Today I managed to make both GRASPeR and CAPS produce sensible results (using the data I gave them). It seems that the high ground near the peak of bleaklow cannot produce much Sphagnum in the future. However, GRASPeR gives a more pessimistic prediction using data completely same as today's. I guess the reason is that I have too many "pseudo absence" data in my dataset, i.e. all sites where I did not visit or did not see Sphagnum were labelled "absence". I may need to randomly reduce the number of "absence" to a reasonable level. Besides, I would like to change the prediction dataset for the GRASPeR, see what if Peak District have more precipitation in the future? It would be interesting.
On the other hand, CAPS has even poorer documentation than GRASPeR. I need to ask the author why the replication number is limited and why the coverage drops drastically after certain years.
没有评论:
发表评论