Search USGS⌕ Search

SEARCH · Search USGS

Results for “Behavioral Ecology”

Search indexed USGS publications on groundwater, aquifers, geologic maps, mineral resources and earthquakes. Explore source records by subject and place.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 667 records · Page 37Linked to original sources

Savanna tree density, herbivores, and the herbaceous community: Bottom-up vs. top-down effects

Herbivores choose their habitats both to maximize forage intake and to minimize their risk of predation. For African savanna herbivores, the available habitats range in woody cover from open areas with few trees to dense, almost‐closed woodlands. This variation in woody cover or density can have a number of consequences for herbaceous species composition, cover, and productivity, as well as for ease of predator detection and avoidance. Here, we consider two alternative possibilities: first, that tree density affects the herbaceous vegetation, with concomitant “bottom‐up” effects on herbivore habitat preferences; or, second, that tree density affects predator visibility, mediating “top‐down” effects of predators on herbivore habitat preferences. We sampled sites spanning a 10‐fold range of tree densities in an Acacia drepanolobium‐dominated savanna in Laikipia, Kenya, for variation in (1) herbaceous cover, composition, and species richness; (2) wild and domestic herbivore use; and (3) degree of visibility obstruction by the tree layer. We then used structural equation modeling to consider the potential influences that tree density may have on herbivores and herbaceous community properties. Tree density was associated with substantial variation in herbaceous species composition and richness. Cattle exhibited a fairly uniform use of the landscape, whereas wild herbivores, with the exception of elephants, exhibited a strong preference for areas of low tree density. Model results suggest that this was not a response to variation in herbaceous‐community characteristics, but rather a response to the greater visibility associated with more open places. Elephants, in contrast, preferred areas with higher densities of trees, apparently because of greater forage availability. These results suggest that, for all but the largest species, top‐down behavioral effects of predator avoidance on herbivores are mediated by tree density. This, in turn, appears to have cascading effects on the herbaceous vegetation. These results shed light on one of the major features of the “landscape of fear” in which African savanna herbivores exist.

Laikipia County↗

Idiosyncratic spatial scaling of biodiversity–disease relationships

High host biodiversity is hypothesized to dilute the risk of vector-borne diseases if many host species are ‘dead ends' that cannot effectively transmit the disease and low-diversity areas tend to be dominated by competent host species. However, many studies on biodiversity–disease relationships characterize host biodiversity at single, local spatial scales, which complicates efforts to forecast disease risk if associations between host biodiversity and disease change with spatial scale. Here, our objective is to evaluate the spatial scaling of relationships between host biodiversity and Borrelia (the bacterial taxon which causes Lyme disease) infection prevalence in small mammals. We compared the associations between infection prevalence and small mammal host diversity for local communities (individual plots) and metacommunities (multiple plots aggregated within a landscape) sampled by the National Ecological Observatory Network (NEON), an emerging continental-scale environmental monitoring program with a hierarchical sampling design. We applied a multispecies, spatially-stratified capture–recapture model to a trapping dataset to estimate five small mammal biodiversity metrics, which we used to predict infection status for a subset of trapped individuals. We found that relationships between Borrelia infection prevalence and biodiversity did indeed vary when biodiversity was quantified at different spatial scales but that these scaling behaviors were idiosyncratic among the five biodiversity metrics. For example, species richness of local communities showed a negative (dilution) effect on infection prevalence, while species richness of the small mammal metacommunity showed a positive (amplification) effect on infection prevalence. Our modeling approach can inform future analyses as data from similar monitoring programs accumulate and become increasingly available through time. Our results indicate that a focus on single spatial scales when assessing the influence of biodiversity on disease risk provides an incomplete picture of the complexity of disease dynamics in ecosystems.

