Introduction

The northern spotted owl (Strix occidentalis caurina; hereinafter NSO) was listed as “threatened” under the Endangered Species Act in 1990 (U.S. Fish and Wildlife Service, 1990). Because of continued population declines, the U.S. Fish and Wildlife Service (hereinafter USFWS) recently determined that uplisting NSO to “endangered” status was “warranted but precluded” by other listing priorities (U.S. Fish and Wildlife Service, 2020). Given the species’ protected status, any proposed activities on Federal lands that might impact NSO require consultation with USFWS, and part of that consultation often includes surveys to determine presence and occupancy status of this threatened species in the area of the proposed activity (U.S. Fish and Wildlife Service, 1992, 2012).

Detection of NSOs in forest stands incorporates two processes: (1) the probability that the territory is occupied (in other words, one or more NSO are present on the territory) and (2) the probability that if an NSO is present on the territory, it is detected. Thus, survey protocols and analyses that allow these two probabilities to be estimated separately provide the best estimates of territory occupancy and are important for documenting when planned management activities will negatively impact resident NSOs through habitat modifications or disruption to essential breeding activities. The primary method of surveying for NSOs in long-term demography studies includes as much as six call-back surveys conducted at fixed points along a route at night, or during walk-ins to previously occupied territory centers during the day (Franklin and others, 1996; Yackulic and others, 2012). As has been documented for some time, the invasive barred owl (Strix varia; hereinafter BO) suppresses NSO response rates to traditional call-back surveys (Olson and others, 2005; Crozier and others, 2006; Duchac and others, 2020). Barred owl presence on a territory also negatively impacts occupancy dynamics of NSO in the landscape (for example, Olson and others, 2005; Kroll and others, 2010; Dugger and others, 2011, 2016; Franklin and others, 2021). Therefore, NSO surveys that account for the probability of detection and BO presence provide the most robust estimates of NSO site occupancy.

Previously published estimates of NSO detection rates were based on per-visit detections (Olson and others, 2005; Kroll and others, 2010; Dugger and others, 2011), rather than equal-length survey periods that might include multiple territory visits (Yackulic and others, 2014; Dugger and others, 2016; Franklin and others, 2021). Early estimates also included detections of any owl—single or pair—and these per-visit detection rates were generally higher than detections of pairs only (Olson and others, 2005; Kroll and others, 2010). However, an owl pair is the ecological unit required to maintain viable populations so pair occupancy is vitally important to understanding population status and trends (Dugger and others, 2011). Recent efforts to understand occupancy dynamics of NSO have focused on detections of NSO pairs only (Yackulic and others, 2012, 2014, 2019; Dugger and others, 2011, 2016; Franklin and others, 2021). In these analyses, the estimation and modeling of NSO pair occupancy dynamics (for example, colonization, local extinction, annual occupancy) were the primary objective, so although the probability of detection was also always modeled, it was often not reported (for example, Dugger and others, 2016; Franklin and others, 2021).

The objective of this report is to present study-area specific estimates of the probability of detection for NSO pairs from twelve 2-week seasonal survey periods (“survey-specific”) using data from a recent meta-analysis of 11 NSO demographic study areas across the species range (Franklin and others, 2021). The NSO meta-analysis was not focused on estimating the probability of detecting NSO pairs, but such estimates were generated as a by-product of the pair occupancy modeling and might provide insight into potential changes into the continued effects of BOs on NSO detection rates. Since these estimates were not provided in Franklin and others (2021), they are presented in this report.

Study Areas

Populations of NSO that occurred on Federal and non-Federal lands within the range of the species were the basis of this analysis. To address our objectives, we used detection/non-detection survey data from historically monitored territories on 11 NSO demography studies included in the most recent meta-analysis evaluating population status and trends (table 1; fig. 1; Franklin and others, 2021). This analysis included three study areas in Washington (Cle Elum [CLE], Rainier [RAI], Olympic Peninsula [OLY]), five study areas in Oregon (Coast Ranges [COA], HJ Andrews [HJA], Tyee [TYE], Klamath [KLA], and South Cascades [CAS]), and three study areas in Northern California (Northwest California [NWC], Hoopa [HUP], Green Diamond Resources [GDR]; fig. 1). These study areas primarily represent lands under Federal administration, but some private, Tribal, and mixed ownership lands are also included (table 1). Detailed descriptions of each study area are available in Anthony and others (2006).

Table 1.    

General characteristics of 11 study areas used to estimate probability of detection of northern spotted owl (Strix occidentalis caurina) pairs in Washington, Oregon and California, 1993–2018.

[WA Douglas-fir refers to Pseudotsuga menziesii. Abbreviations: NW, Northwest; WA, Washington; OR, Oregon; CA, California; OR-CA, Oregon–California; km², square kilometers]

Table 1.    General characteristics of 11 study areas used to estimate probability of detection of northern spotted owl (Strix occidentalis caurina) pairs in Washington, Oregon and California, 1993–2018.

