Defining dual-axis landscape gradients of human influence for studying ecological processes.

Ecological processes are strongly shaped by human landscape modification, and understanding the reciprocal relationship between ecosystems and modified landscapes is critical for informed conservation. Single axis measures of spatial heterogeneity proliferate in the contemporary gradient ecology literature, though they are unlikely to capture the complexity of ecological responses. Here, we develop a standardized approach for defining multi-dimensional gradients of human influence in heterogeneous landscapes and demonstrate this approach to analyze landscape characteristics of ten ecologically distinct US cities. Using occupancy data of a common human-adaptive songbird collected in each of the cities, we then use our dual-axis gradients to evaluate the utility of our approach. Spatial analysis of landscapes surrounding ten US cities revealed two important axes of variation that are intuitively consistent with the characteristics of multi-use landscapes, but are often confounded in single axis gradients. These were, a hard-to-soft gradient, representing transition from developed areas to non-structural soft areas; and brown-to-green, differentiating between two dominant types of soft landscapes: agriculture (brown) and natural areas (green). Analysis of American robin occurrence data demonstrated that occupancy responds to both hard-to-soft (decreasing with development intensity) and brown-to-green gradient (increasing with more natural area). Overall, our results reveal striking consistency in the dominant sources of variation across ten geographically distinct cities and suggests that our approach advances how we relate variation in ecological responses to human influence. Our case study demonstrates this: robins show a remarkably consistent response to a gradient differentiating agricultural and natural areas, but city-specific responses to the more traditional gradient of development intensity, which would be overlooked with a single gradient approach. Managing ecological communities in human dominated landscapes is extremely challenging due to a lack of standardized approaches and a general understanding of how socio-ecological systems function, and our approach offers promising solutions.

Introduction

Rapid expansion of the global human population has led to increasing concern for natural systems and biodiversity. Anthropogenic landscape modification profoundly influences resource availability and habitat quality, which in turn, determines patterns of species distribution and abundance [1]. Given the explicit link between patterns of landscape structure and ecological processes, and the extent of human modification to the landscape, informed conservation and ecosystem management requires reliable descriptors of landscape heterogeneity gradients with an anthropogenic focus [2, 3]. Nevertheless, well documented variability in the quality, complexity, and ecological relevance of quantitative measurements of landscape structure have contributed to a lack of a general and scalable understanding of how ecological processes respond to landscape heterogeneity, particularly along gradients of human modification [3-6].

The need for ecologically meaningful measures of landscape heterogeneity (i.e., composition and configuration of landscape features) to understand drivers of ecosystem responses is well recognized [7], and over time numerous conceptual, theoretical, and applied approaches have been posited [4, 5, 8, 9]. These approaches range from the patch mosaic (fragmentation) paradigm, which, while valuable in some contexts, is arguably overly simple in heterogenous landscapes [10-12], to various metrics of patch complexity and distribution [13]. Efforts to improve the ecological relevance and realism of landscape metrics has led to the development of models thought to better represent the continuous nature of landscape heterogeneity and ecological processes by extending the patch-centered perspective to incorporate the composition of the surrounding landscape [14]. Regardless of the metrics used, successful integration of ideas in spatial ecology across systems and scales requires an improved appreciation for what landscape descriptors are measuring, and how they relate to ecological processes [15]. That is, reliable and accurate measures of landscape heterogeneity are a prerequisite to understanding patterns of ecological response across scales.

Efforts to understand and quantify ecological responses across anthropogenic gradients has resulted in some general, though equivocal, predictions about patterns of ecological response to spatial heterogeneity in human dominated landscapes. For instance, a negative relationship between species richness and human disturbance has been demonstrated in birds [16, 17], invertebrates [18, 19], plants [20, 21], and other taxa [22, 23]. Moreover, this relationship is often non-linear, with a peak in richness in areas of intermediate human modification [24, 25]. At the species level, however, responses vary, and depend on the ecology of the species in question [26-28]. While fragmentation and human population density have been linked to decreases in movement and home range size in many species [29-32], much of the literature suggests no relationship [33-35], or uncertain relationships [36, 37] between a range of ecological processes (e.g., population size, species distribution) and landscape change. These apparent contradictions suggest that measured responses to gradients of landscape heterogeneity are context or locale specific and has led to calls for improved measures of human-dominated landscapes that move towards a more general understanding of ecological dynamics in human-dominated ecosystems [6, 38].

Attempts to improve the applicability and scalability of landscape metrics used in ecological analyses has led to almost exclusive use of one-dimensional gradients of variation (e.g., percent impervious surface), even though landscape heterogeneity, and in particular the myriad ways humans alter landscapes, is multi-dimensional [39, 40]. Highly dimensional landscapes, when compressed into one-dimensional descriptors, are likely to fall short in terms of ecological realism, i.e., the landscape as perceived by a species or community, limiting the ability to infer links between landscape patterns and ecological processes, with important consequences regarding how ecosystem processes are understood and managed in the Anthropocene. We propose an extension of the typical one-dimensional approach, which involves the identification of multiple axes of landscape heterogeneity in the context of human influence.

In this paper, we develop a multi-dimensional approach to defining landscape heterogeneity that can be used for making inferences about species distributions in human dominated landscapes. We demonstrate the generality of our multi-dimensional gradient approach by jointly analyzing urban-exurban landscapes in ten geographically and ecologically distinct US cities, identifying two significant and biologically relevant axes of variation. We demonstrate the utility of our approach in a case study analysis of American robin (Turdus migratorious) occupancy. Specifically, by jointly analyzing detection—non-detection data from the same ten landscapes, we investigate continental-scale consistencies in species responses to two gradients of human influence.

Methods

We selected ten geographically distinct medium sized cities (population between 200,000 and 500,000), widely distributed throughout the contiguous United States, representing the Level I ecoregions as defined by the U.S. Environmental Protection Agency [41]. These were Worcester (Massachusetts), Lexington (Kentucky), Jackson (Mississippi), Lincoln (Nebraska), Lubbock (Texas), Salt Lake City (Utah), Albuquerque (New Mexico), Bakersfield (California), Portland (Oregon), and Spokane (Washington, Fig 1). For each city, we extracted 30-m resolution landcover data from the freely available 2016 National Land Cover Database [42] for a 50-by-50 kilometer window surrounding the city center (coordinates extracted from www.latlong.net). This spatial extent extends well into exurban regions and thus represented the full extent of landscape heterogeneity for each city. To test for sensitivity to the extent, we repeated the analysis at alternative windows and found no difference in our results (Effects spatial extent: 30 x 30 km city window in S1 File).

Fig 1

Map of 10 study cities.

Map showing the locations of all study cities for the landscape quantification and ecological case study. Background colors represent unique Level 1 EPA Eco-Regions. Study cities are represented by numbered red points. 1—Portland, OR, 2—Bakersfield, CA, 3—Spokane, WA, 4—Salt Lake City, UT, 5—Albuquerque, NM, 6—Lubbock, TX, 7—Lincoln, NE, 8—Jackson, MS, 9—Lexington, KY, and 10—Worcester, MA. Ecoregion GIS data was sourced from the US EPA—Ecoregion spatial database (https://www.epa.gov/eco-research/ecoregions-north-america). Map was produced using the package ‘map’ in R.

Landscape composition was fairly evenly split between three dominant lands cover categories when aggregated among all cities: developed (20.03%), forests (23.98%), and agriculture (31.28%) and contained fifteen of the nineteen Anderson Land-Cover classes used by the NLCD. The remaining four (‘Perennial Ice-Snow’, ‘Dwarf Scrub’, ‘Sedge-Herbaceous’, ‘Lichen’) are restricted to Alaska or high elevation locations. At the individual city level, landscape composition was more variable; forested classes dominated Worcester and Spokane (39.08%, 30.22%), Albuquerque was largely scrubland (46.81%), Lexington dominated by pasture (62.31%), and agriculture in Lincoln, Bakersfield, and Lubbock (47.61%, 42.2%, 72.15%). Details for each city are provided in Table 1.

Table 1

Table of study cities.

