Head-started Agassiz’s desert tortoises Gopherus agassizii achieved high survival, growth, and body condition in natural field enclosures

We measured survival, growth, and body condition of 8 hatchling cohorts of desert tortoises Gopherus agassizii (living in predator-resistant outdoor pens in the Mojave Desert, California, USA) over 11 yr to evaluate head-starting methods. At 11 yr of age, 7 times as many of the first cohort had survived than if they had been free-living tortoises. Subsequent improvements in predator control, food and water supplementation, and pen structure increased survival from 7 to 10 times that under wild conditions in younger cohorts. Annual survival averaged 96%. Carapace length (CL) increased 6.95 mm yr-1, similar to that of free-living tortoises. Annual growth rates varied with calendar year (possibly reflecting food and water supply), age, cohort (year hatched), mother, and in 4 dry years, with crowding. Most of the first cohort grew to a releasable size (CL >100 mm) by their 9th year. Body condition indices remained high, indicating little dehydration despite droughts in 8 of the 11 years, because irrigation offered drinking opportunities. Head-started tortoises developed fully hardened shells (≥98% of adult shell hardness) earlier (10.1 vs. 11.6 yr), but at a larger CL (117 vs. 104 mm) than did free-living tortoises. Selective feeding in head-start pens decreased subsequent germination of favored wildflower species, apparently by reducing the natural seedbank. Consequently, we reseeded and irrigated each autumn to promote subsequent spring food supply. We irrigated in early summer to enable drinking and ensuing consumption of dry, dead plants and Bermuda grass hay, a supplement. These procedures can greatly improve juvenile survivorship, and increase numbers of hard-shelled, midsized juveniles to help augment wild populations.


INTRODUCTION
Head-starting is one means to augment populations of Agassiz's desert reducing or eliminating death by predation and the physical environment (e.g., dehydration and starvation during droughts) and by fostering juvenile growth and resistance to these threats (Morafka 1994, Morafka et al. 1997, Nagy et al. 1997, McGovern 2019. A primary goal is to add older, larger juveniles, with higher survival probabilities, to wild populations to enhance reproduction rates of local females and improve population rates of natural increase. This goal is consistent with specific recovery guidance under the Endangered Species Act (USFWS 1994(USFWS , 2011a. Desert tortoises have low egg and juvenile survivorship (Turner et al. 1986, 1987a, Karl 1999, Bjurlin and Bissonette 2004, but have high adult survivorship and long lives (Turner et al. 1984, 1986, 1987a, Curtin et al. 2009), as do many chelonians (e.g., Wilbur and Morin 1988, Congdon and Gibbons 1990, Kuchling 1999. But adult growth is slow, sexual maturity is reached late (ca. 12-20 years in female desert tortoises; Turner et al. 1987b, Germano 1994, and females have low fecundity (ca. 8 eggs per year;Turner et al. 1986, Mueller et al. 1998, Wallis et al. 1999, Lovich et al. 2015. These We built predator-resistant enclosures in an area of high tortoise density (Woodman et al. 2001) and good-quality tortoise habitat (Barrows et al. 2016) containing creosote bush, white bursage and galleta grass vegetation in the south-central Mojave Desert. The fenced enclosures, with overhead netting, excluded most terrestrial and avian predators. Initially, we kept living conditions as natural as possible for enclosed juvenile tortoises (i.e., native vegetation and substrate) to optimize their fitness after being released. However, subsequent droughts depleted native food supplies and threatened juvenile health and survivorship, so we supplemented food and water. The two main 'head-starting' factors we provided to juveniles were protection from predation and provision of adequate plant food and drinking water via a sprinkler irrigation system. We evaluated the effects of these relatively simple treatments on annual survivorship and growth while inside head-start enclosures, and compared them to the age-specific survivorships and growth rates of free-living G. agassizii (Turner et al. 1987aand 1987b, Medica et al., 2012, Nagy et al. 2015b). Additionally, we measured how shell hardness, which should convey resistance to predation by common ravens (Corvus corax) and other predators (Nagy et al. 2011), changed with body size and how it and growth were affected by irrigation. as pens for the females' hatchlings during their first winter. This allowed us to identify the mother of every hatchling (Nagy et al. 2016). Subsequently, we enlarged some 7.7 x 7.7 m pens by removing partition sections to form larger communal enclosures for entire juvenile year classes (or cohorts). We also moved some entire cohorts into larger, undivided enclosures. All enclosures used a sprinkler system to (1) supplement rainfall and (2) encourage plant germination and growth during drought years. Our rain gauges placed within sprinkler footprints measured an average of 12.7 mm (range 5.1 to 22.4 mm) of water per hour. When necessary, native fire ant colonies were controlled with ant-species-specific poison bait in tortoiseproof stations.

