates explain affect rate Route-level are useful visualizing change, generating hypotheses causes comparing regions species, testing with relevant covariates.

Language: Английский

---

Reassessment of French breeding bird population sizes using citizen science and accounting for species detectability

PeerJ, Journal Year: 2024, Volume and Issue: 12, P. e17889 - e17889

Published: Aug. 27, 2024

Higher efficiency in large-scale and long-term biodiversity monitoring can be obtained through the use of Essential Biodiversity Variables, among which species population sizes provide key data for conservation programs. Relevant estimations assessment actual are critical conservation, especially current context global erosion. However, knowledge on size varies greatly, depending status ranges. While most threatened or restricted-range generally benefit from exhaustive counts surveys, common widespread tends to neglected is simply more challenging achieve. In such a context, citizen science (CS) powerful tool engagement various volunteers, permitting acquisition long term over large spatial scales. Despite this substantially increased sampling effort, detectability issues imply that even may remain unnoticed at suitable sites. The structured CS schemes, including repeated visits, enables model detection process, reliable inferences estimates. Here, we relied French scheme (EPOC-ODF) comprising 27,156 complete checklists 3,873 sites collected during 2021–2023 breeding seasons estimate 63 bird using hierarchical distance (HDS). These estimates were compared previous expert-based atlas estimations, did not account issues. We found former lower than those estimated HDS 65% species. Such prevalence likely due conservative inferred semi-quantitative assessments used atlas. also with long-range songs as Common Cuckoo ( Cuculus canorus ), Eurasian Hoopoe Upupa epops ) Blackbird Turdus merula had, contrast, higher our models. Our study highlights need rely sound statistical methodology ensure ecological adequate uncertainty estimation advocates reliance support monitoring.

Language: Английский

---

Joint Urban Modeling With Graph Convolutional Networks and Crowdsourced Data: A Novel Approach

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 57796 - 57805

Published: Jan. 1, 2024

Graph Convolutional Networks (GCN) are a potent and adaptable tool for effectively processing analyzing continuous spatial data. Despite the substantial potential of GCN in various domains, most existing data prediction models confined to defining weights solely based on distance. To overcome this limitation, study proposes novel approach obtain second-level embedding Points Interests (POIs) by employing Delaunay Triangulation (DT), Random Walk, Skip-Gram model training. Subsequently, enhanced features obtained through aggregation strategies regional embedding. The integrated grid data, including longitude latitude coordinates, features, target values, then integrated. Finally, is utilized training fitting achieve final value. By considering influence prediction, can more accurately reflect distribution relationships actual environment. Furthermore, we have experimentally validated effectiveness approach, demonstrating that it significantly enhances accuracy when compared original model's approach.

Language: Английский

---