City, State	Population	Level I Ecoregion	Open Water	Devel.	Forests	Scrub Grass	Crop Pasture
Worcester, MA	185,877	ER5 –Northern Forests	3.33	23.15	65.51	2.54	6.11
Spokane, WA	208,916	ER6 –NW Forested Mountains	1.23	14.46	31.31	31.91	20.62
Salt Lake City, UT	200,591	ER6 –NW Forested Mountains	11.02	23.00	36.37	23.34	5.29
Portland, OR	583,776	ER7 –Marine West Coast Forest	3.12	37.06	23.43	7.85	28.54
Lexington, KY	323,780	ER8 –Eastern Temperate Forests	0.56	15.31	16.65	1.03	66.44
Jackson, MS	164,422	ER8 –Eastern Temperate Forests	4.85	30.11	43.29	12.16	18.87
Lubbock, TX	255,885	ER9 –Great Plains	0.15	12.58	0.23	14.89	72.15
Lincoln, NE	287,401	ER9 –Great Plains	1.54	13.02	5.86	29.4	50.82
Albuquerque, NM	560,218	ER10 –North American Deserts	0.23	17.66	15.96	67.71	3.37
Bakersfield, CA	383,679	ER11 –Mediterranean	0.51	13.96	1.19	37.93	46.41

List of ten urban-exurban regions used for landscape comparisons, including population size (2010 census) and US-EPA Ecoregion. Values for land cover types represent the percent coverage in a given city.

Landscape analyses followed the landscape quantification framework outlined by Padilla and Sutherland [3]. Our decisions regarding the types of landscape features relevant for analysis, the data to represent those features, and the spatial scales of analysis were made to reflect a typical ecological analysis—definitions of, and justification for, these decisions are provided in Table 2. In general, the landscapes within which our cities were set were characterized by a mosaic of natural (forests and wetlands) and un-natural (crop and developed) land-cover categories which are captured well in the NLCD classification system.

Table 2

Landscape analysis decision table.

	Decision	Justification
1) Landscape Features	Physical land-cover and demographic land-use	‘Land-cover’ categories (i.e. forest, shrub) track changes in ‘natural’ landscapes, while ‘land-use’ (devel., crop) tracks the human footprint and approximate population density
2) Spatial Data	Remote-sensed, National Land Cover Data (2016)	NLCD land-cover data is readily available and is a consistent data-source to represent landscape features in all 10 study cities
3) Spatial Scale	500-m and 1,500-m Gaussian kernel	Spatial extent (50 x 50-km) chosen to capture sufficient spatial and ecological heterogeneity. Primary spatial grain (500-m kernel) selected to represent breeding home range of American robin. 1,500-m as a common scale in ecological research selected to compare effects of scale.

Decisions made within landscape gradient framework for analyzing urban landscapes in jointly across study cities and in the city-specific analysis. This follows the framework outlined in Padilla & Sutherland 2019. Justification provided here is in light of dual research goals. First, to quantify landscape pattern in 10 distinct cities, and second, to evaluate occupancy patterns of American robin in response to landscape gradients.

The NLCD is a 30-m resolution raster dataset where each landscape pixel is classified as a single cover type. Ecosystems are influenced both by characteristics of a fixed location, and by the local landscape context surrounding a given location [43]. Therefore, for each NLCD land-cover class, we extracted a binary raster surface (1 if focal class, 0 if otherwise), and to account for landscape surrounding a given location (i.e., landscape context) we computed the spatially weighted average for each pixel using a Gaussian kernel spatial smooth, resulting in a continuous surface ranging from 0 (no focal class within smoothing kernel) to 1 (smoothing neighborhood entirely focal class). The width of the kernel defines the spatial grain of analysis, and therefore should be selected with the research specific ecological process in mind [44]. We selected a 500-m smoothing kernel for our analysis based on the typical breeding home range size of our case study focal species, the American robin [45]. We tested sensitivity of landscape quantification to this choice by replicating the analysis with a 1500-m spatial scale and found no effect of scale selection of downstream inference (Effects of smoothing scale in S1 File). All processing of the spatial data was conducted in R Version 3.5.3 [46] using the ‘raster’ [47], ‘FedData’[48] and ‘smoothie’ [49] packages.

To identify dominant patterns of variation in these landscapes, we used Principal Components Analysis (PCA). PCA is one of several methods for summarizing a large number of potentially correlated variables into fewer uncorrelated axes of variation (others include factor analysis, non-metric multidimensional scaling, correspondence analysis), and it is particularly well suited to exploratory ordination and gradient analysis [50]. Using a matrix of class-specific smoothed landscape variables, we conducted PCA on the data for all cities combined. Dominant principal components were identified and selected based on a cumulative weight cut-off of the broken stick method, which retains components that explain more variance than would expected than dividing variance randomly among all components [51]. These were used to produce a spatially explicit gradient of habitat heterogeneity based on the resolution of the input data, where the value for each pixel in the resulting raster surface is the PCA weighted average calculated as the sum of that pixel’s smoothed NLCD values multiplied by the corresponding PC weight for each NLCD value. We also conducted this analysis for each city independently in order to determine how well the combined (i.e., all cities) gradients described city-specific gradients. Output for our PCA analyses are reported in the Results section under Landscape Gradient Analysis.

Ecological case study

We evaluated the utility of multi-dimensional landscape heterogeneity gradients for ecological applications using a real-world case study. Specifically, using occupancy modelling we tested whether simultaneous consideration of multiple landscape gradients alters inferences about ecological responses relative to the traditional single-gradient approach. We analyzed American robin detection-nondetection data under an occupancy modelling framework using the gradients as covariates. We selected the American robin because it is a widespread generalist species, present in all ten focal cities, and because it is widely considered to be human-adaptive.

Robin detection histories were analyzed using a single-season hierarchical occupancy model, which estimates site occupancy probabilities while accounting for imperfect detection [52]. Stationary, complete checklists in which non-reporting of a species assumed to be non-detection, from surveys conducted from April 1st through September 30th 2018 were extracted from the eBird online database [53] using the R package ‘auk’ [54]. In this analysis, detection data from all cities were pooled in a single analysis. Because there was substantial variation in the number of eBird locations in each city (i.e., each unique fixed eBird survey site), and to improve balance and reduce regional bias in sample size, the data were randomly thinned to a maximum of 250 locations (Table 1 in S2 File).

The standard occupancy model consists of two sub-models: a logit-linear model describing site- and occasion-specific detection probability (p), which can be modelled using site- and occasion-specific covariates, and a second logit-linear model describing site-specific occupancy probability (ψ), that can be modelled using site-specific covariates. To account for variation in detection, we considered the following covariates: city (categorical factor), sampling date, and date2 to allow for peaks or troughs in detection, and site-specific landscape gradient values. Sampling date was scaled (0–18) such that a one unit increase in the date variable represented 10 calendar days, which facilitated parameter interpretation and model convergence. A total of 26 possible detection models were considered, which included all additive combinations and only single interaction terms (Table 2 in S2 File). For occupancy, we included the effect of city, again as a factor, each of the site-specific landscape gradient values, and all combinations of city-gradient interactions for a total of 16 candidate models (Table 2 in S2 File).

We adopted a two-stage modeling approach whereby we fit and compared all possible combinations of detection covariates, each with the ‘global’ (most complex) model for the occupancy component [55]. Using Akaike’s Information Criterion (AIC) to rank models, the best supported model for detection was carried over to the second stage, where we compared competing models for occupancy. Finally, the model selected for inference was validated by examining model residuals and performing goodness of fit tests. Occupancy analysis was conducted in the package ‘unmarked’ [56], while AIC model selection and goodness of fit tests were done using the ‘AICcmodavg’ package [57]. All analyses, were conducted in R Version 3.5.3 [46].

Results

Landscape gradient analysis

Principal components analysis of the combined (i.e., all cities) landscape data identified three axes of variation, explaining 37.1% of the cumulative variance in the data (Table 3). When each city was considered independently, the same three axes explained between 42.60 and 54.89% of the variance (Table 1 in S1 File), demonstrating the scalability of emergent landscape gradients across scales. However, using the broken stick method [51], only the first two axes exceeded the 22.1% cumulative variance threshold for combined and city-specific analyses. The principal component explaining the largest proportion of data variation for the combined data (16.7%) was strongly negative for developed land-cover classes, with neutral or positive loadings for forested, open, and agricultural classes (Table 3). Developed classes are characterized by a high degree of impervious surface, buildings, and associated human population density, whereas the others are predominantly non-impervious natural (wetlands) or un-natural (pasture) landscapes. Thus, this first descriptor of landscape pattern can be interpreted as a transition from hard (characterized by impervious and human presence) to soft (unpaved natural or agricultural), which we refer to as a hard-to-soft gradient.