Ecography↗

Foraging trade-offs along a predator-permanence gradient in subalpine wetlands

We conducted a series of field and laboratory experiments to determine the direct and indirect effects of a top predator, the tiger salamander ( Ambystoma tigrinum nebulosum ), on larvae of two species of limnephilid caddisflies ( Limnephilus externus and Asynarchus nigriculus ) in subalpine wetlands in central Colorado. Asynarchus larvae predominate in temporary wetlands and are aggressive intraguild predators on Limnephilus larvae, which only predominate in permanent basins with salamanders. We first conducted a field experiment in mesocosms (cattle tanks) to quantify the predatory effects of different life stages of salamanders on the two caddisfly species. Two life stages of the salamanders (larvae and paedomorphs) preferentially preyed on Asynarchus relative to Limnephilus. Subsequent laboratory experiments revealed that high Asynarchus activity rates and relatively ineffective antipredatory behaviors led to higher salamander detection and attack rates compared to Limnephilus. In a second field experiment (full factorial for presence and absence of each of the three species), we found that salamander predation on Asynarchus had an indirect positive effect on Limnephilus : survival was higher in the presence of salamanders + Asynarchus than with just Asynarchus. In the laboratory we compared the predatory effects of salamanders with and without their mouths sewn shut and found the observed indirect positive effect on Limnephilus survival to be mainly the result of reduced numbers of Asynarchus rather than salamander-induced changes in Asynarchus behavior. We argue that indirect effects of predator–predator interactions on shared prey will be mainly density-mediated and not trait-mediated when one of the predators (in this case, Asynarchus ) is under strong selection for rapid growth and therefore does not modify foraging behaviors in response to the other predator. The reciprocal dominance of Limnephilus and Asynarchus in habitats with and without salamanders probably reflects a trade-off between competitive superiority and vulnerability to predation. The high activity levels and aggressiveness that enable Asynarchus to complete development in temporary habitats result in strong asymmetric competition (via intraguild predation) with Limnephilus. In permanent habitats these same behaviors increase Asynarchus vulnerability to salamander predation, which indirectly benefits Limnephilus. This and previous work implicate salamanders as keystone predators that exert a major influence on the composition of benthic and planktonic assemblages in subalpine wetlands.

Ecology↗

Legacy or colonization? Posteruption establishment of peregrine falcons (Falco peregrinus) on a volcanically active subarctic island.

How populations and communities reassemble following disturbances are affected by a number of factors, with the arrival order of founding populations often having a profound influence on later populations and community structure. Kasatochi Island is a small volcano located in the central Aleutian archipelago that erupted violently August 8, 2008, sterilizing the island of avian biodiversity. Prior to the eruption, Kasatochi was the center of abundance for breeding seabirds in the central Aleutian Islands and supported several breeding pairs of peregrine falcons ( Falco peregrinus ). We examined the reestablishment of peregrine falcons on Kasatochi by evaluating the genetic relatedness among legacy samples collected in 2006 to those collected posteruption and to other falcons breeding along the archipelago. No genotypes found in posteruption samples were identical to genotypes collected from pre-eruption samples. However, genetic analyses suggest that individuals closely related to peregrine falcons occupying pre-eruption Kasatochi returned following the eruption and successfully fledged young; thus, a genetic legacy of pre-eruption falcons was present on posteruption Kasatochi Island. We hypothesize that the rapid reestablishment of peregrine falcons on Kasatochi was likely facilitated by behavioral characteristics of peregrine falcons breeding in the Aleutian Islands, such as year-round residency and breeding site fidelity, the presence of an abundant food source (seabirds), and limited vegetation requirements by seabirds and falcons.

Alaska↗

Potential causes of shedding aggregations in prairie rattlesnakes

Aggregation is common across taxa and typically confers clear benefits to group members (e.g., allo-parenting, group defense, thermoregulation, access to resources). But aggregation can also be costly. The mechanisms that underpin aggregation—and the cues that elicit it—inform our understanding of how animals resolve tradeoffs among selection pressures. Snakes sometimes form conspicuous aggregations associated with hibernation, gestation, or parturition. Aggregation during ecdysis has also been described in some species, but infrequent observations and the synchronicity of ecdysis have confounded attempts to deduce the mechanism(s) responsible for the behavior. We documented aggregation during ecdysis in a population of asynchronously shedding prairie rattlesnakes ( Crotalus viridis ) and tested predictions generated from four hypotheses proposed to explain this behavior. We found that individuals undergoing ecdysis were more likely to aggregate. Our data did not support the hypothesis that rattlesnakes aggregate to improve their thermal efficiency, but we found some support for the reproductive facilitation and thermal landscape hypotheses as possible explanations for aggregation during ecdysis.

