Currently, I am doing a project to estimate the aboveground biomass of mangrove at the same time also taking the environmental parameter (temp, pH, DO, soil salinity). My experimental design is sampling in only one place and with only one plot. Within the plot, I divided into 4 subplots to get 6 replications of parameter readings in each subplot. How can I determine the relationship between environmental parameter and aboveground biomass ? can I perform Correlation test?
Ng pooi Yan
First, Based on your assumption consider the parameters you want to find the relation in between or want to analyse.
Second, yes the the simplest method is to perform the correlation test and to find out how various parameters are correlated (well-correlated, anti-correlated or not-correlated).
Further you can extend your observations to perform different type of analysis such as time series, geographical distribution, etc. Moreover have a look at publications titled under the key words " effect of weather parameters on biomass/crop yield "
Hope this will help you.
You may want to start by refreshing your knowledge of pseudoreplication, starting with Hurlbert's seminal 1984 paper (Article Pseudoreplication and the Design of Ecological Field Experiments
).The first question I have is how many experimental units you have. It sounds from your question that the answer may be n=1, giving you 0 degrees of freedom. If that is the case, then dividing that experimental unit into 4 subplots and taking 6 samples within each sub-plot still gives you an n of 1, and analyzing it otherwise would be pseudo-replication.
If the experimental unit is your sub-plot, that would be n=4, and you would need to average the 6 sub-samples before running a correlation analysis. But n=4 is far too few for this type of analysis.
Even if your experimental units were the sub-sampling locations, n=24 is marginal for a correlation analysis, but could be acceptable by some reviewers.
So, how do you determine the correct experimental unit? If the dependent variable is mangrove aboveground biomass, that implies that your experimental unit has to be at least large enough to encompass an entire mangrove tree but probably no larger than a population of mangrove trees in order for your measurements to be independent. Not being familiar with mangrove biology, I can't say where the independence level would be. If growth of individual mangrove trees is independent of surrounding trees, then a plot encompassing the individual tree could be large enough. If mangrove growth is similar to the temperate trees I am familiar with, then growth of one tree is strongly affected by resource availability that is affected by other nearby individuals (shade, water usage, allelopathy, etc.). In that case, you'll want your experimental unit to be an entire stand of mangrove trees or larger.
You have only one replicate i.e. the plot. To be replicated, you need other plots distantly located from each other. The subplots are subsamples, not replicates, sort of like a split-plot experiment. But you can still analyse the data. It's just that it will only have local significance. You then need to determine if your data are normally distributed by drawing histograms or plotting residuals. If they are, you can do standard correlations or regression. If not, you need to use non parametric tests such as Kendall rank correlation :)
Andrew Paul McKenzie Pegman Thanks for replying me. Yes, I already did the normality test and all data are normal distributed except salinity data. And I calculate the aboveground biomass follow Komiyama et al. formula, so that let say my plot has total 8 species and I sum up all species aboveground biomass. In this case, I failed to run data as the data amount is not "roughly same". eg. temperature row data I get 24 data but aboverground biomass row has only one data (total aboveground biomass). How I arrange data only that can run the correlation analysis?
- Badges
- Science topic