Table 3

Dominant principal component axes.

NLCD Layer		Obs. Freq.	PC1	PC2	PC3
Std.Dev.			1.581	1.295	1.182
Variance Explained (%)			16.7	11.1	9.3
Water	11—Open Water	2.70	0.042	0.030	-0.04
Developed	21—Developed Open	6.49	-0.360	0.055	0.017
	22—Developed Low Intensity	6.72	-0.545	0.047	-0.040
	23—Developed Medium Intensity	4.78	-0.553	0.015	-0.125
	24—Developed High Intensity	1.61	-0.392	-0.007	-0.148
Barren	31—Barren Land	0.79	0.039	-0.017	-0.116
Forest	41—Forest Deciduous	10.17	0.119	0.469	0.171
	42—Forest Evergreen	8.96	0.154	0.269	-0.349
	43—Forest Mixed	2.14	0.087	0.433	0.001
Shrubland	52—Scrub/Shrub	10.56	0.151	-0.012	-0.554
Herbaceous	71—Grassland/Herbaceous	10.85	0.140	-0.339	-0.282
Cultivated	81—Pasture/Hay	11.25	0.053	0.082	0.446
	82—Cultivated Crops	19.41	0.119	-0.491	0.387
Wetlands	90—Woody Wetlands	2.81	0.043	0.378	0.157
	95—Herbaceous Wetlands	0.77	0.032	0.075	-0.039

Dominant Principal Component axes. The Obs. Freq. column displays the percent composition of each land cover category in all cities combined. For each principal component, standard deviation, percent variance explained, and rotated variable loadings are displayed. Variables with a strong weight are in bold. The first two axes were selected because cumulative variance exceeded the 22.1% broken stick threshold.

The second principal component explained 11.1% of the variation and showed a strong differentiation between the land use classes at the soft end of the hard-to-soft gradient. Specifically, this axis distinguishes between human modified but un-developed areas (cultivated croplands) and more natural areas (forests or wetlands). This axis is intuitively interpretable as a shift from modified agricultural landscapes, to un-developed natural regions, or, brown-to-green. While the hard-to-soft axis does not distinguish between dominant types of soft landscapes, the second accounts for this variation between brown and green regions and is a valuable counterpart to component one producing a triangular distribution (Fig 2).

Fig 2

Conceptual diagram of multi-dimensional landscape gradient: A conceptual description of the triangular distribution captured by a multi-dimensional landscape definition that differentiates between urban, agricultural, and natural portions of the landscape along dual axes of variation.

Hard and soft portions of the landscape are sorted along the vertical axis, while brown and green regions along the horizontal. This results in a multi-dimensional perspective where heterogeneity is maximized at the center of both axes.

The third principal component explained 9.3% of the total variation and was not retained to produce a gradient surface as the cumulative variance of the first two principal components exceeded the broken stick cutoff. However, it is interesting in that like PC2, the third principal component reflected a divergence between modified and un-modified undeveloped areas. While PC2 differentiated natural deciduous and mixed forests from modified croplands, the third axis is a gradient from evergreen forests and scrub, to pastures (Table 3). Both PC2 and PC3, therefore, can be interpreted as brown-to-green in different habitat and land-use types.

Due to the ecological complementarity of the two dominant components, and our focus on highlighting the value of simultaneously considering multiple dimensions of human influence, our approach considers these axes jointly. However, it is worth noting that on their own, these gradients are analogous to traditional approaches that consider single gradients in isolation. The hard-to-soft gradient is consistent with traditional urban gradients focusing on the built environment (e.g., percent impervious surface or housing density) [3, 6, 58], or it’s complement, percent forest cover. The more agricultural brown-to-green gradient, though less common in urban ecology, has been used in agro-ecological investigations [59, 60]. Our approach allows us to investigate ecological responses to both important characteristics of human influence simultaneously.

As a test of whether these axes were consistent locally and at varying spatial extents, we conducted the same analyses of NLCD data for each city independently as well as jointly using a 30x30 km window. Both city specific, and 30x30 km analyses revealed the same dominant axes of variation as the 50x50 km combined analysis. As expected, the component weights of NLCD classes and absolute values of axes differed, nevertheless, interpretation of these axes remained consistent (Effects of smoothing scale: 1,500-m scale in S1 File).

Our robin analysis included data from a total of 1,703 sampling locations (sites) across all cities (min: 31 in Bakersfield, max: 250 in Worcester, Albuquerque, Portland, and Salt Lake City). There were a total of 5,779 sampling visits across all cities, with a mean number of visits per site of 1.95 (range: 1–172, Table 1 in S2 File). The overall proportion of sites with a minimum of one observation (i.e., naïve occupancy) was 0.43, which varied by city from 0.387 in Spokane, to 0.482 in Bakersfield (Table 1 in S2 File).

Of the 26 detection models considered, only eight converged, largely due to the complex model structure. The AIC-top model (AIC wt = 1.0) included additive effects of both landscape gradients, a quadratic effect of date, and a city by date interaction term (Table 4). In the second step, we used the best supported detection model to evaluate 16 candidate occupancy models. Here, a single model held the majority of support (AICwt = 0.91, Table 4) and included the effects of both landscape gradients, city, and an interaction between city and the hard-to-soft gradient. The second model was identical to the top model apart from the inclusion of one additional term, the interaction between city and brown-to-green. Given the lack of support for the additional terms, as indicated by the model ranking [61], model evaluation and inference that follows is based on the top model. Examination of model residuals and a Chi-Square goodness of fit test showed adequate model fit.

Table 4

Model selection table.

	Detection Model Structure	K	AIC c	ΔAIC c	AICwt	- LogLik
1	~ city*date+date2+HS+BG~ ψ	63	6544.02	0.0	1	-3206.45
2	~ city*date+HS+BG~ ψ	62	6611.83	67.81	0	-3241.44
3	~ city*HS~ ψ	60	6998.33	454.31	0	-3436.84
4	~ city~ ψ	43	7001.75	457.73	0	-3456.69
5	~ date~ ψ	42	7038.66	494.64	0	-3476.20
6	~ BG~ ψ	42	7450.53	906.51	0	-3682.13
7	~ 1 ~ ψ	41	7450.69	906.67	0	-3683.27
8	~ HS~ ψ	42	7452.32	908.30	0	-3683.03
	Occupancy Model Structure	K	AIC c	ΔAIC c	AICwt	- LogLik
1	~ p  ~ city*HS+BG	44	6530.44	0	0.91	-3219.98
2	~ p  ~ city*(HS+BG)	53	6535.21	4.77	0.09	-3212.80
3	~ p  ~ city*HS	43	6542.84	12.40	0	-3227.23
4	~ p  ~ city*(HS*BG)	63	6544.02	13.58	0	-3206.45
5	~ p  ~ city+HS*BG	36	6567.86	37.42	0	-3247.10
6	~ p  ~ city+HS+BG	35	6572.90	42.46	0	-3250.67
7	~ p  ~ city+HS	34	6578.90	48.46	0	-3254.71
8	~ p  ~ city*BG+HS	44	6580.24	49.80	0	-3244.88
9	~ p  ~ HS*BG	27	6580.55	50.11	0	-3262.81
10	~ p  ~ city*BG	43	6581.49	51.05	0	-3246.56
11	~ p  ~ city+BG	34	6582.03	51.59	0	-3256.27
12	~ p  ~ city	33	6587.86	57.42	0	-3260.23
13	~ p  ~ BG	25	6490.33	59.89	0	-3269.76
14	~ p  ~ HS+BG	26	6591.46	61.02	0	-3269.29
15	~ p  ~ HS	25	6600.95	70.51	0	-3275.07
16	~ p  ~ 1	24	6601.16	70.72	0	-3276.21

Model selection results for both detection and occupancy components of the American robin analysis based on sample size corrected AIC. K denotes the total number of parameters in the model and AICwt is the model weight. Detection was assessed with the global occupancy model and the best model for detection was used in all models for occupancy. Here, HS refers to the hard-to-soft gradient, while BG denotes brown-to-green.