Study site and weather
We used weather records from the National Weather Service (NWS) station nearby, approximately 10 km away, and having similar elevation and topography (Nagy et al. 2016). Using these records we calculated long-term average annual rainfall. Because desert rains can be localized, we installed three rain gauges within TRACRS to obtain more accurate precipitation data.
Annual rainfall data are reported as total precipitation measured between 1 October and 30 September the following year. Prior to this study, URTD (Upper Respiratory Tract Disease), and specifically mycoplasmosis, was present in some tortoises at Sand Hill, so we used aseptic handling techniques (U.S. Fish and Wildlife Service, 2009). We examined female tortoises for clinical signs of URTD and took blood samples to quantify antibodies to Mycoplasma agassizii and M. testudineum (Christopher et al. 1999 1999, Christopher et al. 2003]. All juveniles hatched at TRACRS had negative ELISA results for Mycoplasma agassizii and were subsequently moved to experimental enclosures.

Food availability and food supplementation
We measured food availability as plant cover to avoid harming the food supply to our tortoises. Mid-April of 2006 to 2018 we measured annual plant cover (zero to 100%) of combined native and exotic species growing in one of our enclosures. We used a 1 m 2 Daubenmire square of PVC (polyvinyl chloride) pipe gridded (100 0.01m 2 squares) using fishing line cross-strung at 10-cm intervals. In each of the 24 pens, we placed the square within the irrigation spray zone and then outside the spray zone, with placement judged to capture the representative cover in the spray zone and in the nonspray zone. For each year we calculated average annual plant cover as a) the sum of % cover for all annuals in each m 2 plot (n = 48), b) as the mean of dry (not irrigated) and wet (irrigated) % cover for each 7.7 x 7.7 pen (n = 24 each), and c) as the mean % cover for all 24 pens. We also measured species richness as the number of forb species in each m 2 plot and total numbers of herb species (i.e., forbs and grasses) in the enclosure (sum of all species in the 48 Daubenmire plots). Due to drought-induced shortages of herbs in spring of 2007, 2012 and 2013, we provided potted, nursery-grown plants (African daisies, Osteospermum spp. and Bermuda grass, Cynodon dactylon) to enclosures as needed. At the end of those seasons, we removed the potted remains. Beginning in 2012, we sowed seeds of native forbs (Malacothrix glabrata, Chaenactis fremontii, Plantago insularis, and Salvia columbariae) every autumn to replenish the soil seed bank with food species. Beginning in summer 2013, we added dry Bermuda grass hay to all occupied enclosures to supplement the dry herbs that tortoises were eating then. Head-start tortoises ate each of these species.

Egg procurement
Each spring from 2006 through 2016 we radio-tracked wild, adult female tortoises (transmitter model AI-2, Holohil Systems) that lived within 5 km of the head-start facility and used some ELISA-negative females as egg donors inside TRACRS. We examined females for oviducal eggs (via palpation and primarily by x-ray radiography; MinXray Portable models HF8015 and X750G;Gibbons and Greene 1979;Wallis et al. 1999). When a female's radiograph showed moderately to heavily shelled (calcified) eggs, we transferred her to an individual TRACRS pen to oviposit. We provided each female at least two burrows to use as refugia or for nesting. Some females dug additional burrows. We avoided close monitoring of females to avoid influencing when and where they nested, and we did not search burrows for nests so as to avoid disrupting or altering egg placement and nest conditions. Both of these may influence incubation temperatures, potentially altering hatchling sex ratios (see Nagy et al. 2016 and references therein) and nest success. Females oviposited after 1 to 4 weeks at TRACRS, were offered water to drink, and were released to their home burrow.
Radiography confirmed that many egg-donor females produced second clutches after being released.

