Trophic mismatches are commonly reported across a wide array of taxa and can have important implications for species participating in the interaction. However, to date, examples of trophic mismatch have centrally focused on those induced by shifts in climate. Here we report on the potential for humans to induce trophic mismatch by shifting the phenology of fire. Globally, anthropogenic fire ignitions are phenologically mismatched to that of historic lightning ignitions but the effects of this phenological mismatch on trophic interactions are poorly understood. Using fire records from 1980-2016 from the southeastern USA, a hotspot of anthropogenic fire, we demonstrate that there is a temporal mismatch between anthropogenic and lightning lit fires in this region. The peak of anthropogenic ignitions (i.e., 45% during March and April) occurred 3 months earlier than the peak in lightning-ignited fires (i.e., 44% occurred during June and July), a pattern consistent with reports from several other regions and continents. We demonstrate with a field experiment conducted at a nutrient-poor site in the southeastern U.S., that anthropogenic fire phenology shifts nutrient pulses in resprouting plants so that they mismatch herbivore reproductive demands. Consequently, plant nutrient quality in four commonly consumed forages was below the threshold to meet lactation requirements. Neonates subsequently were more likely to starve when born far from areas burned during the peak month of lightning fire phenology. Our data indicate that human activities may be an additional causative agent of trophic mismatch.

Regional fire data

We downloaded wildland fire occurrence data for all fires that occurred in 11 states in the southeastern United States for the time period of 1980–2015 from the federal fire occurrence database (https://wildfire.cr.usgs.gov/firehistory/data.html). The states chosen included Alabama, Arkansas, Georgia, Florida, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Virginia and West Virginia. We extracted all fires known to be caused by lightning and all fires that were known to be prescribed, along with their respective ignition dates. We used the start date listed for each fire to assign it to month of ignition. We then plotted the average percentage with standard error across years by month of anthropogenic and lightning ignited fires.

Field experiment study area

We sampled forage quality and neonate survival at Fort Bragg Military Installation (Fort Bragg), North Carolina, USA (35.1°N, −79.2°W). The 73 469 ha area was located in the Sandhills physiographic region of the longleaf pine Pinus palustris ecosystem. This ecosystem and the species within evolved over millennia with relatively frequent (3-year average), low-intensity surface fires occurring due to lightning, and native American activities over the past 10 000–20 000 years (Outcalt 2000). For the purposes of restoring ecosystem function and conserving endangered species, the United States Department of Defense has managed forested stands on a 3-yr fire regime (Lashley et al. 2014b). The parturition phenology of the local ungulate (i.e. white-tailed deer Odocoileus virginianus) peaks in early June (Chitwood et al. 2015a). Thus, because the peak nutritional demand occurs during lactation 3–6 weeks after parturition in white-tailed deer (Hewitt 2011), the peak nutritional demand based on reproductive phenology of this herbivore in this area is during July.In this study area, soil productivity is particularly poor (Lashley et al. 2015b). Thus, deer may be sensitive to shifts in resource pulse phenology. As evidence of this nutritional burden, a relatively large portion of neonates starve on the site, regardless of fire timing, as compared to similar studies in more productive soil regions (Chitwood et al. 2014, 2015b). Likewise, diet selection is relatively narrow and concentrated on obtaining exceedingly limited quantities of phosphorus (Lashley et al. 2015b, 2016). Moreover, predation risk is relatively high (Chitwood et al. 2014, 2015b, 2017) and may limit female selection of the highest quality resources during lactation (Lashley et al. 2015c). Importantly, this population does not have access to anthropogenic subsidies as is common in other parts of their range, so changes in starvation should be related to available nutrition in the native plant community.

Field experiment study design

In a randomized block design, we selected four upland longleaf pine forest stands in each of 3 separate watersheds (blocks), averaging ~8 km apart, with similar soil types (Candor Sands complex) and similar basal area (45–60 m² ha⁻¹). We randomly assigned stands to each of four fire phenologies relative to our plant sampling period (see next section): fires ignited in June of the previous year (1 year-since-fire), and fires ignited in the same year in February (i.e. early anthropogenic phenology), April (i.e. late anthropogenic phenology) and June (lightning phenology). The February fire phenology was meant to represent the onset of the anthropogenic fire season (Brennan et al. 1998, Cox and Widener 2008). The April fire phenology was intended to represent the end of the anthropogenic fire season (Platt et al. 1988, Robbins and Myers 1992, Streng et al. 1993, Glitzenstein et al. 1995, Kirkman et al. 1998, Hiers et al. 2000, Knapp et al. 2009). The June fire phenology was meant to represent the peak in lightning fires for this region (Knapp et al. 2009). Each block contained a replicate from each treatment, and 1 year-since-fire was intended to be the control for comparison to fires ignited in the same year because previous reports suggested that any nutritional benefits would be lost after a single growing season (Dills 1970, Wood 1988, Carlson et al. 1993, Van de Vijver et al. 1999, Long et al. 2008, Nichols et al. 2021).