There was a significant quadratic effect (estimate± SE) of survey date on detectability (-0.012± 0.002), such that detection probability increased, reached a peak, and declined. Robin detection varied significantly along the brown-to-green axis, with robins more likely to be observed in more ‘green’ landscapes (0.14± 0.05), and showed a negative relationship with hard-to-soft, though confidence intervals included zero (-0.38± 0.04). Date of peak detectability ranged from April 1st in Bakersfield (date = 0.0) to July 23rd in Portland (date = 11.3), while maximum detection probability ranged from 0.33 in Worcester, to 0.87 in Jackson (Fig 3; Table 3 in S2 File).

Fig 3

Detection probability of American robin: Robin Detection probability as a function of survey date for each city predicted from the top model.

Grey shaded area represents 95% confidence intervals and solid is the expected value. Date of peak detection probability varied between cities, but tended toward the start of the study period, which coincides with robin breeding behavior.

Robin occupancy varied by city and with both gradients. Holding both gradients at 0 (overall scaled average), robin occupancy ranged from a low of 0.47 (0.08) in Albuquerque, New Mexico to a high of 0.99 (0.074) in Jackson, Mississippi. Robin occupancy was positively related to the brown-to-green axis (0.52± 0.01), suggesting that robins are more likely to occur in more forested areas than in areas characterized as predominantly open or agricultural. This effect was universal across all cities. In contrast, and interestingly, direction and magnitude of the hard-to-soft gradient effect varied by city, i.e., the responses to the gradient describing the transitions from built to vegetative environments was specific to each city (Fig 4). For example, occupancy was positively associated with the hard-to-soft gradient in Spokane (1.61± 0.819), but negatively associated with hard-to-soft in Worcester (-1.72± 0.765).

Fig 4

Predicted robin occupancy along dual landscape gradients: Surface plots depicting robin occupancy on both the brown-to-green (y axis) and hard-to-soft (x axis) for each of the ten cities.

The color scale ranges from low occupancy (red) to high occupancy (blue). Variation in surface plots highlights the differences in landscape composition in study cities, and the variable response to urbanization along the hard-to-soft gradient.

Discussion

Analysis of spatially heterogeneous landscapes surrounding ten metropolitan regions revealed two statistically important and ecologically intuitive axes of variation, which offers an exciting alternative to the conventional one-dimensional approach to investigating ecological responses in human-dominated landscapes. Despite regional variation in landscape composition (Table 1 in S1 File) the dual-gradient approach we present here consistently distinguished between two distinct types of anthropogenic influences: a hard-to-soft gradient capturing a continuum of the built human environment, and a brown-to-green gradient capturing the human agricultural footprint (Fig 2). Our analysis shows that in addition to being fundamental properties of the landscape, considering these axes jointly provides ecological insight that would otherwise be overlooked using a single-axis approach (Table 3). This multi-dimensional perspective highlights the importance of considering the complexity of human-dominated landscapes and identifies a triangular distribution of human influence that presents an intuitive and generalizable framework for understanding patterns of ecological function and developing management strategies in human-dominated landscapes.

Landscape metrics that are adaptable to a variety of ecosystem contexts are needed to improve understanding of human-dominated ecosystems and effectively synthesize local and regional conservation efforts. Prior attempts to produce universal metrics for human footprint or urbanization have thus far failed to result in methodological consistency or broad uptake, in part due to methodological complexity and data requirements. For example the HERCULES method [62] requires users to classify the landscape into categories of building, surface cover, and vegetation using LiDAR data. Seress et al. [63] describe another method that also requires some user based classification of satellite imagery into categories of buildings, vegetation, and road to train a semi-automated model. Metrics proposed as generalizable for use in human-natural systems also tend to focus on one axis of landscape modification, typically urbanization, rather than the full spectrum of changes to the landscape [64, 65]. Recently, a human modification gradient [66] has been produced that incorporates all aspects of the human footprint, however, it results in a single metric making it difficult to distinguish differential effects of agriculture or urbanization, as we have demonstrated here.

Multiple metrics have been used to analyze and quantify spatial change in human-dominated systems. Large suites of input variables ranging from landscape configuration measures to human population density have been used to identify multiple important features of change in urban landscapes in several notable instances [9, 65, 67, 68]. In most of these cases, however, multiple univariate measures are identified (e.g., using multi-variate analyses) as representative of landscape change along urban-rural gradients. Meanwhile, Berland and Mason [67], noted that dominant factors or principal components could perhaps be used to directly represent urbanization rather than selecting the variables with highest loading. Ultimately, regardless of the number of metrics utilized, or how they were derived, prior multi-metric research has tended to focus on identifying one aspect of landscape change, namely urbanization. Furthermore, despite identifying multiple important measures, none of these explicitly promote a dual-axis or multi-dimensional application of these measures.

The multi-dimensional landscape gradient approach we propose here offers the flexibility to balance regional adaptability with local specificity and ecological realism to better understand more mechanistically the relationships between landscape structure and ecological processes [58, 69]. We use an established multi-variate statistical approach to succinctly describe spatial heterogeneity and employ readily available NLCD data to incorporate complexity of the entire landscape into a clear and consistent dual axis of human-influence. Although the NLCD dataset is limited to the United States, it employs a nearly identical landscape classification system as other products, including the European Space Agency’s GlobCover data [70] or Copernicus Global Land Cover [71], and therefore should be applicable globally wherever such landcover data are available. In addition to PCA, other multi-variate methods have been suggested as alternatives when identifying landscape gradients, such as factor analysis [9], and non-metric multidimensional scaling [72]. More recently, PCA approaches that better account for similarities and differences in multi-group (e.g., multi-city) data have been developed [73], which may be particularly applicable when analyzing landscape gradients in multiple regions simultaneously. Our analysis has demonstrated the importance of considering multiple axes of variation in landscape gradients, and fits within a methodological framework centered on transparent and ecologically informed analysis [3]. It is important to note, however, that city-specific means (i.e., mean effect of landscape on occupancy) in our global analysis may influence interpretation and comparison of effect size between groups. Ultimately, the multi-variate method selected by researchers should be informed by the study’s goals, objectives, and types of data available [50]. As the human population continues to grow, the urban, industrial, and agricultural infrastructure must be restructured to ensure future ecological integrity, and the resulting debate over how to effectively do this has led to discussions of land-sharing, i.e., integrating natural systems into the mix of human land-uses, versus land-sparing, i.e., where natural and human systems are concentrated in large, individualized patches. Due to a traditional one-dimensional perspective of landscape heterogeneity, this discussion has largely taken place for agricultural [74], and urban [75] systems in isolation. In reality, however, urban, agricultural, and natural landscapes are inherently inter-mixed. Viewing the land-sharing versus land-sparing debate through a multi-dimensional lens of landscape heterogeneity views the landscape mosaic as a more realistic integrated agro-urban-natural system. Furthermore, the species that will benefit or suffer most from any specific sharing or sparing management, depends entirely on the landscape context within which they are evaluated [76]. Determining how to design a conservation strategy and manage a heterogeneous regional landscape for this species would require that the entire human-natural mosaic be considered and could be facilitated with a multi-dimensional approach to landscape context.

American robins are widely considered to be urban-adaptive and are thought to benefit from urbanized (e.g., hard) landscapes with human habitation [77, 78]. However, our results consistently predicted higher occupancy in more forested (green) regions over areas predominantly agricultural (brown), while the effect of the hard-to-soft axis on robin occupancy varied by city both in terms of magnitude and direction likely due to regional variation in composition of the soft landscape (Fig 4). Regional variation in the effect of hard-to-soft on robin occupancy demonstrates the need to consider and decouple multiple dimensions of landscape heterogeneity and suggests that ecological response to human-dominated landscapes is highly nuanced and regionally variable. While highly adaptable and able to exploit many habitat types, robins showed a preference for natural areas in proximity to urbanization (i.e., green-and-hard) over those in more agricultural landscapes. Our approach synthesizes prior research on the species where single landscape gradients were considered in isolation. In urban contexts higher presence and survival of robins was reported in residential yards, woodlots and golf courses [79, 80], while studies in agricultural landscapes found that robins were more common in habitat fragments surrounded by urbanization than those surrounded by agriculture [81].

