Use of auto‐germ to model germination timing in the sagebrush‐steppe

Abstract Germination timing has a strong influence on direct seeding efforts, and therefore is a closely tracked demographic stage in a wide variety of wildland and agricultural settings. Predictive seed germination models, based on soil moisture and temperature data in the seed zone are an efficient method of estimating germination timing. We utilized Visual Basic for Applications (VBA) to create Auto‐Germ, which is an Excel workbook that allows a user to estimate field germination timing based on wet‐thermal accumulation models and field temperature and soil moisture data. To demonstrate the capabilities of Auto‐Germ, we calculated various germination indices and modeled germination timing for 11 different species, across 6 years, and 10 Artemisia‐steppe sites in the Great Basin of North America to identify the planting date required for 50% or more of the simulated population to germinate in spring (1 March or later), which is when conditions are predicted to be more conducive for plant establishment. Both between and within the species, germination models indicated that there was high temporal and spatial variability in the planting date required for spring germination to occur. However, some general trends were identified, with species falling roughly into three categories, where seeds could be planted on average in either fall (Artemisia tridentata ssp. wyomingensis and Leymus cinereus), early winter (Festuca idahoensis, Poa secunda, Elymus lanceolatus, Elymus elymoides, and Linum lewisii), or mid‐winter (Achillea millefolium, Elymus wawawaiensis, and Pseudoroegneria spicata) and still not run the risk of germination during winter. These predictions made through Auto‐Germ demonstrate that fall may not be an optimal time period for sowing seeds for most non‐dormant species if the desired goal is to have seeds germinate in spring.

However, tracking seed germination timing in the field can be challenging, resource intensive, and time-consuming. Additionally, knowledge gained from short-term field germination studies is often lacking due to high annual variability in weather conditions at the time of the experiment (Hardegree, Jones, Roundy, Shaw, & Monaco, 2016). Subsequently, to gain general inferences from germination studies, labor-intensive studies need to be repeated for multiple years.
Progress toward germination for many cool-season species can be predicted through a wet-thermal accumulation model where soil moisture must exceed a base water potential (Ψ b ) for germination to occur (Finch-Savage, Steckel, & Phelps, 1998;Rawlins, 2009;Rawlins, Roundy, Davis, & Egget, 2012;Roundy, Hardegree, Chambers, & Whittake, 2007). The base water potential used is derived through laboratory experimentation (Roundy et al., 2007). Though there are many factors that influence the rate of seed germination and number of germinable seeds, adjusting Ψ b is expected to correct for impacts from environmental conditions, after-ripening and seasonal changes in dormancy cycling (Bradford, 2002). Subsequently, once Ψ b is determined, seed germination timing and number of germinable seeds may be accurately predicted from soil temperature. Field trials have validated wet-thermal accumulation models (Rawlins, Roundy, Egget, & Cline, 2012;Rawlins, Roundy, Davis et al., 2012), and confirmed their utility in predicting seed germination in a number of settings, with a wide variety of species (Cline, Roundy, & Christensen, 2018a,b;Hardegree, Sheley et al., 2016). Despite the simplicity of wet-thermal accumulation models, a relatively large amount of data and processing is required to develop the models and estimate seed germination timing in the field.
To overcome the logistical challenges associated with predicting seed germination timing, we created a programmed workbook called "Auto-Germ" that allows users to efficiently process seed germination data and predict seed germination timing in the field. Our workbook utilizes Visual Basic for Applications (VBA) in Microsoft Excel (Microsoft Corporation, Redmond, Washington, USA) to create wet-thermal accumulation models as well as calculate various other germination indices from laboratory constant temperature trials. Auto-Germ also provides users with an interface to apply the wet-thermal accumulation models to estimate germination timing in the field from historic soil moisture and temperature data sets.
Auto-Germ's predictive germination modeling capabilities have the potential to educate practitioners in knowing how their planting dates may influence germination timing and subsequently the growing conditions that impact seedling establishment. The Artemisia spp. (sagebrush)-steppe ecosystem in the Great Basin region of the western United States is an example of an imperiled ecosystem that would benefit from improved restoration practices (Hardegree, Jones et al., 2016;Suring, Rowland, & Wisdom, 2005). In this region, seeding is used to reclaim degraded sites that have been impacted by wildfires, invasive species, and various human disturbances (Davies, Bates, Madsen, & Nafus, 2014;Knick et al., 2011;Noss, 1995). In the Artemisia-steppe, seeding typically occurs in autumn, with the expectation that seeds will remain dormant in the soil and then germinate in the spring (Crawford et al., 2004;Madsen, Davies, Boyd, Kerby, & Svejcar, 2016;Richards, Chambers, & Ross, 1998). However, planting too early in the year can result in seeds germinating prior to winter and then experiencing high mortality over the winter period (James & Svejcar, 2010). Winter mortality may occur as a result of freezing conditions (Boyd & Lemos, 2013;Rinella, 2011). Roundy and determined that across 14 Artemisia-steppe sites there was an average of 58 freeze-thaw periods for the upper 1-3 cm of soil between October and March. Seedbed freezing conditions have been shown to alter the physiological responses of Artemisia tridentata Nutt. (Asteraceae) (big sagebrush) in the Great Basin (Loik & Redar, 2003), and has the potential to further inhibit plant survival of perennial grasses such as Pseudoroegneria spicata [Pursh] A. Love (bluebunch wheatgrass) (Boyd & Lemos, 2013). Mortality may also occur to seedlings over the winter period as a result of drought, pathogens, and expenditure of seed carbohydrate resources (James et al., 2011;Madsen et al., 2016). Subsequently, in this region understanding the seeding date required to prevent premature germination and subsequent winter mortality is paramount to improve the effectiveness of restoration projects.
Our objectives were to provide instructions on how to use Auto-Germ and demonstrate the utility of the program through a case study that (a) calculated various germination indices under different constant temperatures on 10 different species commonly used for restoration projects in the Great Basin and (b) for these same species model seed germination timing across 6 years and 10 Artemisiasteppe sites to estimate the planting date required for 50% or more of the simulated population of seeds to germinate in spring (March 1st or later) when conditions are predicted to be more conducive for plant establishment.
There are four main steps for processing data in Auto-Germ, which include: (a) entering laboratory data, (b) wet-thermal model creation, (c) entering field data, and (d) model application. Each step is initiated by clicking a button in Auto-Germ on the Home worksheet (note macros and content must be enabled to use Auto-Germ). Auto-Germ provides instructions on the Home worksheet for each step (Supporting information Figure S1).