Study area	Acronym	Area (km2)	Landowner	Ecological region
Washington
Cle Elum	CLE	1,784	Mixed	WA mixed conifer
Rainier	RAI	2,167	Mixed	WA Douglas-fir
Olympic	OLY	2,230	Federal	WA Douglas-fir
Oregon
Coast Ranges	COA	3,922	Mixed	OR coastal Douglas-fir
HJ Andrews	HJA	1,604	Federal	OR Cascades Douglas-fir
Tyee	TYE	1,026	Mixed	OR coastal Douglas-fir
Klamath	KLA	1,422	Mixed	OR-CA mixed conifer
South Cascades	CAS	12,372	Federal	OR Cascades Douglas-fir
California
NW California	NWC	460	Federal	OR-CA mixed conifer
Hoopa	HUP	356	Tribal	OR-CA mixed conifer
Green Diamond Resources	GDR	1,340	Private	CA coast
Total		18,683

¹

Study area size was 3,377 km² in 2014 meta-analysis.

Methods

Results

Detection probabilities for within-season surveys for NSO pairs generally indicated a quadratic pattern, increasing from the earliest surveys in March through mid-season (May–June), decreasing again late in the season for five study areas (CLE, OLY, HJA, KLA, CAS; table 3; fig. 2). The pattern for COA and TYE was quadratic but the highest detection rates occurred earlier in the season, in late March or early April, and then decreased through the rest of the season (table 3; fig. 2). For the three California study areas (NWC, HUP, GDR) and one study area in Washington (RAI), the highest per-survey pair detection rates occurred during the earliest surveys in March and April and then decreased through the season to lows in August (fig. 2B, I–K). The additive effect of BO was negative for all study areas, with substantial reductions in per-survey NSO pair detection rates on all areas when BOs were present (table 3; fig. 2).

Estimates of cumulative seasonal detection of NSO pairs (${\widehat{p}}^{*}$) from a maximum of six surveys were highest when based on either the single highest estimate or the six highest estimates. However, only CLE achieved $p^{*}$ greater than (>) 0.90 and represented the highest estimates of per-survey probability of detection when BOs were present (table 4). For most study areas the estimates of $p^{*}$ based on the six highest $p_{.j}$ estimates varied from 0.60 to 0.75, but a low of 0.51 was observed on OLY and a high of 0.91 on CLE when BOs were present (table 4). In contrast, estimates of $p^{*}$ based on six surveys were >0.90 when BOs were not present on a territory for all $p_{.j}$ configurations except for the lowest estimates (table 5), or when surveys 2,4,6,8,10, and 12 were used on OLY and HUP (table 5).

Despite low detection rates when BOs were present on a territory, conditional estimates of occupancy for territories where NSO pairs were never detected generally were also very low, particularly for the lowest per-survey pair detection rates and unconditional probability of occupancy for NSO pairs from 2018 when BOs were present (table 6). The probability that a territory was occupied when an NSO was not detected over six surveys was highest for study areas with the highest NSO pair occupancy rates and the lowest per-survey pair detection rates within seasons for both territories with and without BOs (table 6).

Discussion

For various ecological and analytical reasons, recent occupancy analyses for NSOs have focused on detections of NSO pairs (Dugger and others, 2011, 2016; Yackulic and others, 2012; 2014; Franklin and others, 2021) rather than detection of any owl, single or pair. From the perspective of understanding population status and trends, territorial occupancy of NSO pairs is the relevant ecological metric. We observed variation in the survey-specific probability of detecting a pair over the course of the breeding season with five study areas having the highest per-survey detection rates in May, consistent with earlier findings by Yackulic and others (2014) on TYE. However, we observed the highest per-survey detection rates earlier in the season (March or April) for all California study areas, and on RAI in Washington, and COA and TYE in Oregon, suggesting that the timing of the most efficient survey effort might be different depending on the study area.

Our estimates of the probability of detection of NSO pairs based on 2-week time periods during the season (per-survey estimates) were 29–59 percent lower than previous estimates for any owl (single or pair) on the same study areas (OLY, CLE, COA, HJA, TYE, CAS; Dugger and others, 2009; table 7). However, our findings were similar to mean per-visit detection rates of owl pairs in Oregon and California across territories with and without BOs (Olson and others, 2005, Kroll and others, 2010). Estimates of the probability of detection of NSO pairs from these earlier studies varied by season but were often less than (<) 0.40 (0.22–0.67; fig. 3 in Olson and others, 2005; 0.27–0.67; fig. 3 in Kroll and others, 2010). On the CAS study area, previous estimates of per-visit detection rates for NSO pairs when BOs were present were often <0.20, as compared to rates >0.50 when BOs were absent (Dugger and others, 2011), consistent with current findings for CAS at least when BOs were present in territories (table 3). Thus, detection rates of pairs in response to BO presence may not have changed substantially in the last 10 years, at least in some areas.