Wyoming↗

Integrating biology, field logistics, and simulations to optimize parameter estimation for imperiled species

Conservation of imperiled species often requires knowledge of vital rates and population dynamics. However, these can be difficult to estimate for rare species and small populations. This problem is further exacerbated when individuals are not available for detection during some surveys due to limited access, delaying surveys and creating mismatches between the breeding behavior and survey timing. Here we use simulations to explore the impacts of this issue using four hypothetical boreal toad ( Anaxyrus boreas boreas ) populations, representing combinations of logistical access (accessible, inaccessible) and breeding behavior (synchronous, asynchronous). We examine the bias and precision of survival and breeding probability estimates generated by survey designs that differ in effort and timing for these populations. Our findings indicate that the logistical access of a site and mismatch between the breeding behavior and survey design can greatly limit the ability to yield accurate and precise estimates of survival and breeding probabilities. Simulations similar to what we have performed can help researchers determine an optimal survey design(s) for their system before initiating sampling efforts.

Ecological Modelling↗

Fuels guide and database for intact and invaded big sagebrush (Artemisia tridentata) ecological sites—User manual

The Fuels Guide and Database (FGD) is intended to provide fuel loading and vegetation information for big sagebrush ( Artemisia tridentata ) ecological sites in the Morley Nelson Snake River Birds of Prey National Conservation Area (hereinafter the NCA) in southern Idaho. Sagebrush ecosystems in the NCA and throughout much of the Great Basin are highly influenced by non-native plants that alter successional trajectories and promote frequent wildfires, especially due to fine-fuel loadings that are highly variable over time and space. These dynamic fuel conditions can increase uncertainty when attempting to project fire risk and fire behavior. The FGD was developed to help quantify and assess these dynamic fuel loadings, and it provides access to fuels data across a range of conditions, from relatively intact sagebrush-bunchgrass communities to degraded communities dominated by nonnative annual grasses and forbs. The FGD can be queried for a variety of environmental conditions, and it provides tabular data, reports, and photographic records of fuels based on user queries. This report describes the FGD, including overall data content and data-collection methods, as well as instructions for installing and using the database.

Data Series↗

The Life of P: A biogeochemical and sociopolitical challenge in the Everglades

• Phosphorus (P) is an essential element for all life forms, yet to understand its life cycle and impact we need to grasp not only the biogeochemical life of P, but also how P intersects with human activities and values. • Phosphorus is the limiting nutrient in the oligotrophic Everglades ecosystem. Thus, the anthropogenic addition of P to the landscape and its subsequent transport, transformation, and persistence throughout the Everglades are critical to both the visual appearance and the ecological integrity of the Everglades ecosystem. • Legal and political attention to P in the Everglades have created a powerful social-cultural legacy. This legacy includes the continuing influence of P in Everglades restoration, the dissemination of knowledge about P cycling and mitigation among rural and urban Everglades residents, and social and political realignment of management priorities. • Management decisions based on legal mandates require not only robust scientific data about P-related processes, but also information on the behavior, well-being, and political decisions of tribal nations, agricultural enterprises, anglers, and others who use the South Florida landscape.

Florida↗

Complex patterns of genetic population structure in the mouthbrooding marine catfish, Bagre marinus, in the Gulf of Mexico and U.S. Atlantic

Patterns of genetic variation reflect interactions among microevolutionary forces that vary in strength with changing demography. Here, patterns of variation within and among samples of the mouthbrooding gafftopsail catfish ( Bagre marinus , Family Ariidae) captured in the U.S. Atlantic and throughout the Gulf of Mexico were analyzed using genomics to generate neutral and non-neutral SNP data sets. Because genomic resources are lacking for ariids, linkage disequilibrium network analysis was used to examine patterns of putatively adaptive variation. Finally, historical demographic parameters were estimated from site frequency spectra. The results show four differentiated groups, corresponding to the (1) U.S. Atlantic, and the (2) northeastern, (3) northwestern, and (4) southern Gulf of Mexico. The non-neutral data presented two contrasting signals of structure, one due to increases in diversity moving west to east and north to south, and another to increased heterozygosity in the Atlantic. Demographic analysis suggested that recently reduced long-term effective population size in the Atlantic is likely an important driver of patterns of genetic variation and is consistent with a known reduction in population size potentially due to an epizootic. Overall, patterns of genetic variation resemble that of other fishes that use the same estuarine habitats as nurseries, regardless of the presence/absence of a larval phase, supporting the idea that adult/juvenile behavior and habitat are important predictors of contemporary patterns of genetic structure.

Alabama, Florida, Louisiana, Mississippi, Texas↗

Transglobal spread of an ecologically relevant sea urchin parasite

Mass mortality of the dominant coral reef herbivore Diadema antillarum in the Caribbean in the early 1980s contributed to a persistent phase shift from coral- to algal-dominated reefs. In 2022, a scuticociliate most closely related to Philaster apodigitiformis caused further mass mortality of D. antillarum across the Caribbean, leading to >95% mortality at affected sites. Mortality was also reported in the related species Diadema setosum in the Mediterranean in 2022, though the causative agent of the Mediterranean outbreak has not yet been determined. In April 2023, mass mortality of Diadema setosum occurred along the Sultanate of Oman's coastline. Urchins displayed signs compatible with scuticociliatosis including abnormal behavior, drooping and loss of spines, followed by tissue necrosis and death. Here we report the detection of an 18S rRNA gene sequence in abnormal urchins from Muscat, Oman, that is identical to the Philaster strain responsible for D. antillarum mass mortality in the Caribbean. We also show that scuticociliatosis signs can be elicited in Diadema setosum by experimental challenge with the cultivated Philaster strain associated with Caribbean scuticociliatosis. These results demonstrate the Philaster sp. associated with D. antillarum mass mortality has rapidly spread to geographically distant coral reefs, compelling global-scale awareness and monitoring for this devastating condition through field surveys, microscopy, and molecular microbiological approaches, and prompting investigation of long-range transmission mechanisms.

ISME Journal↗

Biology and impacts of Pacific Islands invasive species. 14. Sus scrofa the feral pig (Artiodactyla: Suidae)

Feral pigs ( Sus scrofa L.) are perhaps the most abundant, widespread, and economically significant large introduced vertebrate across the Pacific island region. Unlike many other nonnative invasive species, feral pigs have both cultural and recreational importance in the region, complicating their management. Today, Pacific island feral pigs are a mixture of several strains of domestic swine, Asiatic wild boar, and European wild boar. Due to their generalist diet and rooting behavior, feral pigs alter soils and watersheds and negatively impact native and nonnative flora and fauna. As a result, feral pigs have played a role in the extinction of several species of plants and animals on Pacific islands and have negative effects on both ecotourism and agricultural industries in the region. Despite numerous published studies on feral pigs in the Pacific island region, of which the majority include systematic analyses of original empirical data, some fundamental aspects of feral pig ecology remain poorly characterized, at least partly due to the remote and inaccessible environments that they often inhabit. To address these knowledge gaps, effort should be made to integrate research conducted outside the Pacific island region into local management strategies. This review summarizes the origins, history, ecology, environmental effects, and current management of feral pigs in the Pacific island region; integrates regional scientific findings with those of other insular and continental systems; and identifies current knowledge gaps requiring further research to inform the ecology and management of this impactful invasive species.

Hawai'i↗

Latitudinal gradients and sex differences in morphology of the Black Oystercatcher (Haematopus bachmani)

Environment and behavior are widely understood to affect bird morphology, which can lead to differences among subspecies or populations within a wide-ranging species. Several patterns of latitudinal gradients in morphology have been described, though Allen's and Bergmann's rules are the most well-known and have been tested and confirmed across a diversity of taxa and species. These state that individuals at higher latitudes will have larger bodies (Bergmann's Rule) but smaller extremities (Allen's Rule) to conserve heat in colder climates. Migratory behavior also can influence avian morphology, particularly wing shape, where migratory birds tend to have longer, more pointed wings than residents. The Black Oystercatcher ( Haematopus bachmani ) is a large, partially migratory shorebird species restricted to intertidal habitats and distributed from Alaska to Baja California, spanning about 35° of latitude. A large proportion of Black Oystercatchers that breed in Alaska are migratory, where nearly all individuals breeding in British Columbia through the southern end of their range remain resident through the annual cycle. Their broad latitudinal range and diversity in migratory behavior may drive geographic variation in morphology. Here we evaluate three explanations for geographic variation in morphology of the Black Oystercatcher using data from seven sites across two regions: Alaska and British Columbia. We found evidence consistent with Allen's but not Bergmann's rule; birds in Alaska have shorter bills than those in British Columbia, and these findings held when controlling for body size using wing length. Despite regional differences in migratory behavior, we detected no difference in the wing shape of birds in Alaska and British Columbia. Differences between sexes and among sites suggest that multiple factors drive patterns of morphological variation in the Black Oystercatcher.