Plant sampling and analysis

We selected 4 native plant species that occurred in every replicate of each fire treatment. Because deer eat plants of several growth forms, we selected 2 trees, 1 shrub and 1 forb to ensure the plants represented responses across functional groups. The trees collected were common persimmon Diospyros virginiana and sassafras Sassafras albidum, the shrub was dwarf huckleberry Gaylussacia umosa and the forb was fragrant goldenrod Solidago odora. We selected these species because they are common in the study area and commonly consumed by white-tailed deer (Lashley et al. 2015b, 2016).In each month of the growing season (i.e. May–September), we remotely established a plot center in each replicate of each treatment using a geographic information system. We navigated to the a priori selected plot center and collected the foliage of the nearest 10 plants of each species that were in the understory strata (i.e. < 1.5 m tall), separately bagging young leaves and the mature plant parts not typically eaten by this herbivore (Lashley et al. 2014a). We used previous data collected on site to determine that 10 plant samples was robust to the expected variation in intraspecific plant nutritional value (Lashley et al. 2015b). We separated physiologically mature and immature plant parts because plant maturity affects quality, and we were interested in how fire affects relative maturity of plant tissue, quality of young leaves, as well as quality of the whole plant. We assumed that secondary plant compounds were not significantly affecting nutritive quality based on results presented in Jones et al. (2010) that demonstrated tannin defensive compounds in forages consumed by herbivores in this region were generally low. If a plant was discolored, malformed or damaged (by herbivory or otherwise), we did not collect the plant tissues and instead sampled the next nearest plant. To avoid biases associated with forage handling, we followed the protocol presented by Lashley et al. (2014a) by transporting samples within 3 hours to a convection oven and drying forages to constant mass at 47°C. After drying samples, we measured weight to the nearest 0.01 g of the young and mature plant parts and shipped samples to the Clemson University Agricultural Service Laboratory, which was certified by the United States National Forage Testing Association.The lab performed a standard full nutrient array with chemical determination methods to yield the percent of each sample of young and mature plant parts that was crude protein (i.e. nitrogen × 6.25; CP), acid detergent fiber (ADF) and neutral detergent fiber (NDF). For the same samples, we obtained measurements for macro-nutrients phosphorus (P), potassium (K) and calcium (Ca), and micro-nutrients magnesium (Mg), zinc (Zn), copper (Cu), manganese (Mn), iron (Fe), sulfur (S) and sodium (Na). After obtaining the nutritional parameters for physiologically young and mature plant parts, we calculated the whole plant nutritional value by weighting each sample by the relative proportion of physiologically young and mature plant parts and their associated nutritional values.For the purposes of understanding the effects of fire phenology mismatch on available nutrition for white-tailed deer, we calculated the phosphorus requirements of a lactating white-tailed deer conservatively based on the minimum concentration needed to obtain adequate phosphorus assuming forage abundance was not limiting maximum possible physiological intake. We used this nutrient specifically because it was formerly deemed the limiting nutrient on this study site (Lashley et al. 2015b). However, we recorded the array of nutrients because it was part of a standard analysis at the lab. We assumed a maximum daily intake for a 45 kg animal (i.e. average adult female on site (Lashley et al. 2015b)) was 4.8% of the body weight or 2.16 kg day⁻¹ (dry matter), which is the reported physiologically limited possible dry matter intake for female white-tailed deer during peak lactation (National Research Council 2007). Our intention with this calculation was simply to compare the forage quality in terms of phosphorus availability in the plants following each respective fire phenology to determine if those plants would meet the phosphorus requirement for an average size female with one fawn in the study area. We estimated the phosphorus concentration of the plants would need to be a minimum of 0.025% for a lactating female which is consistent with previous estimates (McEwen et al. 1957, Barnes et al. 1990). In JMP Pro 11.0 (SAS Corporation, Cary North Carolina, USA), we fit general linear mixed models with restricted maximum likelihood to evaluate the effects of fire treatments on the proportion of biomass contributed by young leaves, the nutritional quality of young leaves and the nutritional quality of the whole plant. We included random effects of drainage (i.e. block) and plant species to control for influences on nutritional quality not related to fire.

Influence of fire phenology on deer reproductive success

To determine the influence of fire phenology on deer reproductive success, we radiotagged pregnant female white-tailed deer in winter to identify birth site locations relative to burned areas on the landscape and measure the subsequent survival of the neonates. Each female was fitted with a vaginal implant transmitter (VIT) to aid in the discovery of birth sites and hours-old neonates. We fitted each neonate with an expandable, breakaway VHF collar that had a 4-hr motion-sensitive mortality switch. We monitored neonates intensively (i.e. every 4–8 hrs) for the first month of life via VHF and continued monitoring survival at reduced time intervals until fawns reached 16 weeks (Chitwood et al. 2015a). Thus, survival of neonates was our proxy for reproductive success in this study. When we detected a mortality signal from the collar, we tracked to the collar to determine cause of mortality using field evidence and, when predation was evident or suspected, DNA swabs for residual predator saliva on the carcass and/or radiotag (Chitwood et al. 2015a). We necropsied all carcasses to finalize cause of mortality; individuals with no signs of predation that had lost body mass since capture and had empty digestive tracts were classified as starvation (Chitwood et al. 2015a). We used the birth site location of each neonate to calculate a straight line distance to the nearest area burned during the lightning season (i.e. June in the study area). This allowed us to determine if proximity to areas burned in the lightning season affected the likelihood of neonate starvation. Using a binary logistic regression in JMP Pro 11.0, we used the straight line distance from each birth site to the nearest area burned during the lightning phenology to predict the probability of starvation. Our rationale for using straight line distance to areas burned in lightning season was that in this resource limited environment, which does not contain anthropogenic subsidies or agriculture, the predicted pulse in available nutrients following fire would serve as the highest quality foraging opportunity for lactating females in this system and thus, serve as a primary means to meet the demands of lactation (Chitwood et al. 2015a, 2017, Lashley et al. 2015b, Nichols et al. 2021). All protocols presented herein were approved by the North Carolina Wildlife Resources Commission and the NCSU IACUC (no. 10-143-O).