| Step 1-Germination count data input
The first step is to input germination count data from constant temperature laboratory trials into the Data Entry worksheet (Supporting information Figure S2), which is accessed by clicking the Data Entry button. To input new data, click the Start Over button on the Data Entry worksheet. In Auto-Germ, the data organization must match the sheet setup, where column A is temperature in Celsius, column B is replicate (or block), column C is plot ID, column D is treatment, column E is the number of seeds planted per sample, and everything from column F to the right is measurement dates and their respective germination counts. The planting date is entered into cell B8. The workbook processes up to 100 germination date entries and 1,000 samples. Under each measurement date, enter the number of seeds that germinated between the last count time and the current one. Do not enter cumulative germination count data on this sheet. Entries in the columns labeled as rep/block and plot ID are optional. If the user does not want to produce wet-thermal accumulation models, germination metrics will be calculated through Auto-Germ without temperature data. Auto-Germ will not operate if empty cells are included under the columns labeled as temperature, treatment, seeds planted, planting date, and the germination measurement columns. The treatment column can be used to signify a number of different variables. For example, if seed treatments are being analyzed the type of seed treatment would be placed in this column. If species were being compared the treatment column would contain the name of the species.

| Step 2-Wet-thermal model creation
Once the data is entered, return to the Home worksheet and click the The synchrony of germination was calculated as follows: where Z = synchrony of germination C n i ,2 = combination of the seeds germinated in the ith time, two by two; n i = number of seeds germinated on the ith time.
The time to reach each percent germination was calculated as follows: where T N = time (days) to subpopulation germinatio; t a = incubation day when subpopulation germination was reached; t b = incubation day before subpopulation germination was reached; n a = number of germinated seeds on day that subpopulation germination was reached; n b = number of germinated seeds on day before subpopulation germination was reached; N = number of germinated seeds equal to the percentage of the total subpopulation of interest.
The Data Averages worksheet displays the same metrics for the average of each treatment and temperature combination. The Standard Error worksheet displays the standard error for each calculation on the Data Averages worksheet. The Polynomial Equations worksheet contains second order polynomial equations with their associated coefficient values (A, B and C), the R 2 value for each germination percentage of each treatment, and the corresponding graphs depicting germination rate as a function of temperature (Supporting information Figure S3). To create new polynomial equations the newly created sheets need to be exported or deleted.