Nest and egg success
We recorded the hatchling emergence events for each nest laid, and we uniquely marked emerging hatchlings on vertebral scutes with a permanent marking pen (Sharpie TM ) and a small printed label epoxied to one scute. Emergence success was calculated as number of hatchlings emerging out of the number of eggs laid (typically equaling clutch size from radiographs). We compared our emergence success measurements with the life table value in Turner et al. (1987a) for free-living tortoises at Goffs, California.

Juvenile survivorship
We captured most juveniles twice each year, in Spring (late March, early April) and in Autumn (late August, early September), to measure survival and growth. Despite extensive search efforts, if we repeatedly ceased detecting individuals, we assumed they died shortly after they were last seen alive. Annual survivorship calculations were based only on those juveniles confirmed visually to be alive or dead (or repeatedly missing and assumed dead) a year later, and were compared to age-specific survivorship estimates for free-living juveniles (Turner et al. 1987a;Bjurlin and Bissonette 2004).

Growth measurements and analyses
We measured body mass to 0.1 g using portable digital scales and standardized orthogonal, straight-line shell dimensions to 0.1 mm with digital calipers (Nagy et al. 2002). These included carapace length (CL, the distal measure at nuchal and supracaudal scute notches), shell width (SW, the distal measure at notches between left and right marginal scutes 5 and 6), and shell height (SH, the distal, vertical measure of plastral and carapacial scutes measured perpendicular to the SW measure). We based growth measurements on shell lengths rather than body masses, which can vary widely due to differences in hydration, reproductive mass and gut fill rather than somatic growth (Nagy and Medica 1986, Jacobson et al. 1993, Henen 1997, Nagy et al. 2002. We analyzed growth rates as annual changes in carapace length (CL), tested these for effects of age, year, cohort, mother and individuals, and we compared annual growth rates with those of juveniles in three wild populations (see 3.4 below for details).

Biomass density effects on growth rate
We used linear least-squares regression (using SPSS-Statistical Package for the Social Sciences) and analysis of covariance (ANCOVA, Zar 1999) to evaluate relationships of CL growth rates (i.e., mm y -1 , from autumn to autumn) to tortoise biomass density (= biodensity: g tortoise per m 2 ground surface) among pens. First, we estimated growth rates as the mean increase in CL per year for 19 to 37 pens per year, with a minimum of three individuals per pen. We estimated each pen's biodensity as the sum of autumn body masses, at the beginning of the 12 months, of all individuals in that pen, divided by ground surface area of that pen.

Body condition and shell hardness indices
We calculated body condition index (BCI) as the ratio of body mass (g) to shell volume (cm 3 ) estimated as the product of standardized carapace length, width and height (in cm, Nagy et al. 2002). For juveniles this index varies primarily with hydration state and gut fill. We analyzed autumn BCI (late August, early September) of all juveniles each year, as growth rates were greatest in spring, which is when body mass fluctuates considerably.