Alaska↗

Do mule deer surf peaks in forage quality while on summer range?

Many animals track ephemeral peaks in food abundance and quality that propagate across landscapes. Migrating ungulates, in particular, track waves of newly emerging plants from low-elevation winter ranges to high-elevation summer ranges—known as “green-wave surfing.” Because plants lose crude protein and gain insoluble fiber with maturation, ruminants are expected to exploit peaks in forage quality among individual plants (i.e., Forage Maturation Hypothesis). Although ample evidence supports the long-standing hypothesis that migratory ungulates surf peaks in forage quality during migration, the hypothesis that ungulates track peaks in forage quality at a small scale (i.e., microsurf while on summer range) remains less known. We studied a partially migratory population of mule deer ( Odocoileus hemionus ) in Wyoming, USA, to understand whether temperate ungulates optimize the use of high-quality forage as plants grow and senesce on disparate summer ranges. Specifically, we evaluated how crude protein, digestible energy, and relative abundance changed throughout the growing season and whether deer altered their diet to reflect species-specific changes in plant phenology. In support of the Forage Maturation Hypothesis, forage quality declined as large-scale patterns of phenology progressed away from a remotely sensed metric of peak green-up for most plant species on the summer ranges of deer that migrated short (<50 km), medium (50–130 km), and long distances (>130 km). Declining rates in forage quality among plant species were heterogeneous, providing deer with the phenological diversity required to microsurf. Deer changed their diet throughout the growing season and prioritized the consumption of some plants, including Rosa woodsii and Purshia tridentata , as the rank of forage quality increased ( p < 0.01). In light of the complexities common to studies on foraging behavior, our findings suggest that deer may have some potential to microsurf on summer range when heterogeneity in resource phenology is prevalent. Moreover, our findings validate the accuracy of remote sensing in quantifying peak forage quality for plants within sagebrush shrublands and montane habitats.

Wyoming↗

Noose carpets: A novel method to capture rails

Avian research may require investigators to capture birds for many reasons, including to collect measurements and attach leg bands or transmitters. The effectiveness and efficiency of capture techniques varies by species, habitat, and time of year. Rails are particularly challenging to capture because of their secretive behavior and the dense vegetation they inhabit. As such, basic natural history questions for many rail species remain unanswered. We paired audio lures with modified noose carpets to capture and study 69 Yuma Ridgway's rails ( Rallus obsoletus yumanensis ) in the southwestern United States during 2016–2018. We compared results with other more commonly used capture methods, and our results show that noose carpets paired with audio lures can be an effective tool to capture rails, thereby facilitating studies of their ecology and life history. Noose carpets are easy to use, cheap to build and maintain, and effective over a wide range of conditions. This method could be used to capture rails other than the Yuma Ridgway's rail by adjusting the noose size, noose line weight, and audio lures to match the target species. Published 2020. This article is a U.S. Government work and is in the public domain in the USA.

Arizona, California↗

Recent history of Saker Falcon studies in Mongolia

This report clarifies recent studies on the saker falcon in Mongolia. In the last five years, three museum studies appeared on the Altay falcon (is it a gyrfalcon, saker or separate species). These showed that all of the most distinct Central Asian summer specimens were from only two mountain ranges. However, there is a continuum between sakers and Altay falcons and the two should be considered synonymous. Of field studies focused on the saker in Mongolia, the first was conducted by Baumgart in the 1970s. The next (my own) began in 1994 with a Mongolia-wide study of the nesting ecology of the species. From that study, we now have over 150 breeding territories with over 200 eyries described. Resulting publications dealt with reproductive performance of the falcons and unusual breeding situations. Not only were many nests in odd situations, many were also composed mainly of manmade materials. We found that birds sometimes became entangled in twine and cloth in such nests and either died or would have died without our intervention. Our work also led to observations of novel social behavior including the first documentation of siblicide for any falcon and the description of a new falcon display, splayed-toes-flight. Aware that saker populations in Kazakhstan and elsewhere were plummeting due to over harvest for falconry and seeing a growing harvest in Mongolia, in 1997, we began efforts to build artificial eyries for the falcons. To date, over 150 artificial eyries have been created. In 1998, there was a great expansion of saker field work after the National Avian Research Center (NARC) of the United Arab Emirates became directly involved in hiring Mongol students and scientists. Those efforts resulted in an immediate estimate of the breeding population. Continuing work promises to provide good information on home range, food habits, productivity and other topics.