| Step 3-Field data input
To estimate seed germination timing in the field from the polynomial equations, the user needs to create worksheets containing field soil temperature and water potential data. is temperature, and column C is water potential. The user must input their own field data worksheets to apply the model. The field data worksheets must be located in-between the Home and Data Entry worksheets. If there are any other worksheets besides field data in this location, the program will not operate correctly.

| Step 4-Field germination predictions
At this point, two options are available for the user to choose from. The first option is to predict the time to reach the previously specified germination percentages based on a planting date.
The second option is to predict the dates a certain germination percentage is reached based on a range of planting dates. Before clicking either button, make sure that steps 1-3 are complete and that the Polynomial Equations worksheet is located in the workbook somewhere after the Data Entry worksheet. If Polynomial Equations are missing or has a changed name, Auto-Germ will not operate.
To predict the times to reach the previously specified ger-  Each table corresponds to a field data sheet. The graphs of the tables are located on the right.

| Workbook Options
Workbook Options is the last heading on the Home sheet. The View Data button will create a new workbook that contains all of the data generated from steps 2 and 4, but will not remove any new worksheets. The new workbook containing generated data may be saved. The Export Data button will export the data that was generated in steps 2 and 4 to another workbook that can be saved, and data will be removed from Auto-Germ. The Start Over button will completely reset Auto-Germ and delete all the data generated, but will not affect worksheets located before Data Entry.

| Laboratory methods
We developed wet thermal-time models for 10 seedlots of species commonly used in restoration projects in the Great Basin. We in- A range of constant temperatures was used to germinate the seeds (5, 10, 15, 20, and 25°C). The study was setup using a randomized block split-plot design, with temperature comprising the split plot. Seven repetitions were used for each species, at every temperature. In each repetition, 25 seeds were placed in a 9 cm diameter petri dish that contained a single layer of blotter paper.
Five ml of water was initially added to each petri and additional water was added as petri dishes dried throughout the study. Petri dishes were closed in plastic bags by block to prevent the loss of water. Germinated seeds were counted every 1-3 days, for 60 days. Seeds that had germinated were counted, recorded, and removed from the petri dishes. Germination count data was then processed in Auto-Germ.
Auto-Germ was used to calculate final germination percentage, T 50 , synchrony, and mean germination time. We then used mixed model analysis in JMP ® (Version 13, SAS Institute Inc., Cary, NC, USA) to first determine the significance (p ≤ 0.05) of these four indices with respect to species, incubation temperature, and their interactions (unless determined to not be significant). In the model, blocks were considered random, while incubation temperature and species were both considered fixed. We tested for differences in responses to species at the incubation temperatures of 5, 10, 15, 20, and 25°C using a Tukey pairwise comparison test (p ≤ 0.05). Final germination was squared and the log of T 50 , synchrony, and mean germination time was taken to normalize the data.

| Field germination predictions
Wet-thermal accumulation models for each species was applied to historical soil temperature and water potential data from the

| Germination indices
Incubation temperature, species, and the interaction between these two factors affected final germination percentage (F = 10.5,  Figure 1).
Synchrony values fluctuated greatly between temperatures for all species (Figure 1). There were five species that However, at 10°C mean germination time was lower for P. spicata by 7 days and at 25°C, T 50 was lower for A. tridentata by 2 days (Figure 2).   September), and ended as early as late November-early December (23 November, 6 December; Figure 5).