City-specific variation in robin response to landscape heterogeneity reiterates the importance of landscape context on biotic response. The size (i.e., spatial extent) and density (i.e., human population) of human-dominated landscapes significantly impacts the direction and magnitude of biotic response, with larger and more densely populated cities typically resulting in a stronger negative response [82, 83]. Had our analysis centered on larger or smaller urban regions the specifics of robin response may have differed, however, our core findings—the importance of considering multiple landscape gradients and regional variation in response—would likely have remained. Though, additional research into the impacts of size of human-dominated landscapes on the use of multi-dimensional landscape gradients is warranted. We saw that robin occupancy was demonstrably influenced by both axes of human-modification across the continental United States, suggesting that a continued reliance on one-dimensional landscape descriptors may result in ecosystem pattern being misinterpreted as inherent stochasticity (e.g., noise), when in fact it reflects an overlooked component of the landscape. Specifically in our context, an analysis using a conventional hard-soft gradient would have overlooked the value of green (natural) landscapes integrated in hard (urban) regions for robins (Fig 4). Bearing this in mind, management decisions that consider only a single aspect of the human-natural landscape may overlook or misinterpret ecological response and result in ineffective conservation plans [84].

All measures of landscape heterogeneity are imperfect representations of reality and therefore fall short to varying degrees, and it is unlikely that any single metric will be ideally suited to every question of ecological pattern and process [85]. Therefore, extending one-dimensional descriptors to a multi-dimensional perspective can help move toward a more general understanding of landscape mosaics. And yet, oversimplified one-dimensional measures such as percent forest cover, or percent impervious surface continue to dominate the literature [3]. Multi-city analysis of urban ecosystems has experienced rich growth in recent years. This work has highlighted the negative and positive potential impacts of urbanization on biodiversity, while stressing the importance of the regional landscape context in driving the direction and magnitude of biological response [82, 83, 86]. Still, multi-region analysis remains hampered by inconsistencies in study design and methodological limitations [87, 88]. The multi-dimensional, dual-axis understanding of spatial heterogeneity we describe has the potential to improve and standardize existing approaches to producing ecologically relevant landscape metrics leading to improvements in multi-region research and valuable insight into patterns of ecological response within and across human-dominated systems.

Supplemental landscape gradient analysis.

(DOCX)

Click here for additional data file.

American robin occupancy analysis: Supplementary tables and figures.

(DOCX)

Click here for additional data file.

17 Jun 2021

PONE-D-21-15482

Defining dual-axis landscape gradients of human influence for studying ecological processes

PLOS ONE

Dear Dr. Sutherland,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Although both reviewers had generally favorable perspectives of your manuscript, they both also had a list of concerns.  Most of these are minor and will be straightforward to incorporate.  The major concern identified by both reviewers had to do with the rigor of the PCA analysis.  Please carefully consider whether this analysis is the most appropriate for your data and questions.  Reviewers provided a number of references to consult as you revise the manuscript.

Please submit your revised manuscript by Aug 01 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Janice L. Bossart

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

3. We note that Figure 1 in your submission contain map images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

a) You may seek permission from the original copyright holder of Figure 1 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

b) If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

3. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: NOTE: The review is written in markdown style format. I've also attached it as a pdf, which may be a little more intelligible.

# Review for:

*Defining dual-axis landscape gradients of human influence for studying ecological processes*

This paper describes a way to separate multiple axes of human modification on the landscape through the use of spatially weighted averaging of NLCD data via a Gaussian kernel and application of PCA, which is simple and provided intuitive results (a win win in my opinion). The statistical analysis as described appears sound, and using a common and easy to identify bird species helps alleviate any concerns I may have potentially had with using community science data from ebird (though you perhaps want to sell other readers on this in the methods section). I pretty much have one primary concern (top-level thought number 1), and then some relatively minor suggestions / improvements to the manuscript. I'm waiving anonymity (I'm Mason Fidino), so if there are any questions that arise from this review please reach out to me at mfidino@lpzoo.org.

## Top-level thoughts

1. One aspect I've wondered about a lot, as someone who is also interested in how species distributions differ within and among cities, is how applicable PCA is when applied to multiple cities combined. Certainly, when applied this way PCA will identify the average gradient of urban intensification across cities, but then we do not know how well that average gradient applies to each city. This can also cause statistical issues in terms of model interpretation. For example, while the global mean of each gradient identified via PCA will be zero (assuming the data has been scaled), the city specific means when subset from that global gradient are not likely going to equal that global mean. Similarly, the standard deviation of each gradient will not be equivalent to the global standard deviation. Depending on the extent of these differences, I would imagine it could be very hard to compare effect sizes of regression coefficients from a global PCA that has then been subset among multiple cities (which is why we generally want to have mean = 0 and sd = 1 for a given continuous covariate).

There are, of course, other styles of PCA that help address this (e.g., multi-group PCA, or more recently network PCA), but it's something I've never really seen used in ecology (but perhaps should be when we start thinking about quantifying gradients across cities).

```

Eslami, A., Qannari, E. M., Kohler, A., & Bougeard, S. (2014). Multivariate analysis of multiblock and multigroup data. Chemometrics and Intelligent Laboratory Systems, 133, 63-69.

Codesido, S., Hanafi, M., Gagnebin, Y., González-Ruiz, V., Rudaz, S., & Boccard, J. (2020). Network principal component analysis: a versatile tool for the investigation of multigroup and multiblock datasets. Bioinformatics.

```

The authors did apply single-city PCAs (available in supplemental material), but the explanation within the main text (and supplemental table) still have not especially assuaged the concerns I have here. For example, the single-city PCAs center on an individual city, whereas the multi-city PCA centers on the global average. How different the average of each variable is among cities is then could possibly have a big influence on what is being observed. At a minimum, perhaps it would help to city-mean center your variables before applying the global PCA (and then not mean centering when calling `princomp()` or `prcomp()` in `R`)?

2. With this many cities, and given the interest in quantifying differences within and among cities, I wonder if it would help to treat city as a random effect within the model. Currently, all regression coefficients have to be interpreted relative to some baseline city for comparison whereas a multi-level model you can estimate an average response and deviation from that response for each city. Of course, this adds in the complexity of needing to do a Bayesian analysis, but even if the authors do not want to specify a model in JAGS, NIMBLE, or Stan the `ubms` package facilitates unmarked style analyses while adding in the ability to include `lme4` style random effects into the linear predictors for occupancy and detection. Definitely not necessary, but something to consider here.

https://kenkellner.com/ubms/index.html

3. Multi-city urban ecology research is still relatively new, and so there are not many papers out yet about how differences among cities may influence what we observe within a city. However, I'd encourage the authors to think about the results they found in the context of other multi-city studies (some that come to mind are below). Please note that while I try my best to not encourage authors to cite my papers in the review process, in Fidino et al. (2020) we specifically quantified how common urban mammals urbanization response differed across cities due to landscape level differences among cities, which I think is right in line what what you are observing with your own analysis.

```

Aronson, M. F., La Sorte, F. A., Nilon, C. H., Katti, M., Goddard, M. A., Lepczyk, C. A., ... & Winter, M. (2014). A global analysis of the impacts of urbanization on bird and plant diversity reveals key anthropogenic drivers. Proceedings of the royal society B: biological sciences, 281(1780), 20133330.

Beninde, J., Feldmeier, S., Werner, M., Peroverde, D., Schulte, U., Hochkirch, A., & Veith, M. (2016). Cityscape genetics: structural vs. functional connectivity of an urban lizard population. Molecular ecology, 25(20), 4984-5000.

Fidino, M., Gallo, T., Lehrer, E. W., Murray, M. H., Kay, C., Sander, H. A., ... & Magle, S. B. (2020). Landscape-scale differences among cities alter common species’ responses to urbanization. Ecological Applications, n/a (n/a), e2253. doi: https://doi. org/10.1002/eap, 2253.

Gagné, S. A., Sherman, P. J., Singh, K. K., & Meentemeyer, R. K. (2016). The effect of human population size on the breeding bird diversity of urban regions. Biodiversity and conservation, 25(4), 653-671.

