Geog477 – Lab 6 Due: November 21, 2007
Lab 6: Term Project - Part II
Analysis – Report Writing
Drought Impacts in the Mid-Atlantic del Sur (Virginia, North Carolina, South Carolina)
A time-series of NDVI observations will be created in order to examine the dynamics of the growing season and to monitor the drought-impacts on primary production. The main objective of this lab is to create a set of cumulative NDVI images for our study area and calculate the region’s average cumulative ‘greenness.’ The goal is to identify those years which differ substantially from the average year for our study region.
You should also revise your introduction based on our comments and write up your methods section and bullet some preliminary results. The next lab will be our final lab for this project.
Vegetation Indices Background: The Normalized Difference Vegetation Index (NDVI)
There are many resources to learn more about NDVI. The manual for the MOD13 vegetation indices algorithm is probably the most comprehensive source. It is located at
\Data\termProject\help\MOD13_Veg_Indices_Algorithm.pdf
The user’s guide is maintained by the University of Arizona and can be found at:
http://tbrs.arizona.edu/project/MODIS/UserGuide_doc.php
The simplest explanation for NDVI is shown below as the reflectance in the NIR minus the reflectance in the red, normalized by the sum of the two;
The following text is taken directly from the first source and should give you some background on some of biological processes that are commonly linked to NDVI:
“Many studies have shown the NDVI to be related to leaf area index (LAI), green biomass, percent green cover, and fraction of absorbed photosynthetically active radiation (fAPAR) (Asrar et al., 1984; Baret and Guyot, 1991; Goward and Huemmrich, 1992; Sellers, 1985; Sellers, 1986; Running and Nemani, 1988; Tucker et al., 1981; Curran, 1980)…Other studies have shown the NDVI to be related to carbon-fixation, canopy resistance, and potential evapotranspiration allowing its use as input to models of biogeochemical cycles (Raich and Schlesinger, 1992; Fung et al.,1987; Sellers, 1985; Asrar et al., 1984; Running et al., 1989; Running, 1990; IGBP,1992).”
If you are interested in pursuing any of the above topics in more detail for this project you are encouraged to track down one of the referenced papers above.
Data
For this lab you will be using the MOD13A2 (Vegetation Indices - NDVI, 1km, 16 days) product. This data was imported and reprojected (using Modis Reprojection Tool (MRT - see http://edcdaac.usgs.gov/landdaac/tools/modis/index.asp to download) to a geographic coordinate system using the North American Datum of 1983 – GCS NAD83. The imported images were then clipped to the North Carolina, South Carolina, and Virginia region using Extract by Mask (batch) tool in ArcGIS ArcToolbox.
The total NDVI record is from 2000-2007. We will focus our analysis on Julian day 81 (March 20th) through 271(Sept. 28th) since this roughly represents the growing season (but realize that in North Carolina alone the growing seasons is up to 100 days longer on the coast than in the mountains). For our analysis we will be working with 13 NDVI images, each representing a 16 day period.
The naming convention for the images is as follows:
0081.c1_km_16_days_NDVI.tif
The first digit refers to the year. The next three digits refer to the Julian day. The above image is for day 81 year 2000. For 2001 the first four digits would be 1081. The c references that the image is clipped, 1_km_16_days_NDVI is self-explanatory.
The valid range for MOD12A2 values is from -2000 to 10,000. In order to scale the data to range from -1 to 1 you must divide the entire scene by 10,000. Values filled with -3000 typically represent water. Refer to the image above for a visual interpretation of the values. Vegetated areas will typically have values greater than zero and negative values indicate non-vegetated surface features such as water, barren, ice, snow, or clouds.
Analysis Methods
In order to automate this process as much as possible the Modeler within Imagine will be used. To accumulate NDVI through the end of the summer we will simply add our first scene – day 081 to our second scene – day 097. Our next step will be to add this composite to the third scene – day 113 and so on. The intermediate steps will be used in later analysis when we plot the accumulation of NDVI through the year and also to calculate an average seasonal NDVI. This will entail summing the accumulated NDVI images for each time period for the entire 8 year record.
For the year 2000 this cumulative model has already been written and run. You should use this as a starting point for the other 7 years. Navigate to \Data\termProject\model\ and copy CumulativeNDVI_2000.gmd into your student directory. In Imagine, open the Modeler | Model Maker. Click Open and navigate to the model you just copied.
This model maker is fairly straightforward. Each represents an NDVI image input or output. The are the functions linking the images. The is used to connect each step. Clicking the hammer will open your toolbar. Since the model is already built you won’t need to use these tools much in this first step. Clicking will give you an overview of the entire model. The text will disappear in this setting. Click on to zoom back to a level that allows the text to be seen. Clicking on the black arrow will allow you to move and rename objects in the model.
What you will need to do is change the INPUT RASTER, OUTPUT RASTER for every step. The adding operation linking the data sets will stay the same. Make sure that you declare the input rasters as FLOAT and set the Data Type in the output as Float Single, not Signed 16-bit! This is very important! To change the input variables simply double click on the boxes. Your life will be made much easier if you stick to a standard naming convention for the output. Use my example in the model or make up your own. JUST BE CONSISTENT as it will get confusing quickly if you are not.
Create 7 different models for the years 2001 through 2007 and save each one separately. This might be a good time to divide and conquer with your classmates so that you don’t have to create all 7 models. Since there are 16 of you working on this project it would be good to form groups of 4-5 and divide this part of the analysis between you. This isn’t mandatory but it will save you a decent amount of time. Each of you will have to create the model and then share the resulting datasets in your group.
The final product of these 8 models will be a set of images accumulating NDVI through the growing season (81, 81+ 97, etc.). Your next step will be to create a set of average cumulative NDVI images for the entire study period. You can do this a number of ways. I recommend building a new model from scratch and simply summing all of the final cumulative images together for each compositing period and then dividing by 8. Something like this:
(00_comp_81_273.tif + 01_comp_81_273.tif +…07_comp_81_273.tif) / 8 = 01_07_comp_81_273.tif
The next step will be to create the seasonal averages. This will entail summing the NDVI for year 2000 day 81, with NDVI year 2001 day 81, etc. and dividing by 8. Then summing the composite 081_097 for each year etc….
I recommend another set of models for this step. This means another 13 models…I know this sounds horrific but once you figure out the first one they should go smoothly after that. This is also why I STRONGLY recommend getting in groups.
Finally – you should make a difference map of the cumulative NDVI for years 2002 and 2007 from the average total cumulative NDVI (2000-2007). Include two images showing your results and offer an interpretation of what you see.
You will be graded on the completeness of your models and this final difference map. Please include in your report whom you teamed up with.