4.6 Article

Nearest neighbour distance matching Leave-One-Out Cross-Validation for map validation

相关参考文献

注意:仅列出部分参考文献,下载原文获取全部文献信息。
Article Computer Science, Information Systems

Spatial machine-learning model diagnostics: a model-agnostic distance-based approach

Alexander Brenning

Summary: This paper proposes SPEPs and SVIPs as novel assessment and interpretation tools for spatial prediction models, focusing on prediction distance. The suitability of these tools is demonstrated in two case studies, showing their effectiveness and reliability across different models. The paper also highlights the limitations of related cross-validation techniques and provides recommendations regarding the spatial prediction horizon.

INTERNATIONAL JOURNAL OF GEOGRAPHICAL INFORMATION SCIENCE (2022)

Article Ecology

Spatial cross-validation is not the right way to evaluate map accuracy

Alexandre M. J-C Wadoux et al.

Summary: Recent studies suggest that cross-validation statistics in mapping studies are often optimistic due to the neglect of spatial autocorrelation. A numerical experiment on above-ground biomass mapping demonstrates that standard cross-validation results in smaller bias compared to spatial cross-validation. Spatial cross-validation lacks theoretical foundation and should not be used for map accuracy assessment, while standard cross-validation is inadequate for clustered data. Design-unbiased accuracy assessment can be achieved through probability sampling and design-based inference without considering spatial autocorrelation.

ECOLOGICAL MODELLING (2021)

Article Environmental Sciences

Validation of the US Geological Survey's Land Change Monitoring, Assessment and Projection (LCMAP) Collection 1.0 annual land cover products 1985-2017

Stephen V. Stehman et al.

Summary: The US Geological Survey has released annual land cover and land cover change products for the CONUS, with an overall accuracy of 82.5% as assessed by a reference sample dataset. Accuracy varies by land cover class, with lower accuracy in the eastern US, and specific user and producer accuracies were reported for different land cover types. Change accuracy improved when allowing for a +/- 2-year window for agreement, and footprint accuracy for areas experiencing land cover change was reported.

REMOTE SENSING OF ENVIRONMENT (2021)

Article Ecology

Predicting into unknown space? Estimating the area of applicability of spatial prediction models

Hanna Meyer et al.

Summary: Machine learning algorithms are popular for spatial mapping due to their ability to fit complex relationships, but their use is limited to data similar to the training set. The study proposes a method to assess the area where a prediction model can be reliably applied, using a Dissimilarity Index (DI) to define the Area of Applicability (AOA) and map estimated performance. Simulation studies show comparable prediction errors within the AOA to cross-validation errors, emphasizing the importance of considering the relationship between DI and cross-validation performance.

METHODS IN ECOLOGY AND EVOLUTION (2021)

Editorial Material Ecology

National climate and biodiversity strategies are hamstrung by a lack of maps

Guido Schmidt-Traub

NATURE ECOLOGY & EVOLUTION (2021)

Article Mathematics, Interdisciplinary Applications

Geostatistical Learning: Challenges and Opportunities

Julio Hoffimann et al.

Summary: Statistical learning theory is essential for applied machine learning, but doesn't address the challenges of geospatial statistical learning where spatial correlation and covariate shift pose unique challenges. Existing methods are inadequate for model selection in a geospatial context and further research is needed to develop more effective approaches.

FRONTIERS IN APPLIED MATHEMATICS AND STATISTICS (2021)

Article Multidisciplinary Sciences

Spatial validation reveals poor predictive performance of large-scale ecological mapping models

Pierre Ploton et al.

NATURE COMMUNICATIONS (2020)

Article Ecology

dsmextra: Extrapolation assessment tools for density surface models

Phil J. Bouchet et al.

METHODS IN ECOLOGY AND EVOLUTION (2020)

Review Geography, Physical

Geospatial data mining for digital raster mapping

Bruce K. Wylie et al.

GISCIENCE & REMOTE SENSING (2019)

Article Multidisciplinary Sciences

Soil nematode abundance and functional group composition at a global scale

Johan van den Hoogen et al.

NATURE (2019)

Article Environmental Sciences

A comparison of resampling methods for remote sensing classification and accuracy assessment

Mitchell B. Lyons et al.

REMOTE SENSING OF ENVIRONMENT (2018)

Review Ecology

Outstanding Challenges in the Transferability of Ecological Models

Katherine L. Yates et al.

TRENDS IN ECOLOGY & EVOLUTION (2018)

Article Computer Science, Information Systems

Estimating the prediction performance of spatial models via spatial k-fold cross validation

Jonne Pohjankukka et al.

INTERNATIONAL JOURNAL OF GEOGRAPHICAL INFORMATION SCIENCE (2017)

Article Biodiversity Conservation

virtualspecies, an R package to generate virtual species distributions

Boris Leroy et al.

ECOGRAPHY (2016)

Article Environmental Sciences

Mapping Daily Air Temperature for Antarctica Based on MODIS LST

Hanna Meyer et al.

REMOTE SENSING (2016)

Review Ecology

Mapping where ecologists work: biases in the global distribution of terrestrial ecological observations

Laura J. Martin et al.

FRONTIERS IN ECOLOGY AND THE ENVIRONMENT (2012)

Article Geography, Physical

Evaluation of transfer functions in spatially structured environments

R. J. Telford et al.

QUATERNARY SCIENCE REVIEWS (2009)

Article Biochemistry & Molecular Biology

Super learner

Mark J. van der Laan et al.

STATISTICAL APPLICATIONS IN GENETICS AND MOLECULAR BIOLOGY (2007)

Article Computer Science, Interdisciplinary Applications

Multivariable geostatistics in S: the gstat package

EJ Pebesma

COMPUTERS & GEOSCIENCES (2004)