Falco↗

Dynamic risk from Mexican wolves and mountain lions influences elk foraging behavior

Foraging time is a major component of ungulate activity budgets but can be limited by anti-predator behaviors (e.g., vigilance). Multitasking can reduce the nutritional costs of vigilance under heightened predation risk, but this may depend on the response of prey to risk from multiple predators across a complex spatiotemporal landscape. Mexican gray wolves ( Canis lupus baileyi ) and mountain lions ( Puma concolor ) are primary predators for elk ( Cervus canadensis ) in the Mexican wolf experimental population area in east-central Arizona and west-central New Mexico. We observed elk foraging across varying levels of wolf risk throughout all seasons and diel periods to quantify proportions of foraging, intense vigilance, and multitasking at the individual and herd levels. We quantified encounter and kill risk from Mexican wolves and mountain lions using habitat selection functions and utilization distributions. We modeled elk behaviors as functions of predicted risk for both predators in addition to temporal and environmental covariates and accounted for human presence. Our results indicate that individual elk reduced foraging in areas with higher predicted risk from Mexican wolves or mountain lions and increased intense vigilance and multitasking in areas with higher wolf risk. A reduction in the proportion of bedded elk in the herd during all diel periods under increased wolf risk supports previous findings. These results also suggest that elk compensate for higher intense vigilance and reduced foraging during foraging bouts by increasing cumulative foraging bouts per day at the cost of resting. Additionally, the probability of multitasking for individuals depended on an interaction between short- and long-term wolf risk, and the likelihood of intense vigilance was highest under the greatest combined spatial and temporal risk from wolves. This research provides insight into the fine-scale and complex behavioral responses of elk to their primary predators and implies a need for researchers to consider these non-consumptive effects in future studies of predator–prey dynamics.

Arizona, New Mexico↗

Energetic costs of mange in wolves estimated from infrared thermography

Parasites, by definition, extract energy from their hosts and thus affect trophic and food web dynamics even when the parasite may have limited effects on host population size. We studied the energetic costs of mange ( Sarcoptes scabiei ) in wolves ( Canis lupus ) using thermal cameras to estimate heat losses associated with compromised insulation during the winter. We combined the field data of known, naturally infected wolves with data set on captive wolves with shaved patches of fur as a positive control to simulate mange-induced hair loss. We predict that during the winter in Montana, more severe mange infection increases heat loss by around 5.2 to 12 MJ per night (1240 to 2850 kcal, or a 65% to 78% increase) for small and large wolves, respectively accounting for wind effects. To maintain body temperature would require a significant proportion of a healthy wolf's total daily energy demands (18-22 MJ/day). We also predict how these thermal costs may increase in colder climates by comparing our predictions in Bozeman, Montana to those from a place with lower ambient temperatures (Fairbanks, Alaska). Contrary to our expectations, the 14&deg;C differential between these regions was not as important as the potential differences in wind speed. These large increases in energetic demands can be mitigated by either increasing consumption rates or decreasing other energy demands. Data from GPS-collared wolves indicated that healthy wolves move, on average, 17 km per day, which was reduced by 1.5, 1.8 and 6.5 km for light, medium, and severe hair loss. In addition, the wolf with the most hair loss was less active at night and more active during the day, which is the converse of the movement patterns of healthy wolves. At the individual level mange infections create significant energy demands and altered behavioral patterns, this may have cascading effects on prey consumption rates, food web dynamics, predator-prey interactions, and scavenger communities.

Montana↗

Achieving interpretable machine learning by functional decomposition of black-box models into explainable predictor effects

Machine learning (ML) models are often based on complex black-box architectures that are difficult to interpret. This interpretability problem can hinder the use of ML in fields like medicine, ecology, and insurance, and has boosted research in interpretable machine learning (IML). Here, we propose a novel approach for the functional decomposition of black-box predictions, which is a core concept of IML. This approach replaces the prediction function with a surrogate model consisting of simpler subfunctions, providing insights into the direction and strength of the main feature contributions and their interactions. Our method is based on a concept termed “stacked orthogonality”, which ensures that the main effects capture as much functional behavior as possible. To compute the subfunctions, we combine neural additive modeling with an efficient post-hoc orthogonalization procedure. Our method yielded plausible results in an analysis of stream biological condition in the Chesapeake Bay watershed (United States).

Chesapeake Bay watershed↗