1
$\begingroup$

I am investigating the elevation characteristics and sediment accretion effects of two distinct ecosystems within the Wadden Sea, impacted by the bioinvasion of one species (ecosystem one) and displacement of a native species (ecosystem two). Using high-resolution digital elevation models (DEMs) generated via structure-from-motion (SfM) from drone flights in 2020 and 2022, I aim to quantify differences in elevation and the impact on sediment accretion between these ecosystems. My DEMs were processed uniformly, from data recording to generation (SfM), resulting in the exact resolution. For this step, I solely focus on elevation values. No errors, resampling or conversions are taken into account.

Due to the bioinvasion of one species and the displacement of one native species in the Wadden Sea, the origin ecosystem is transformed into a new ecosystem. Knowing where both species can live and interact with their surroundings is crucial as both ecosystems impact their environment. Here, I focus on the elevation relative to the local mean sea level and quantify the surrounding sediment accretion enhanced by these ecosystems. To state if the ecosystems differ in their effects on sediment accretion, I want to know if there are significant differences in the accretion affected by these two ecosystems. I want to quantify the magnitude of differences and the extent of divergence between the ecosystems/sites to state if the means "differ significantly" with a p-value < 0.05. There is no certain threshold that I follow.

First, I want to compare the ecosystems' means to determine whether they significantly differ in their living conditions regarding elevation. In which elevation do the ecosystems feel comfortable to live? Do they have optimal living zones (highest abundance at a specific elevation)?

Secondly, as I have mentioned, the ecosystems impact the surrounding sediment accretion (vertical growth/volumetric changes) to varying degrees. Thus, I want to know if their impact on sediment accretion significantly differs by comparing sediment accretion results.

I plan to employ statistical tests to assess the significance of elevation and sediment accretion differences between ecosystems. My approach is so far:

  1. Normal distribution: Given the large dataset size and non-normal distribution of elevation data, I will skip normality testing and provide corresponding histograms and Q-Q plots in the Supplementary Material for reference.

  2. Variance homogeneity: The large data sets make Levene's test unnecessary. Further, the standard deviation in the descriptive statistics differs.

  3. Test on equal means: Welch t-test (two groups: ecosystem 1 vs ecosystem 2) and Welch ANOVA (study site 1 vs study 2 vs study site 3 …). I have non-normally distributed data and heterogeneity in the variance, and I am interested in comparing the means; thus, I have decided to use the Welch options. The Friedman test will be used to analyse temporal changes in elevation.

For this study, 'differing significantly' will be defined as follows:

a. H0: The elevation means of the ecosystems do not differ (p<0.05). b. H1: They differ.

  1. Post-hoc: Games-Howell test.

I seek guidance on whether this approach suits or needs adjustments. I would greatly appreciate any references or suggestions on these topics. Thank you!

Improve this question
$\endgroup$
6
  • 1
    $\begingroup$ It's not entirely clear what you are asking here. Are you just asking if the steps you have performed so far are correct? What specifically are you trying to test? There is some information about what you have and what you have done, but not explicitly what research question you are trying to answer here. $\endgroup$
    – Shawn Hemelstrand
    Commented Apr 23 at 9:48
  • 1
    $\begingroup$ I agree with @ShawnHemelstrand but, to put it another way, when you say "differ" what do you mean? Different mean? Different median? Different shape of distribution? Or what? Also, with large data sets, p values can be small when the differences are tiny. $\endgroup$
    – Peter Flom
    Commented Apr 23 at 11:00
  • $\begingroup$ Yeah I should say that your RQs are clearly listed in the beginning of the question, but it's not clear how you are seeking to fit a mathematical model to a theoretical model (e.g. Do you need ANOVA? Does a regression work here? Which one? And which pieces should I include in my model?). Part of why that is important is highlighted to some degree by Peter, who is seeking to understand which estimates you care about with respect to these differences. $\endgroup$
    – Shawn Hemelstrand
    Commented Apr 23 at 11:09
  • $\begingroup$ I want to state whether the elevations of these ecosystems can exist at the same elevation relative to the water level and seabed or only exist at different elevations. Can I find differences in their optimal elevation for living? Do the ecosystems have different living conditions and optimal growth zones regarding the elevation/water level? I'm mainly interested in the mean (average elevation values) to describe the central tendency of the habitat. I added the descriptive statistics (all in meters) to give more information. For me, it is difficult to decide which models are suitable here. $\endgroup$
    – OmteK.
    Commented Apr 23 at 12:13
  • 1
    $\begingroup$ You are right not to expect normal distribution. Thanks! My DEMs were processed uniformly, from data recording to generation (SfM), resulting in same resolution. I solely focus on elevation values (descriptive statistics). No errors, resampling or conversions are taken into account. I want to quantify the magnitude of differences and the extent of divergence between the ecosystems/sites to state if the means and variance "differ significantly" with a p-value < 0.05. There is no certain threshold that I follow. $\endgroup$
    – OmteK.
    Commented Apr 24 at 7:43

0

Reset to default