Statistics
Results are indicated as mean + standard deviation (SD) and sample size (n). Differences between means were evaluated using two-tailed Student's t-tests and one-way or two-way ANOVA (analysis of variance), considering probability values p < 0.05 statistically significant. We used least-squares linear regression to evaluate relationships between annual growth rates and plant cover as an index of annual food supply, precipitation (rain plus irrigation) as an index of annual water supply, and pen-specific tortoise mass (biodensity) as an index of annual food demand. We used General Linear Models ANOVA to evaluate main effects of year, age, cohort and mother on growth rates, and nested (or hierarchical) ANOVA to evaluate age, cohort and mother effects, nested with effects of year, because the dataset was not crossed (Zar 1999). One-way Repeated Measure ANOVA provided the same main and post-hoc results as one-way ANOVA with mother as a random factor.
To compare regression slopes and elevations, we used analysis of covariance (ANCOVA; Zar 1999). We used non-linear, least-squares analyses (SigmaPlot 11) to analyze fit for curvilinear relationships (e.g., exponential rise to asymptotes), and Spearman Rank Order correlations (r s ) for other nonlinear, heteroscedastic and non-normal data. We used Z tests to compare regression correlation coefficients (r 2 ; Zar 1999).
Juvenile shells harden asymptotically to adult values (Nagy et al. 2011), so the difference between juvenile and adult SHI (the same as compressibility) converges asymptotically on zero as juveniles grow.
Logarithmic transform of this convergence results in linear relationships to tortoise age (years) and size (CL), enabling us to estimate the age and size at which juvenile shells reach 98% and 99% of adult SHI. We also used ANCOVA to compare SHI-to-CL regressions, and SHI-to-Age regressions, to evaluate their shell hardening trajectories to those of head-started tortoises experiencing only natural rainfall (Nagy et al. 2011). If slopes were similar among groups, we compared their elevations (at p < 0.05; Zar 1999). If two regression slopes differed significantly (t-test with p < 0.05), we used a Zerbe test (Zerbe et al. 1982, Loehr et al. 2006) to determine at which covariate values (CL or Age) the two groups differed in elevation. We used the same procedures to test for growth rate differences among sites. Annual plant cover varied from less than one percent to over 36 %, and forb species richness varied from two to 19 species (Table 1). In 2012-2013, irrigation after January did not stimulate new germination.  Table 2). Annual emergence success was not correlated with rainfall, irrigation amounts, or average air temperature during mid-incubation (all p > 0.39). Our head-start process improved emergence success (73.8 ± 5.2 %, n = 8) compared to those in wild conditions at Goffs, California (55.2 %; t 7 = 10.15, p = 10 -5 ). With one exception (see Juvenile survival below), vertebrate predators apparently did not enter the enclosures. In regular inspections of nesting pens and burrows, we saw no nest disturbance or egg predation by vertebrates or invertebrates, including fire ants (Solenopsis xyloni) or other ant species, and no indirect evidence of predation (e.g., broken eggshells, dead embryos, digging, or footprints).
Annual survival was nearly 100 % in the latter years (2014)(2015)(2016)(2017). For the three oldest cohorts, survival to nine years, the age when about half of the surviving individuals were large enough to release, was 48.6 juveniles per We were able to determine causes of death for some juveniles. If carcasses were not found, absences from spring and autumn censuses indicated the tortoise died underground during winter. Inspection of the few smaller juvenile carcasses we found implicated death by ants and beetles attacking the exposed yolk sac and umbilicus, or soft, moist or incompletely closed umbilical scar. Additionally, older juveniles that were overturned or trapped in vegetation or fencing, likely overheated in the sun and died. We detected no dead juveniles that may have frozen after emerging to drink winter rain. The lowest annual survivorship (66.7 %, Table 2 these tortoises died of thermal exposure. We suspect these structures also contributed to the death of tortoises we found upside down near these bars.
Following the removal of these bars, and continual netting inspection and repair efforts, mortality rates stabilized at very low levels in this enclosure.