```

4. How universal these trends observed here are conditional on the cities used. As the paper only focuses on cities that range in populations from 200K to 500K I am still left wondering how applicable this is to smaller or larger cities. Certainly worth a few sentences to explore the idea in the discussion.

## Introduction

---

### Top-level thoughts

1. This manuscript gets right to the point, which is refreshing.

2. There is a little bit of passive voice that creeps in, especially with the use of 'understanding'. A small edit, but it would be nice to see these in the active voice and edit such sentences to remove some unnecessary gerunds.

3. As a near permanent change on the landscape, I often wonder why we equate urban intensity as a "disturbance." (Line 71). Would it be more clear to say that species richness is greatest at intermediate levels of urban intensification (or perhaps human modification to keep with the phrase you are using), when landscape heterogeneity is often at it's greatest?

4. The introduction uses the phrase "ecological processes" enough, which is perhaps a little too vague at times. For example, the manuscript states that fragmentation and human population density are not related to a range of ecological processes (line 75), but it would help to be more specific about what these ecological processes are.

### Line by line comments

Line 41: I understand the point of this sentence, but the wording makes it difficult. I think it is because it is unclear what 'expands' is acting upon, the link between 'human population' and 'ecological footprint', and the assumption that this means natural landscapes are being transformed due to this.

Line 72: replace "are more variable," with "vary"

Line 75: missing a word here.

## Methods:

---

### Top-level thoughts

1. A sentence in the PCA section of the methods wherein you give a heads up that you'll describe these gradients within the results section would be helpful. I got a little confused at what is only presented in the methods, and it had me wondering whether or not the manuscript was actually going to define the primary gradients captured in the analysis, provide the loadings, etc (which was in the results).

2. A little more explanation about what a "location" is could be helpful. Were surveys thinned so you could assume they were spatially independent?

### Line by line comments

Line 119: What do these percentages represent? Composition across all landscapes? If so, that does not indicate that landscape composition is similar. We would instead be interested in how much these dominant landscape types vary among cities. For example, based off of table 1, forest landcover is clearly not similar across cities.

Line 173: At a minimum cite either the MacKenzie et al. book, a paper, etc. for occupancy modeling.

## Results

---

### Top-level thoughts

1. Loadings are just as important as the description of what the gradient represents. I'm assuming that is what is listed in Table 3 (PC1, PC2, and PC3), but a little more explanation in the table header can make that explicit.

2. What are the values presented in parentheses? Regression coefficients and p values? Regression coefficients and standard errors? Please provide a little more information.

3. Figure 4 does not especially capture the variation that is being explained. Something similar to Fig 3 for each city could help with this (perhaps replacing figure 3 if there are space constraints, which is focused on detection with one that is focused on within vs among city variability in occupancy).

## Discussion

---

### Top-level thoughts

Since there was no comparison here to one-dimensional approaches for urban gradients I am not certain how the analysis described here represents a challenge. No doubt there is still merit in the PCA approach used here, but perhaps it would be better to sell the reader on the insights which you may have missed otherwise.

Thinking about it even further, and perhaps I'm getting a little off-topic, but urban ecologists have tried to capture urban intensity in a variety of ways. Traditionally, we used categorical metrics (e.g., urban vs rural) which are not useful for a variety or reasons. Following this we tried more continuous metrics that were simple to calculate like distance to city center, which while better than categories it effectively assumes a city is an onion, steadily decreasing in urbanization from it's core through each "onion layer". Modern cities, however, are not onions, so again not a great approach. Currently, I would argue we are in the "one-dimensional" vs "PCA" style approach like this manuscript details. While I fit into the PCA camp a bit more, I'm still a little agnostic to which approach is better, because both have their uses. For example, I would argue that a one-dimensional approach makes it easier to fit multi-level models because that single gradient can provide you a way to incorporate within vs among city variation as two separate continuous covariates (e.g., average of a given gradient within each city, and then a city-mean centered site-specific gradient, Fidino et al. 2020).

## Figures.

Figure 3 legend says it's about detection but the y-axis says occupancy.

Reviewer #2: I appreciate the efforts in suggesting a standardized approach for defining multi-dimensional gradients, as this is key for understanding species responses to human modifications (as shown for the American Robin) and for informing management and conservation efforts. However, previous attempts in this direction have been made and have already suggested a multi-dimensional approach instead of using one-dimensional measure (e.g. Luck and Wu 2002; Hahs and McDonnell 2006; Schwarz 2010; Modica et al. 2012; Suarez-Rubio and Krenn 2018), thus this should be acknowledged and discuss in the manuscript.

Although most statistical analyses were appropriate, a Factor Analysis (FA) should be performed instead of the PCA. du Toit and Cilliers (2011) carefully argue why PCA is not the correct method when the purpose is identifying uncorrelated axes of variation and distinguishing the different aspects forming a multi-dimensional urbanization gradient ‒despite its earlier use. Additionally, subsequent studies used this methodology in their research e.g. Berland and Manson (2013), Leveau (2013), Suarez-Rubio and Krenn (2018). The manuscript's main conclusions are based on the PCA results, which might change considerably if the data were reanalyzed using FA.

A better description of the methods, in particular the landscape quantification is needed. Although it was mentioned that the landscape analysis followed the framework of a previous publication, it is important for the repeatability of the study to briefly describe how the quantification was performed. In Table 2, a description of decisions is included but it is unclear how the physical land-cover differ from the NLCD and from where the land-use data are coming from. In addition, it seems that the study only included landscape composition variables, why configuration variables are not included, this was shown to be relevant in previous studies (see references in paragraph above).

Minor comments:

Line 28 It would be nice if the name of the focal species is mentioned in the abstract.

Line 102 As “medium size city” have different meanings in other urban settings around the world, it will be relevant to include the average size of the cities studied and in Table 1 include the size of each city.

Line 220 There are several tables included in S1, to which one in particular this refers to?

Line 287 Table 3 or 4?

Line 304-306 Where are these results depicted?

Lines 320-326 I assume that these results are presented in S2 Table 3, if so, please make sure that the numbers are the same as some of them are not (e.g. -1.68 vs. -1.72).

Line 342 I do not think this is supported by Fig. 3.

Line 369 Although the GlobCover data is a good source, it is quite old (2009), so better also to include the Copernicus Global Land Cover.

Lines 351-361 Please refer to the comment above.

Line 383 to which species this refers to, the American Robin?

Table 1: There is no need to include % next to each number under the land-cover classes.

Fig.1 For the international readers not familiar with the location of the cities in the US, including the name of the city in the map or even a number as an ID of the city will be highly appreciated. In addition, it will be better to be consistent in the way you refer to them, sometimes the name of the city is used (Line 78) and sometimes the state where the city is located (Line 318, S2 Table 3).

Fig. 2 In the lower left corner, I think that % of forest should be high and not low.

Fig. 3 Based on the text (Lines 302-304), the y-axis in the figures should be detection probability, correct?

S1 will benefit from including first the information about the spatial extent (city window) and the then the smoothing scale and city specific analysis.

The manuscript is well-written. But there are few things that should be revised:

Line 233 Replace boldened by bold.

Line 236 It should be 11.1%

Line 132 and 400 Instead of including the citation in author-year format include the corresponding number.

References should be checked for accuracy and consistency, e.g. 6, 12, 22, 35, 37, 69.

References:

Berland A, Manson SM (2013) Patterns in Residential Urban Forest Structure Along a Synthetic Urbanization Gradient. Annals of the Association of American Geographers 103:749-763

du Toit MJ, Cilliers SS (2011) Aspects influencing the selection of representative urbanization measures to quantify urban-rural gradients. Landscape Ecol 26:169-181

Hahs AK, McDonnell MJ (2006) Selecting independent measures to quantify Melbourne's urban–rural gradient. Landsc Urban Plan 78:435-448

Leveau LM (2013) Bird traits in urban-rural gradients: how many functional groups are there? J Ornith 154:655-662

Luck M, Wu J (2002) A gradient analysis of urban landscape pattern: a case study from the Phoenix metropolitan region, Arizona, USA. Landscape Ecol 17:327-339

Modica G, Vizzari M, Pollino M, Fichera CR, Zoccali P, Di Fazio S (2012) Spatio-temporal analysis of the urban-rural gradient structure: an application in a Mediterranean mountainous landscape (Serra San Bruno, Italy). Earth Syst Dynam 3:263-279

Schwarz N (2010) Urban form revisited—Selecting indicators for characterising European cities. Landsc Urban Plan 96:29-47

Suarez-Rubio M, Krenn R (2018) Quantitative analysis of urbanization gradients: a comparative case study of two European cities. Journal of Urban Ecology 4:1-14

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Mason Fidino

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: PLOS_ONE_REVIEW.pdf

Click here for additional data file.