Estimates of detection rates for NSO pairs across 2-week survey periods are not necessarily equivalent to per-visit detection rates reported in previous studies as per-visit detections reflected a single complete territory survey (for example, Olson and others, 2005; Dugger and others, 2011), whereas the 2-week time periods used in our analysis could include the results of less than one complete territory survey. Additionally, detection rates for pairs can be substantially lower than estimates for any owl (both singles and pairs; Olson and others, 2005; Kroll and others, 2010). As a result, cumulative seasonal detection rates for NSO pairs ($p^{*}$) from this study when BOs were present were less than 0.95 on all study areas except CLE under a six-survey protocol (table 8). Increasing the number of surveys within a season could increase the probability of detecting an NSO pair at least once during the season, but the number of surveys required to meet a target of $p^{*}$>0.95 or even >0.90 would range from 8 to >12 depending on the study area (table 8). Additionally, the extra calling by field crews could also create more negative consequences for breeding NSO because of the increased territory that defense callbacks elicit, or the potential for increased interactions with resident BO that might occur when NSO respond to callbacks. Alternatively, conducting most surveys earlier in the season, when the probability of detecting pairs is highest (through May on most areas) could improve seasonal detection rates. When BO are not present on a territory, a six-survey protocol had a high probability of detecting an NSO pair at least once during the season on all study areas, except for the very lowest $p{.}_j$ estimates (table 8).

In contrast to detection rates, occupancy dynamics have changed substantially since the mid-to late-2000s. Colonization rates have declined and local extinction rates have increased for NSO populations in all study areas coincident with the increasing presence of BO in the landscape (Franklin and others, 2021). These changes in occupancy dynamics have resulted in strong declines in territory occupancy (fig. 9 in Franklin and others, 2021) and increased rates of movement between territories (in other words, breeding dispersal) by NSOs (Jenkins and others, 2021). This result is consistent with overall declines in rates of population change, suggesting that NSO populations continue to decline range-wide (Franklin and others, 2021). Smaller NSO populations and declines in the probability of pair territory occupancy can impact the cumulative seasonal probability of detecting NSO pairs during surveys. Although a six-survey protocol has a high probability of detecting an NSO pair at least once during the season if BO are not present on the territory, by 2018, BO occupancy on NSO territories ranged from 0.49 to 0.97 (fig. 9 in Franklin and others, 2021). Thus, very few NSO territories included in this analysis were not impacted by BO. Consequently, NSO pair occupancy rates observed in 2018 by Franklin and others (2021) were low in many study areas (<0.25 on seven areas). For these study areas, the probability of never detecting NSO pairs within a season, given that they were present (in other words, closed conditional occupancy) when six surveys were conducted, was generally <0.10 for the highest per-survey detection rates observed when BOs were present. Thus, when occupancy rates are very low, the probability of missing an NSO pair if it was present using a six-survey protocol was low, even though per-survey detection rates were also very low. When BOs were not present, $p_{.j}$ estimates on all study areas were generally high enough that there was a low probability of incorrectly assigning pair occupancy status to a territory under a six-survey protocol (in other words, closed conditional occupancy was <0.10 on most study areas; table 6).

Alternative methods of population monitoring—such as the use of passive acoustic monitoring where as little as 3 weeks of recording can document the presence/absence of NSO (singles or pairs) with 95-percent certainty (Wood and others, 2019; Duchac and others, 2020)—may be needed to continue monitoring NSO for research and management (Lesmeister and others, 2021). However, these estimates of the probability of detection are not directly comparable to our estimates of the probability of detecting NSO pairs. These studies have used random grid sampling that estimated the probability of use rather than occupancy of historical territories and estimated detection of any NSO (singles and pairs; Wood and others, 2019; Duchac and others, 2020). Advances in processing and interpretation of acoustic data are ongoing and distinguishing between the sexes (Dale and others, 2022), territory residence, and breeding status (in other words, transients, resident singles, and pairs) may be possible using calling frequency and timing. However, the delineation of NSO activity centers would likely require additional survey effort (walk-ins in addition to the use of passive acoustic monitoring) to provide the information currently used to protect NSO in forest stands proposed for harvest (Duchac and others, 2020). Alternatively, the trigger for stand protection could shift from identifying an occupied NSO activity center to identifying any stand with some specified level of “use” by NSO (Lesmeister and others, 2021).

Changes in NSO pair occupancy rates across the species range (for example, Franklin and others, 2021), and BO effects on NSO behavior including decreased detection rates, and longer, and more likely movement of breeding birds (Jenkins and others, 2019, 2021) has increased the difficulties of monitoring NSO. However, it is critical to NSO conservation and management to understand where on the landscape productive NSO pairs exist, particularly when proposed management activities may negatively impact those pairs. BO control strategies can be successful at halting or reversing declines in NSO populations (Diller and others, 2016; Wiens and others, 2021); however, these activities can only be successful if suitable habitat is available for NSO to recolonize after BO are controlled.