Annual growth rates
The mean of all annual growth rates was 6.95 mm y -1 (SD + 3.55, n = 3361). Annual growth rates of the eight cohorts averaged from 1.25 mm y -1 to 12.92 mm y -1 (Table A1). Individual growth rates were more variable, ranging from -2.4 mm y -1 (a shrinking shell) to more than +20 mm y -1 . Even growth rates among clutch mates varied considerably. For example, growth rates of five clutch siblings living together during the 2006-2007 year varied from 1.23 to 12.04 mm y -1 (mean 6.71 mm y -1 + 4.21).
One-way ANOVA indicated a strong maternal effect (F 49,3311 = 2.510, p < 10 -8 ) among the 50 mothers, but the p -value was larger than those for other univariate ANOVA (< 10 -30 ). Also, the offspring of only five mothers had significantly low (n = 1) or high (n = 4) means in SNK post-hoc tests. The one mother's group with low growth rates (mean = 5.01, SD = 0.350, n = 18) represented six years of data for 3 hatchings of 2011, which had the lowest rates of all cohorts (Fig. 5), occurred during the early, slow growth ages (Fig. 4), and hatched at the beginning of a five-year drought. The three mothers with high offspring growth rates (20 hatchlings, weighted mean = and did not differ statistically from each other (ANCOVA t 19 = 1.131, p > 0.13). Growth rates were lowest at Fort Irwin (enclosed: 4.19 mm y -1 ; freeranging: 4.38 mm y -1 ) and did not differ from each other (ANCOVA t 10 = 0.363, p > 0.36). The growth rate via regression slope of TRACRS juveniles during their first 11 years (6.41 mm y -1 ) was lower than those at Rock Valley and Goffs (both ANCOVA t > 8.7, p < 0.000001, df = 19 and 20, respectively), and higher than those at Fort Irwin (head-started ANCOVA t 15 = 6.41, p < 10 -5 ; free-ranging t 15 = 6.12, p < 1 x 10 -6 ). Thus, the overall rate of growth of juveniles at TRACRS was intermediate between those of juveniles living in their natural habitats.

Density Effects on Growth Rate
Initial (previous autumn) biomass densities in separate pens ranged from a low of 0.27 g tortoise m -2 in 2006 to a high of 18.10 g m -2 in 2016.
The only significant relationships between annual growth rate and initial juvenile biomass density were negative relationships that occurred during four years (2009-2010, 2010-2011, 2012-2013, and 2013-2014). We hypothesized that any year having a substantial food shortage should show a downward deflection in growth rate at high biodensities, beyond the point where food supply exceeds food demand. However, the results do not support this hypothesis; there was no clear deflection point or threshold (Fig.   8). Growth rates during 2012-2013 were unusually low (also see Table A1).

Body Condition and Shell Hardness Indices
The mean autumn BCIs ranged from 0.48 to 0.58 g body mass cm -3 .
There were two instances where mean BCI values were slightly below 0.   tortoises. Additionally, we calculated the expected ages when unirrigated and irrigated juveniles reached 98% and 99% of adult SHI (Table 4). Shells of irrigated juveniles hardened 1.5 years faster, and at a larger size (12% larger), than did shells of unirrigated tortoises.

Survivorship
We demonstrated head-starting's ability to substantially enhance nest and egg success, and to increase juvenile survival when compared to the wild. Consequently, head-starting can potentially augment Agassiz's desert tortoise (Gopherus agassizii) populations and species recovery (USFWS 2011a) by providing releasable healthy juveniles that are past their highest mortality stages.
In natural habitats, emergence success (percent of emerging neonates per 100 eggs laid) was 55.2% (at Goffs, Turner et al. 1987a) and 68.9% (in the Sand Hill Training Area [Sand Hill] < 5 km from TRACRS, Bjurlin and Bissonette 2004). These two studies documented substantial nest predation by vertebrates (37% at Goffs and 26% at Sand Hill). At Goffs, egg and nest mortality were attributed to 1) infertile eggs (6.1 %), 2) broken eggs (6.6%) and 3) nest predation (32. 1%;Turner et al. 1987a). At TRACRS, we observed no signs of vertebrate predation on nests, and emergence success was relatively high (73.8 % over 8 years; Table 2), but was not as high as expected from Goffs results in the absence of predation (87.3%). It is  Berry et al. 2013], but may also be affected by factors that can reduce fecundity (e.g., food availability; Turner et al. 1986, 1987a, Henen 1993, 1997, Lovich et al. 2015 and slow juvenile growth, among other causes. Consequently, head-starting may be central to bolstering the declining natural populations of Agassiz's desert tortoises. However, biologists express concern about post-release survival (e.g., Heppell et al. 1996, Siegel and Dodd 2000, Reed et al. 2009), with models suggesting that head-starting cannot, logistically and numerically, augment and sustain populations. We anticipate measuring survival rates of released head-started TRACRS juveniles in an effort to evaluate these uncertainties.