16 Aug 2021

see attached response letter

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

16 Sep 2021

PONE-D-21-15482R1Defining dual-axis landscape gradients of human influence for studying ecological processes

PLOS ONE

Dear Dr. Sutherland,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The comments from both reviewers were generally minor so I don’t see any need to send the manuscript out for external review.  That said, both reviewers identified multiple areas where additional detail or modification of existing text will improve readability and/or understanding.  Please explain how you have addressed their concerns in your response.

Consult the PLOS ONE website and follow the format for cited references.  Carefully check that all references cited within the text are listed in the References Cited section and that all listed references are cited in the text.  Please also pay very careful attention that references are free of any errors.  PLOS ONE articles are not copy edited.

I have listed additional needed changes below:

Tables 1 & 4.  Right justify all numbers.

Table 1.  Remove % signs within body of table.  Add a statement to the caption that numbers are percent cover, or some such.

Table 3.  Why isn’t font size of numbers consistent throughout?  Why are some numbers italicized?

Do a ‘Find’ for all uses of ‘which’.  Ensure that the word preceding which is followed by a comma, e.g. Line 88, a comma should follow ‘approach’. I noticed at least 3 locations were the comma is missing.

Consolidate all information on your focal species and why this particular species into one spot.  Currently, the information is spread out in three spots and is redundant (lines 95-96, 161-162, 188-191).  Also, if you decide you want to refer to the scientific name more than once, then the standard practice is to spell the name of the genus out fully on first use and thereafter abbreviate to just the first letter, i.e. T. migratorius.

Line 82. Replace ‘despite the fact that’ with ‘even though’.

Line 157. ‘usinga’ should be ‘using a’.

Line 168-180. Use the same font & size for PCA throughout. My preference is for it to be the same as the general text.

Line 168. Should be ‘PCA is’.

Line 170. Delete the extra space before ‘and’.

Line 171. Add a space after the period.

Line 298. Missing ‘to’.

Line 300. Add a comma after ‘date’.

Line 358. Delete second comma.

Line 366. Should be ‘the human footprint’.

Line 374. Should be ‘the landscape’.

Lines 381-385. Might be better to split into two sentences for clarity.

Line 401. Should be ‘approaches that’.

Line 407. Add a space after the period.

Lines 415-416. ‘Viewing’ and ‘views’?  Maybe reword?

Line 429. ‘suggest’ should be ‘suggests’.

Lines 432-436. Clarity lacking.  Needs modified.  Also, likely better to split into two sentences.

Line 433. ‘species have investigated’??  The species investigated?  I don’t think so.

Line 437. Missing ‘to’.

Line 451. Should be ‘robins’.

Please submit your revised manuscript by Oct 31 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Dr. Janice L. Bossart

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: No

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Hello,

The revisions the authors made are fantastic. I think the only thing that really needs to be added in the discussion is a bit on how interpretation of city-specific regression coefficients from the global PCA approach. Basically, acknowledge that city-specific means of the PCA may vary from the global mean, and how that can influence interpretation of effect size. This isn't a problem with this approach, but I would hate for someone to get excited by this technique, use it, and then compare effect size among cities via the magnitude of regression coefficients. By giving a reader a roadmap to model interpretation with your approach, you'll hopefully ensure that it gets used effectively and correctly.

Small line by line comments:

Line 179: Provide some justification to the reader about why you also conducted this analysis separately (e.g,. determine how well the combined (i.e., all cities) gradient described city-specific gradients).

Table 1: Shouldn't the numeric data be right justified?

Line 296: You report occupancy probabilities later as values between 0 and 1 instead of a percentage I'd be consistent in the way you report it. My personal preference is [0-1], but that is what it is, a preference.

Line 403: I know what you mean here, but readers unfamiliar with the term 'multi-block' coul be confused. Also, in this case, I think you actually mean 'multi-group'. I believe that multi-block is grouping among covariates (column wise) while multi-group means there is categorical structure of the same covariates (row wise). Maybe modify to "...differences in multi-group (e.g., multi-city) data..."

Figure 3: Remove the underscores in Salt_Lake_City, and perhaps add the state abbreviations (e.g., because there is a Portland, Maine as well).

Cheers,

Mason Fidino

Reviewer #2: Thank you for addressing my previous comments. The clarity of the manuscript has improved and key points were acknowledged and added. However, there are minor things, mostly related to style, that should be considered. This is particularly important as PLOS ONE does not copyedit accepted manuscripts.

L82 Replace ‘despite the fact that’ by ‘although’

L113, 289 Please make sure that the heading included here match the ones included in the Appendix.

Table 1 This was previously mentioned, but the changes were not made. The percentages in each of the columns are not needed, please remove them.

L137-143/Table 2 The description of the methodology included in this paragraph is not clear. I had trouble following the procedure and although you cited Table 2, that did not help. What is the difference between ‘land-cover categories’ mentioned in ‘landscape features’ and the ‘land-cover data’ mentioned for the spatial data? You cited a previous publication, but to allow repeatability of the study it is important to clearly describe these steps.

L286 It should be ‘x’

L354 Remove ‘configuration’. The study does not quantify the configuration of the cities, so including here the word ‘configuration’ is misleading. The information included in Table 1 in S1 Appendix includes only the proportion of land-cover i.e., composition.

L415-417, L437-438 Please revise the grammar of these sentences.

References There are many mistakes in the list of references. I mentioned here only few examples. In some cases each word of the title is capitalized in other instances it is not. In 6, Beggs J. editor, should be removed. In 22 first names should be replaced by initials. #46 is not cited in the text.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Mason Fidino

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

14 Oct 2021

Tables 1 & 4. Right justify all numbers.

Numbers are right justified

Table 1. Remove % signs within body of table. Add a statement to the caption that numbers are percent cover, or some such.

% signs have been removed, and relevant statement added to caption.

Table 3. Why isn’t font size of numbers consistent throughout? Why are some numbers italicized?

Thank you, numbers have been resized and italics removed

Do a ‘Find’ for all uses of ‘which’. Ensure that the word preceding which is followed by a comma, e.g. Line 88, a comma should follow ‘approach’. I noticed at least 3 locations were the comma is missing.

Thank you for noticing this grammatical error. Corrected

Consolidate all information on your focal species and why this particular species into one spot. Currently, the information is spread out in three spots and is redundant (lines 95-96, 161-162, 188-191). Also, if you decide you want to refer to the scientific name more than once, then the standard practice is to spell the name of the genus out fully on first use and thereafter abbreviate to just the first letter, i.e. T. migratorius.

We appreciate this suggestion, however, after reviewing the manuscript, we believe that it is best organized as is. The majority of the important life history information on the study species is located in lines 188-191. The information in lines 95 and 96 is a brief reference to the study species in a summary of the manuscript’s objectives, while the reference in lines 161-162 is important information for the statistical methodology described in that section.

We have changed the second reference to the genus species name to read as T. migratorious.

Line 82. Replace ‘despite the fact that’ with ‘even though’.

Changed

Line 157. ‘usinga’ should be ‘using a’.

Revised

Line 168-180. Use the same font & size for PCA throughout. My preference is for it to be the same as the general text.

All occurrences of PCA have been changed to match the style of the general text.

Line 168. Should be ‘PCA is’.

Revised

Line 170. Delete the extra space before ‘and’.

Revised

Line 171. Add a space after the period.

Revised

Line 298. Missing ‘to’.

‘To’ has been added, thank you

Line 300. Add a comma after ‘date’.

Revised

Line 358. Delete second comma.

Revised

Line 366. Should be ‘the human footprint’.

Revised

Line 374. Should be ‘the landscape’.

Revised

Lines 381-385. Might be better to split into two sentences for clarity.

Revised

Line 401. Should be ‘approaches that’.

Revised

Line 407. Add a space after the period.