| D ISCUSS I ON
Our case study demonstrates that Auto-Germ has the potential to enable researchers to efficiently process laboratory germination data and field soil moisture and temperature data to predict various germination indices, including field germination timing. Based on these results, we anticipate that Auto-Germ will be applicable to non-dormant seeds of most species. Both land managers and researchers could benefit from this program by providing them with a better understanding of how seeds may respond to their planting sites' unique soil temperature and moisture regimes.
It should be noted that predictions developed from Auto-Germ should be used as rough assessments to help guide further research and management. Wet-thermal models used in Auto-Germ can overestimate germination rates (more so than other hydrothermal models) but these errors are expected to be minimal (Hardegree et al., 2017;. In previous studies that have validated wet-thermal accumulation models, non-linear regression equations were used from TableCurve 2D (Systat Software Inc., San Jose, CA, USA) curve-fitting program (Rawlins, Roundy, Davis et al., 2012; Roundy et al., 2007). In these studies, the R 2 values of the models ranged from 0.70 to 0.98. For our case study, a more simplified second order polynomial was used to allow processing in Microsoft Excel. This study indicated that second order polynomials provided a similar level of accuracy to predict germination timing as other models (R 2 = 0.71-0.98).
The germination indices calculated showed that individual species react uniquely to differences in soil temperature (Figures 1 and 2). For example, A. tridentata at 5°C had an extremely high T 50 and mean germination time in relation to the other species tested (almost 2× more than L. cinereus, the species with the next highest values; Figure 2). However, as the temperature increased, T 50 and mean germination time decreased to levels similar to the other species. Given this information, it is impractical for land managers to plant different species at the same date and expect similar results in germination timing.
Our case study also showed how these unique germination characteristics affected when species would germinate in the field under historic soil moisture and temperature settings (Figures 3-5).
Auto-Germ was used to calculate when 11 different species would need to be planted to have the majority of germination occur after 1 March, across 6 years, and 10 Artemisia-steppe sites in the Great Basin of North America. Looking at all species collectively by site showed that the required planting date for germination to occur after 1 March was highly variable, with planting dates ranging from September to February, due to differences in the sites soil temperature and moisture (Figure 3). The year of planting was also highly variable when looking at all species collectively by planting year, with required planting date for germination to occur after 1 March ranging from November -January ( Figure 4) These differences between species germination timing ( Figure 2 and 5) may be beneficial when applied to bet-hedging strategies surrounding seed mixes. Rinella  Our findings provide evidence that winter mortality may play a role in the lack of spring emergence seen in restoration efforts due to species germinating prior to or during the winter period and being subjected to freezing conditions. For all species except A. tridentata and L. cinereus, 50% or more of the required planting dates for spring germination occurred by November or later. This means that land managers who seed areas in mid to late fall would run the risk of having germination occur outside of more favorable spring conditions. Premature germination could potentially be mitigated by planting later in the season, however this study shows that seeding would need to take place in early to late winter. Winter seeding can be logistically challenging due to freezing and/or saturated soil conditions impacting the delivery of seed from mechanical equipment.
One potential solution may be to treat the seeds and induce seed dormancy over the winter period. Richardson (2018) demonstrated that seed dormancy can be induced through the addition of the plant hormone abscisic acid (ABA), which is applied to the seed through a seed coating. It may be possible to have seeds that are not suitable for planting in early fall treated with an ABA seed coating so that the seeds germinate in spring when conditions may be more favorable for plant establishment and growth.

| CON CLUS ION
Our research indicates that Auto-Germ provides researchers with a tool to efficiently model germination timing to understand the germination patterns of species across large temporal and spatial spectrums. As shown through our case study in the Great Basin, Auto-Germ was able to generate germination indices and predict seed germination timing in the field, over six different years, for 10 different species commonly used for restoration projects. The results of this research provide new insights into when these species should be planted and can help guide scientists and land managers in developing new restoration technologies and practices.

ACK N OWLED G M ENTS
We would like to thank Rhett Anderson and Gabriel Paulson for assisting in collecting and processing the data, as well as other members of the BYU Seed Enhancement Laboratory.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R ' S CO NTR I B UTI O N S
MM, DW, KS, and NB conceived the ideas and helped program Auto-Germ. WR and RC collected and analyzed the data. BR collected field soil moisture and temperature data. WR, MM, BR, and ZA helped write and organize the manuscript. All authors contributed to the editing of the drafts and gave final approval for publication.

DATA ACCE SS I B I LIT Y
Data are accessible via the Dryad Digital Repository, https://doi. org/10.5061/dryad.r6d4190.