Mortality inside enclosures
Besides preventing deaths by large predators, dehydration and starvation, we reduced juvenile deaths several other ways. We eliminated detectable nest predation by ground squirrels by constructing fences with bands of slippery metal sheeting that rodents could not climb (Nagy et al.

Growth rates
Despite irrigating to counter drought conditions (i.e., to hydrate tortoises and promote food plant growth), our juveniles' overall growth was not higher than average rates in wild juvenile tortoises but was comparable to rates in other irrigated, head-start facilities (4.2 mm y -1 to 11.9 mm y -1 ;

Free-living versus head-start
Average annual growth rates of free-living juveniles in three natural populations were both higher and lower than the average growth rate at TRACRS (Fig. 7). The growth rate differences between the three field populations appear positively related to variation in annual rainfall amounts: 2012, Hillard and Nagy unpublished obs.). At TRACRS, we tried to achieve a "good" herbaceous production year every year by irrigating. So why did TRACRS juveniles not achieve higher growth rates? To address this, we examined the relationships between growth, total precipitation (rain plus irrigation), food availability, and tortoise biomass density.

Precipitation and food availability
Growth responses to precipitation (rainfall and irrigation) and food supply are essential to an understanding of the effectiveness of headstarting efforts and general tortoise biology. Although these responses are complicated or obscured by the large variation in growth rates (Table A1) (Bradshaw 1988 and1997) and other ectotherms (Pough 1980), and central to their species' success in arid environments.
Nonetheless, the growth rates here correlated strongly to plant cover (an indicator of food supply) and correlated mildly to precipitation. irrigated each fall immediately after seed sowing regardless of weather forecast and realized rainfall. This method increased plant cover and juvenile growth rates above those of the first seven years (Fig. 10). Plant species richness also increased in response to regular irrigation and seeding.
We also irrigated briefly, 30-60 minutes, during summer so tortoises could drink, eat the available dry plants (Nagy and Medica 1986, Henen 1997, 2002, and eat the Bermuda grass hay we began providing each summer starting in 2012. With these modifications in irrigation, seeding and dry food supplementation, growth rates in TRACRS increased to levels seen in freeliving juveniles during "good" years (Medica et al, 2012).

Biomass density and food supply
If growth rates were limited just by food supply and not influenced by food quality, we would suspect that annual growth rates would be lower in pens with greater densities of tortoise biomass. However, this did not occur in seven of the 11 years, and in three of the other four years (2009( -10, 2010( -11, 2013, growth rates were near average (Table A1). Except for the very low growth rate in the year with the lowest plant cover, 2012-13, growth rates in the other 10 years varied little and seemed to plateau despite increasing food availability (i.e., plant cover; Fig. 6), suggesting food availability rarely limited growth. At Edwards Air Force Base (EAFB) in 2010 to 2012, head-start juveniles had very low growth rates (3.7 mm y -1 ), but also had low condition indices, poor health, lethargy and high mortalities despite  (Fig. 8). Herbaceous plants available at EAFB's facility, comprised primarily of three non-native annual grasses of low nutritional quality (Hazard et al. 2009, 2010), support Mack et al.'s (2018 suggestion that EAFB pens lacked sufficient mass of preferred herbs to sustain the animals.

2017) at an irrigation-equipped head-start facility that had a plant population
with good native species richness and cover. There, first-year juveniles also showed maternal effects on growth rates, with larger mothers producing larger hatchlings that grew faster and had higher survivorship (Nafus et al.