Revised

Lines 415-416. ‘Viewing’ and ‘views’? Maybe reword?

This text has been reworded for clarity

Line 429. ‘suggest’ should be ‘suggests’.

Revised

Lines 432-436. Clarity lacking. Needs modified. Also, likely better to split into two sentences.

The text has been split into two sentences to improve clarity and readability

Line 433. ‘species have investigated’?? The species investigated? I don’t think so. – This has been changed to “research on the species where single landscape gradients were considered in isolation”

Revised

Line 437. Missing ‘to’.

Revised

Line 451. Should be ‘robins’.

Revised

The revisions the authors made are fantastic. I think the only thing that really needs to be added in the discussion is a bit on how interpretation of city-specific regression coefficients from the global PCA approach. Basically, acknowledge that city-specific means of the PCA may vary from the global mean, and how that can influence interpretation of effect size. This isn't a problem with this approach, but I would hate for someone to get excited by this technique, use it, and then compare effect size among cities via the magnitude of regression coefficients. By giving a reader a roadmap to model interpretation with your approach, you'll hopefully ensure that it gets used effectively and correctly.

We have added the following text to lines 410-412 “It is important to note, however, that city-specific means in our global analysis which may influence interpretation and comparison of effect size between groups.”

Small line by line comments:

Line 179: Provide some justification to the reader about why you also conducted this analysis separately (e.g,. determine how well the combined (i.e., all cities) gradient described city-specific gradients).

Table 1: Shouldn't the numeric data be right justified? � DONE

Line 296: You report occupancy probabilities later as values between 0 and 1 instead of a percentage I'd be consistent in the way you report it. My personal preference is [0-1], but that is what it is, a preference. � we have changed these to 0-1 scale

Line 403: I know what you mean here, but readers unfamiliar with the term 'multi-block' coul be confused. Also, in this case, I think you actually mean 'multi-group'. I believe that multi-block is grouping among covariates (column wise) while multi-group means there is categorical structure of the same covariates (row wise). Maybe modify to "...differences in multi-group (e.g., multi-city) data..." � Thanks for suggestion, this has been added

Figure 3: Remove the underscores in Salt_Lake_City, and perhaps add the state abbreviations (e.g., because there is a Portland, Maine as well). � DONE

Cheers,

Mason Fidino

Reviewer #2: Thank you for addressing my previous comments. The clarity of the manuscript has improved and key points were acknowledged and added. However, there are minor things, mostly related to style, that should be considered. This is particularly important as PLOS ONE does not copyedit accepted manuscripts.

L82 Replace ‘despite the fact that’ by ‘although’

Revised

L113, 289 Please make sure that the heading included here match the ones included in the Appendix.

Revised

Table 1 This was previously mentioned, but the changes were not made. The percentages in each of the columns are not needed, please remove them.

Revised

L137-143/Table 2 The description of the methodology included in this paragraph is not clear. I had trouble following the procedure and although you cited Table 2, that did not help. What is the difference between ‘land-cover categories’ mentioned in ‘landscape features’ and the ‘land-cover data’ mentioned for the spatial data? You cited a previous publication, but to allow repeatability of the study it is important to clearly describe these steps.

We have done our best to improve this portion of the methods, and have revised it as follows “Our decisions regarding the types of landscape features relevant for analysis, the data to represent those features, and the spatial scales of analysis were made to reflect a typical ecological analysis - definitions of, and justification for, these decisions are provided in Table 2”

L286 It should be ‘x’

Revised

L354 Remove ‘configuration’. The study does not quantify the configuration of the cities, so including here the word ‘configuration’ is misleading. The information included in Table 1 in S1 Appendix includes only the proportion of land-cover i.e., composition.

Revised

L415-417, L437-438 Please revise the grammar of these sentences.

Revised

References There are many mistakes in the list of references. I mentioned here only few examples. In some cases each word of the title is capitalized in other instances it is not. In 6, Beggs J. editor, should be removed. In 22 first names should be replaced by initials. #46 is not cited in the text.

Revised

Submitted filename: Response_to_Reviewers_pone.docx

Click here for additional data file.

22 Oct 2021

PONE-D-21-15482R2Defining dual-axis landscape gradients of human influence for studying ecological processesPLOS ONE

Dear Dr. Sutherland,

Thank you for submitting your revised manuscript to PLOS ONE.  Unfortunately, I'm afraid I need to return it once again given the many minor corrections that are necessary.  As I pointed out earlier, PLOS ONE does not use copy editors.  Please submit a revised version after you've carefully dealt with the needed corrections.

I can appreciate that you prefer to leave the information on your case study species in the three separate locations even though it seems redundant to this reader to be told multiple times that your case study species is the American Robin, that the American Robin is Turdus migratorious, and that the American Robin is a widespread generalist found in areas with humans.  At the very least, the 2nd mention of the scientific name is unnecessary since the American Robin was already identified as such the first time.  Please delete that second mention.

Lines 410-412.  Incomplete sentence:  “It is important to note, however, that city-specific means in our global analysis which may influence interpretation and comparison of effect size between groups.”  Given you're largely, but not entirely, incorporating what the reviewer literally suggested, I'm not exactly sure what changes are needed.  The sentence would easily make sense by simply removing the 'which'.  That said, some readers less familiar with your analyses might appreciate more complete inclusion of the 'city-specific means differing from the global mean' verbiage, as the reviewer included.

Many corrections are needed in the references

Use lower case in the titles: L495, L502, L507, L512, L537, L648, L698, L711

Journal name should not be all caps: L529, L526, L547

Italicize scientific names: L571, L680

L562. Delete 1st Ecol

L677. Delete (80-) after 'Science'

L585. Lower case 'Framework'

Please submit your revised manuscript by Dec 06 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Janice L. Bossart

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

28 Oct 2021

Dear Janice Bossart,

Thank you for the detailed editorial review of our manuscript. We have made all final changes, and we trust you will agree that all necessary modifications and edits. Below you will find the final round of revisions, with our edits in red.

1. I can appreciate that you prefer to leave the information on your case study species in the three separate locations even though it seems redundant to this reader to be told multiple times that your case study species is the American Robin, that the American Robin is Turdus migratorious, and that the American Robin is a widespread generalist found in areas with humans. At the very least, the 2nd mention of the scientific name is unnecessary since the American Robin was already identified as such the first time. Please delete that second mention.

Thank you for standing firm on this. We have removed the second reference to the scientific name in line 191. Further, we have removed our description of the species ecology in lines 95 through 96, and it now reads “We demonstrate the utility of our approach in a case study analysis of American robin (Turdus migratorious), occupancy.”

2. Lines 410-412. Incomplete sentence: “It is important to note, however, that city-specific means in our global analysis which may influence interpretation and comparison of effect size between groups.” Given you're largely, but not entirely, incorporating what the reviewer literally suggested, I'm not exactly sure what changes are needed. The sentence would easily make sense by simply removing the 'which'. That said, some readers less familiar with your analyses might appreciate more complete inclusion of the 'city-specific means differing from the global mean' verbiage, as the reviewer included.

We have removed ‘which’ in order to make the sentence complete. Further, to aid in providing the reader with clarity regarding city-specific means, we have added the following to the text: “city-specific means (i.e., mean effect of landscape on occupancy) in our global analysis”

3. Many corrections are needed in the references

Use lower case in the titles: L495, L502, L507, L512, L537, L648, L698, L711

Journal name should not be all caps: L529, L526, L547

Italicize scientific names: L571, L680

L562. Delete 1st Ecol

L677. Delete (80-) after 'Science'

L585. Lower case 'Framework'

All references have been corrected

Submitted filename: Response_to_Reviewers.docx

Click here for additional data file.

2 Nov 2021

Defining dual-axis landscape gradients of human influence for studying ecological processes

PONE-D-21-15482R3

Dear Dr. Sutherland,

Congratulations!  We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing any required amendments.  At that time please also correct the one (very) minor remaining issue:  Line 97 - remove the unnecessary comma after "migratorious)," before submitting the final version that will go to press.

When these required issues have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Dr. Janice L. Bossart

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

9 Nov 2021

PONE-D-21-15482R3

Defining dual-axis landscape gradients of human influence for studying ecological processes

Dear Dr. Sutherland:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Janice L. Bossart

Academic Editor

PLOS ONE