The value of mid-season census and revisions to phytometer experiments


In a set of pilot experiments, we made the following discoveries.
1. Phytometers: Use more than one species of seed to explore importance and responsiveness of treatments in desert ecosystems.
Implications: use three different species. Check phylogenetic relatedness.

2. Phytometers: Plant early and plant more seed. Dormancy, loss, predation, and potentially low germination rates can lead to very low germination and even lower numbers of established plants that can potentially respond to your treatments.
Timing: One-week before earliest ‘substantial’ rainfall event – estimate derived from long-term climatic records.
Implications: germination patterns are likely highly variable for your desert system so be safe in sowing seed early.

3. Design: When adding phytometers into the different levels of treatment, consider providing a clean-slate in terms of the seedbank.  To do this, sieve out all existing seed from the top 3-5cm of sand. If one of your treatments however includes manipulation of the existing ambient vegetation that is present within your system, consider including a procedural control for sowing protocol you do adopt. For instance, add seed to soil surface uniformly throughout  a 25cm diameter sowing ring, spring clean (seed free) sand on top of seeds – no more than 25mm, and add water stabilize seeds and sand from blowing away immediately.

4. Design: Include a procedural control in addition to treatment control particularly for cases wherein you need existing plants for competition/interference trials (see #3 above). Procedural control: disturb soil/sand similar to as if you were sowing seeds but do not add the phytometer seeds, add the clean sand on top of minimal disturbance, add water to stabilize.  Then include the usual treatment-control plots wherein there is no disturbance at all to estimate ambient seedbank effects in terms of emergent plant communities. Replicate procedural and treatment controls for every level of main factors to be certain that variation in ambient seedbank effects and your sowing disturbance effects are effectively estimated.

5. Design: A mid-season census is critical.  Typically, you do 2 very early season census surveys to capture a signal of germination and then emergence for your phytometer seed sowing. This can be up to 14 days long in many desert systems if you add water not just to stabilize your planting but also encourage ‘reasonable’ levels of germination for estimating major factor effects.  In the field, it is difficult to effectively estimate germination as the radicle and cotyledonous leaf can be within the sand. Hence, the second early season census is often the best to estimate emergence.

However, you must also have a mid-season census to ensure that you capture the signal of seedling/plant establishment.  The primary reason is that estimating survival to end of season from number of seeds sown can generate very low statistical power, i.e. 1-10% germination rates for some desert seeds can lead to very low numbers of established plants. Importantly, survival to end of season is not ‘seed’ survival but plant survival.  Consequently, it is assumed you are describing the survival of individuals that not only entered the competition/interaction arena so to speak but also established.  This season 1000 seeds sown generated 10 emergent seedlings.  Of these, only 1-2 became established plants. Using 1000 as the denominator for ‘survival’ generally does not correspond with the assumption of readers, i.e. plant survival, nor does it capture seed survival as seeds could be dormant in the soil, viable, and thus surviving. Using 10 as the denominator is legitimate but if mortality occurs pre-establishment, early in the season, then you are overestimating the capacity for interactions to become important (and reducing your estimate of survival dramatically). Ecologically, if only 1-2 plants established and then persisted to interact with other species and respond to your treatment effects, then using mid-season established plant density as the denominator is more reasonable -provided your clearly indicate maths in methods for reader and use appropriately specific terminology.

6. Design & phytometers: Do not mix the three species of phytometers.  This can introduce interference/competition and pre-emption effects that are both density dependent (i.e. if you add same mass but different numbers of seeds per species) and context or level within treatment specific differences.