2015).
Head-start overcrowding may limit juvenile growth inside head-start facilities via more than one means. Although maximum biodensities inside TRACRS (12-13 g m -2 ; Fig. 8) were more than 100 times that of wild adults in the surrounding habitat (0.097 g m -2 , assuming 100 2.5-kg adults mi -2 , Woodman et al. 2001), growth rates inside TRACRS were low in only one of 11 years, and that was during the most severe food paucity (Fig. 8).
Consequently, extreme food paucity and food composition (e.g., low availability of preferred forbs and grasses) provides one estimate of maximum biodensity in head-start enclosures. Qualitatively, we detected early seasonal reductions of specific food plant species as juveniles emerged from winter brumation. We suspect they consumed their 'preferred' foods as seedlings, before those plants could grow, provide a larger source of food, and set seed that sustains the seed bank. Additionally, the remaining, 'less-preferred' plant species would be consumed less vigorously, subsequently propagating and competing with preferred species. Wild juveniles have much lower constraint on movements, so wild juveniles should have access to much greater areas to forage selectively on more nutritious foods.
In order to be of a manageable size and yet produce useful numbers of large juveniles for release, head-start facilities can have biomass density constraints that reduce food plant diversity, food plant productivity, and soil seed banks. We have countered these reductions by sowing seeds of preferred food plants in autumn and early winter and irrigating deeply and regularly after sowing through the ensuing May. Additionally, we controlled some seed-eaters (ants and rodents) but not small birds that easily ingressed through cyclone fencing and overhead nets.

RECOMMENDATIONS AND CONCLUSIONS
The conclusions and recommendations below are based primarily on results in this study, located in the south-central part of the Mojave Desert.
Rainfall patterns and average annual precipitation amounts vary widely across the species range, from relatively high winter rainfall in the western areas to low, mostly winter rainfall in the central Mojave area, and to relatively high summer rainfall in eastern parts of the range. Similarly, the species composition of tortoise food plants varies from mainly "winter annuals" in the west to a mixture of "winter" and "summer" annuals in the east (see Henen et al. 1997, Wallis et al. 1999, andreferences therein). We suggest that the recommendations below, which are based on results from a relatively low rainfall area with mainly winter rainfall, be applied considering regional differences in climate. In hindsight, some of these recommendations may now seem obvious, but our initial research strategy was to start with a protocol that was current, low-cost and simple to operate, but with the caveat to adaptively manage with more intensive procedures as our results indicated necessary for successful head-starting. 1). To help sustain soil seed banks in head-start enclosures, seeds of preferred plant species should be sown in autumn (October and November) of each year.
2). After sowing seeds, irrigation should commence in October or November, as desert rainfall is, and forecasts are, extremely variable and unpredictable (Louw & Seely 1982, this study).
3). Tortoises were more apt to remain above ground, eat and grow if they were hydrated (Nagy et al. 2015a). To provide drinking water, we recommend irrigating for at least 30 minutes as juveniles emerge from brumation in March, several times in spring (especially during droughts), in early June before summer heat arrives, and in late August and September to enable drinking and eating before brumation. Dehydrated tortoises may emerge in winter rains to drink, and subsequently they may be prone to die ). The bird netting overhead must be inspected frequently to discover and repair degradation and damage caused by sunlight, strong winds, heat, and gnawing rodents, that enables bird depredation of juveniles. 10). Avoid overstocking tortoises (e.g., biomass density > ~ 5 g m -2 ), which compromises tortoise growth, health, and survival, especially during droughts, and during conditions of low food diversity, quality and abundance.
11). This study shows that a head-start facility containing natural habitat and relatively high densities of young Agassizi's desert tortoises can be operated to promote good growth rates. These operations produce wellhydrated 11-year-old juveniles that number 7 to 10 times more than would survive from the same number of eggs laid in the wild.   Nested ANOVA (Nested) used variables age, cohort and mother nested within year (e.g., Mother-year indicated mother nested with year). Degrees of freedom indicated by df 1 and df 2 . All p < 10 -30 except for the simple nesting of mother within year (Mother-year*, p < 10 -15 ). Other forms (e.g., age, cohort and mother simultaneously nested within year) were incomplete, unbalanced designs.    and ensuing pre-brumation predation was high. . Relationship between shell hardness (SHI) and shell size (carapace length, CL) of one to 11 year old desert tortoises living in naturalhabitat enclosures, and experiencing natural rainfall plus irrigation. The red curve represents the transformed linear regression equation calculated for a semilog analysis. The dashed blue line represents a fully-rigid adult shell defined as being incompressible by our digital micrometer.