I'm trying to model some plant species from which I have my own data and from which I also collected data from GBIF. I've downloaded the climatic data from worldclim. I have 2 problems:
1) the MaxEnt keeps showing the message that my species are located out of the range of the climatic data
2) I also keep getting a message telling that I don't have training samples. How do I define which samples are for training and which are for analysis?
Thanks!
See the paper by Elith et al (2011), A statistical explanation of MaxEnt for ecologist, for a good explanantion and examples.
Thank you Santiago! I will check all the data from GBIF! You were very helpful!
Cheers
Thank you Luther! I will go through the paper! I sure its helpful!
Cheers
Viva Bernardo. Não estarás a modelar espécies que tenham ocorrências muito próximas do litoral? Se for o caso a máscara do WorldClim pode eliminar pequenas áreas para as quais tens ocorrências; a solução poderá passar por usar os dados existentes nos píxeis mais próximos (a ferramenta Nibble do ArcMap pode ser útil para isto).
No caso de espécies sujeitas à dinâmica das marés, estes dados também podem ser úteis: SCHWEITZER, P.N., 1993. Modern Average Global Sea-Surface Temperature. U.S. Geological Survey Digital Data Series DDS- 10. Reston, Virginia: U.S. Geological Survey. http://geo-nsdi.er.usgs.gov/metadata/digital-data/10/.
Se a área de ocorrência for apenas o território de Portugal continental, poderás experimentar os índices bioclimatológicos que desenvolvi no âmbito da minha tese (encontras aqui: http://home.isa.utl.pt/~tmh/). A qualidade dos dados de base relativos à precipitação é bastante superior à que encontras no WorldClim.
Por fim, num dos grupos de investigação em que me incluo, temos vindo a desenvolver metodologias robustas e com elevada interpretabilidade para modelar a adequabilidade ambiental para espécies. Caso tenhas interesse, podes espreitar aqui: https://www.researchgate.net/publication/261957868_Mathematical_contributions_to_link_biota_with_environment?ev=prf_pub.
Article Mathematical contributions to link biota with environment
Obrigado Tiago!! Sim são espécies de sapal daí provavelmente o problema que tu dizes. Vou ver os teus links são úteis de certeza!
Abraço
Also remember to check if you have the correct order of latitude and longitude values. For Maxent you need a .csv file where the first value (beside e.g. name of species) is longitude (E-W) and then latitude (N-S)
I think that when you have solved question 1 you will have solved question 2 as well, as I believed you get the error message that you have no training data because you response data is outside the extent of your predictor data.
I guess the problem is the datasets you have, the predictors (your plants) and the responses do not have the same projection. If you have a GIS project, then choose the projection that you usually work in and transform the data into that projection. Loom at the ArcGIS manual for setting the projection right: http://resources.arcgis.com/en/help/main/10.1/index.html#//00660000000v000000
When it comes to training and test data, Maxent will subsample from you data. But remember that this is not an independent validation. It is better than nothing, but you are testing the model with just a subsample of you data. If you have time, then it is much better to collect a dataset independently of the dataset that has been used for modelling, this is the way to fully validation your model. When using a subsample of your data for validation, the validation usually gets better that if you use a true independent dataset.
Good luck
Bernado,
If you have access to ESRI products, there is a great SDM tool box created by Jason Brown from Duke University. It takes you through a step by step process to get your data ready for Maxent simulations, including the download and processing of Worldclim data. You can download the product here: http://sdmtoolbox.org/ and read more on the processing steps involved for Maxent data preparation here: http://www.jasonleebrown.org/SDMtoolbox/current/User_Guide_SDMtoolbox_Chp5.pdf.
Good luck!
As far as I understand you can also randomly divide your species data into separate training and test files -- this is helpful if you are planning to use cross-validation etc but still want to use x% training/y% test data.
I would definitely advise using SDM Toolbox - it will make processing of environmental datasets a lot more efficient and allows projections into EAP etc in much more "user-friendly" way than directly through ARC. It is generally a life-/time-saver when it comes to anything to do with Maxent!
There is a good discussion group which answers most common questions: [email protected]
The online tutorials for Maxent are also very helpful. I would advise using the online datasets and following the tutorial instructions before using your own data.
www.academicjournals.org › app › webroot › article
Maxent modeling for predicting suitable habitat for threatened ...
PDFby S Kumar
- Badges
- Science topic
- Similar topics
- Psychology