I want to do a time series analysis of NDVI using MODIS dataset. Now if I consider a time span of 12 years say, 2002-2014 with MOD13Q1 (250 m spatial resolution and 16 days temporal resolution), I will be having approximately 288 images.
How to handle these many images, and how to present them in results ?
A single image contains 11 subdatasets should I add them in GIS as RGB image or process them individually. While adding the whole of this as a RGB image I am getting pixel values as 2987, 2349,5756, 6345.... Do these values correspond to the number after decimal i.e, 0.2987, 0.2349, 0.5756, 0.6345...
No do not add them as an RGB as your data will be usel;ess The values are -2000 to 10000 and then you need to scale them by 0.0001 to compare to other comparable NDVI measures. The 11 subdatasets contain various measures of quality and other information. You only need the first one. See attached it explains them all. ArcGIS can handle it in the cell statistics tool.
http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009z0000007q000000.htm
No do not add them as an RGB as your data will be usel;ess The values are -2000 to 10000 and then you need to scale them by 0.0001 to compare to other comparable NDVI measures. The 11 subdatasets contain various measures of quality and other information. You only need the first one. See attached it explains them all. ArcGIS can handle it in the cell statistics tool.
http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009z0000007q000000.htm
I would recommend using the MODIS or MODISTools package in R for downloading, clipping, stacking and reprojekting your data. Handling such data amounts in ArcGis can be exhausting.
http://r-gis.net/?q=ModisDownload
You can use QGis.
You have to apply the scaling coefficient of 0.0001, to derive the NDVI values. Layer stack using ERDAS, subset the images in case you are considering a region, convert pixels to ascii. Now that you have the ascii file, you can import it to MATLAB and do the analysis.
Use Idrisi's Earth Trends Modeler (http://www.clarklabs.org/products/Earth-Trends-Modeler.cfm)
You can use ENVI for scalling values, stacking images and do the analisys. Considering a large number of images, first stacking ('basic tools'/'layer stacking') and then proceed with the scalling (by 0.0001). Don´t forget to reorder files according to datas.
I have used ArcPy module (Checked out SpatialExtensionTool) in python for basic raster calculator and looped through the files in the folders for a similar application. os module can help with looping through files.
I would store your grids outside of an ArcGIS Geodatabase and compile them to a Mosaic Dataset using ArcCatalog. By doing so, you can manipulate them and derive outputs in ArcMAP without interfering in the "original" dataset while achieving high performance. Also, you can assess easily the change in time of overlying grids: http://resources.arcgis.com/en/help/main/10.1/index.html#/Time_in_a_mosaic_dataset/009t000001vs000000
After you scale your NDVI by 0.0001, you can use TIMESAT, which is a software package to analyse time-series. It's easy to use in case you are not familiar with R. it allows you to smooth your data (Savitzky-Golay filter, or least-squares fitted assymetric Gaussian or double logistic smooth functions) and to generate seasonality parameters.
hi Deepak,
We have used successfully the TIMESAT package described by Jönsson and Eklundh (2004). The Time Series Product Tool software package developed at the National Aeronautics and Space Administration (NASA) John C. Stennis Space Center uses TIMESAT plus some adaptions for vegetation monitoring by McKellip et al. (2005, 2008) and Spruce et al. (2011). I found a copy of our paper ( Ramsey III, E., J. Spruce, A. Rangoonwala, Y. Suzuoki, J. Smoot, J. Gasser, and Bannister, T., 2011. Daily MODIS trends of Hurricane-induced forest impact and early recovery. Photogrammetric Engineering & Remote Sensing, 77(11) 1133-1143) at "http://onlinedigitalpublishing.com/publication/?i=86396&p=67" that you can access that contains a illustration of using TIMESAT for monitoring of forest damage and recovery. If you can't get to the article from that url, contact me.
http://ibis.colostate.edu/WebContent/WS/ColoradoView/TutorialsDownloads/CO_RS_Tutorial10.pdf
You might want to check out this: http://www.clarklabs.org/products/Land-Change-Modeler-Overview.cfm
Also:
Neeti, N. and Eastman, J.R., (2011) "A Contextual Mann-Kendall approach for the assessment of trend significance in image time series", Transactions in GIS, 15, 5, 599-611.
Hi everyone, I have a question about TIMESAT:
How many are the free parameters in Savitzky-Golay filter?
The e-sensing project is developing a Web Time Series Service WTSS. This service is specifically designed to provide access to time series of imagery. A request to this service must include a latitude & longitude, a time interval, a imagery id (e.g MOD13Q1), and the requested bands. Besides the server component, we have developed clients for C++, JavaScript, python, and R. The code is available at the project's github page, there you will also find documentation and some examples on how to retrieve data from a test running instance with MODIS data of the amazon forest.
https://github.com/e-sensing/
I hope this helps.
Hi all, is it possible to export TIMESAT seasonal metrics to raster format?
Thanks,
TIMESAT seasonal metrics are raster format. You can visualize in TIMESAT Imageview. Besides, you can export to other softwares (envi, qgis, idrisi, etc.)
Hello
Thanks for all,
Dear Bruno Lara thanks, can you send me TIMESAT?
best regards
Sajad
Dear Sajad,
Here: http://web.nateko.lu.se/timesat/timesat.asp?cat=4 you can download TIMESAT's version according your requirements.
Best regards
Bruno.
You can perform maximum value composite (MVC) technique to get a single data from series of multi -temporal data.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing