h13+v13

The h13 v13 box contains a large portion of the southern tip of South America. This area is more commonly known as Patagonia, and consists of the two countries Argentina and Chile. For this project, I will focus on the provinces of Santa Cruz and Chubut (see map below). According to the Global Vegetation Trends 2000-2008 image, this area has been experiencing a decrease in vegetation.

Hypothesis:

I think the reported decrease in vegetation is associated with climate change. Climate change can be seen in Patagonia by the rapidly melting glaciers. Changes in temperature and precipitation lead to environmental challenges such as water shortages, forest fires, and disease outbreaks (Resilience to Climate Change in Patagonia, Argentina), all of which can contribute to the destruction or reduction of vegetation growth. I think it is also important to consider the consistency of the images used to discover the decreased vegetation. When and how often were images compared to compile the Global Vegetaion Trends 2000-2008 image? If images were compiled once a year at the same time, then the results may be misleading as they do not take into account the changes in time and duration of the seasons.



Assignment: Week 4 ||
 * [[image:spring_ndvi_final.jpg]] || [[image:summer_ndvi_final.jpg]] ||
 * **//Spring (May 25, 2010)//** || **//Summer (August 5, 2010)//** ||
 * [[image:Fall_NDVI.jpg]] ||= **H13 V13 2010 NDVI Imagery**
 * //**Fall (November 1, 2010)**// ||  ||

Since there are a lot of shrublands and grasslands within my area, I chose to use NDVI. This will allow me to see where and when there are areas of decreasing and/or increasing greenness. It will also allow me to see areas of increasing greenness due to melting glaciers along the western coast.

Assignment: Weeks 5 & 6


 * There are 15 bands per year.
 * Occurs every 16 days.
 * There are 23 observations per year.
 * There are 513 columns and 202 lines for the area time series.
 * There are 309 columns and 121 lines for the Anthromes files.


 * Class Study ||
 * = Class # ||= Clasification ||= Number of Pixels in Area ||= % Pixels in Area ||
 * = 11 ||= Urban ||= 2 ||= 0.005 ||
 * = 12 ||= Dense ||= 5 ||= 0.013 ||
 * = 21 ||= Rice Villages ||= 0 ||= 0 ||
 * = 22 ||= Irrigated Villages ||= 0 ||= 0 ||
 * = 23 ||= Cropped Pastoral Villages ||= 10 ||= 0.027 ||
 * = 24 ||= Pastoral Villages ||= 4 ||= 0.011 ||
 * = 25 ||= Rainfed Villages ||= 0 ||= 0 ||
 * = 26 ||= Rainfed Mosaic Villages ||= 15 ||= 0.040 ||
 * = 31 ||= Residential Irrigated Cropland ||= 7 ||= 0.019 ||
 * = 32 ||= Residential Raindfed Mosaic ||= 230 ||= 0.617 ||
 * = 33 ||= Populated Irrigated Cropland ||= 0 ||= 0.021 ||
 * = 34 ||= Populated Rainfed Cropland ||= 84 ||= 0.225 ||
 * = 35 ||= Remote Croplands ||= 20 ||= 0.054 ||
 * = 41 ||= Residential Rangelands ||= 111 ||= 0.298 ||
 * = 42 ||= Populated Rangelands ||= 376 ||= 1.009 ||
 * = 43 ||= Remote Rangelands ||= 8,545 ||= 22.929 ||
 * = 51 ||= Populated Forests ||= 460 ||= 1.234 ||
 * = 52 ||= Remote Forests ||= 2,250 ||= 6.037 ||
 * = 61 ||= Wild Forests ||= 72 ||= 0.193 ||
 * = 62 ||= Sparse Trees ||= 40 ||= 0.107 ||
 * = 63 ||= Barren ||= 175 ||= 0.470 ||

(Above colors: Remote forests is yellow, populated forests is blue, remote rangelands is dark brown, populated rangelands is lighter brown, and residential rainfed mosaic is purple.)


 * = Dominant 5 Classes ||
 * = Ranking ||= Classification ||
 * = 1 ||= Remote Rangelands ||
 * = 2 ||= Remote Forests ||
 * = 3 ||= Populated Forests ||
 * = 4 ||= Populated Rangelands ||
 * = 5 ||= Residential Rainfed Mosaic ||

The five dominant land classes represent 722,361,841,692.1719 squared meters or 31.826% of the area.






 * Remote rangelands and populated rangelands are overall less green than the remote forests, populated forests, and residential rainfed mosaic.
 * The greenness of the five dominant land classes appears to be decreasing over time.
 * While the seasons are not very consistent, I think the seasons can be identified for each classification except the residential rainfed mosaic.
 * Remote rangelands and populated rangelands are at their peak greenness when the others are at their minimum greenness.
 * Why? Need to research vegetation.
 * While humans indirectly contribute to climate change which can lead to vegetation change, I do not think less populated areas would directly change my results as the two largest classifications of this area are remote rangelands and remote forests.



As stated earlier, I believe the NDVI shifts in Patagonia are largely due to climate change. So far, my studies have supported this theory.

Let's first look at the west coast which consists of glaciers and forests. The melting glaciers may be contributing to the greenness of the forests which in turn affects the NDVI score. This lab shows us there are areas of high and low variance along the west coast. The low values are probably associated with the forests and the high values are associated with the melting glaciers. I would also like to look at an area along the east coast that I am still unsure about. Notice the area of purple on the January 1 image. This area consists of rangelands and appears to have a large temperature variance during several months of the year. It also has a low, and still decreasing NDVI score. After talking about my area with Bill, I think this area may be experiencing a drought. Possible research for this area may be looking at burned areas and/or soil moisture.

Choosing Landsat Images:




 * [[image:Image2_Landsat_bds.png caption="March 29, 2001"]] || [[image:Image3_Profile_bds.png width="312" height="224" caption="March 29, 2001"]] ||
 * [[image:Image1_Landsat_bds.png caption="April 1, 2002"]] || [[image:Landsat5_Image_bds.png width="313" height="224" caption="April 1, 2002"]] ||

Lab 9 Image Date: March 29, 2001

Based on some discussion in class, I think my image is a volcano.

For the first part of the lab, application of a smoothing technique, I tried applying a medium filter which tends to be good for reducing noise. I first ran the image at 3 kernels and then I ran it at 43 kernels. The 3 kernel image is a sharper image, but in my own personal opinion, the 43 kernel image seems to illustrate more depth. For the edge detection, I chose to use the sobel edge. I first tried running the sobel edge with 0% of the original image, which turned out dark and slightly hard to interpret. I than tried running the sobel edge with 50% of the original image. I personally think the results from this run turned out much better as there is more contrast between the higher and lower ground, allowing the edges to display more clearly.

Although PC 1 is the average and appears to have the cleanest image, PC 2 and PC 3 seem to really highlight features that are not so obvious in the PC 1 image. Notice the small black dots in the PC 2 image. The majority of these dots (which I think are little lakes or craters), are covered by whiteness (possibly fog or steam) in the original image. Also take notice to the streams (or rivers) that stand out well in the PC 3 image. Due to the texture of the original image, it is hard to identify those features.
 * [[image:PC1_March29_bds.png width="480" height="480" caption="PC 1"]] || [[image:PC2_March29_bds.png width="480" height="480" caption="PC 2"]] || [[image:PC3_March29_bds.png width="480" height="480" caption="PC 3"]] ||
 * [[image:PC4_March_29_bds.png width="480" height="480" caption="PC 4"]] || [[image:PC5_March29_bds.png width="480" height="480" caption="PC 5"]] || [[image:PC6_March29_Bds.png width="480" height="480" caption="PC 6"]] ||





Based on the eigenvalue chart, the ENVI stats file, and my calculations from Excel, it is easy to see that band one has the most variability (85.37%). Bands two and three have the second and third most variability with percentages less than 10% each, while bands four, five, and six have the least about of variability with percentages less than 1%. At first glance, my original image and my inverse PC image look the same, but if you look closesly, there are a few things to take note of:
 * Basic Stats || Min || Max || Mean || Stdev || Num || Eigenvalue || % Variability ||
 * Band 1 || -1127 || 6806 || 152.36 || 231.6459 || 1 || 564840.2139 || 0.853717908 ||
 * Band 2 || -714 || 6915 || 612.2 || 292.0734 || 2 || 64246.70262 || 0.09710456 ||
 * Band 3 || -590 || 8136 || 1004.1 || 345.4539 || 3 || 21836.93289 || 0.033005052 ||
 * Band 4 || -416 || 6270 || 1541 || 402.3188 || 4 || 5225.608325 || 0.007898155 ||
 * Band 5 || -169 || 6248 || 1833.1 || 377.3833 || 5 || 3293.300843 || 0.004977602 ||
 * Band 6 || -85 || 5232 || 1503.3 || 314.7068 || 6 || 2181.190979 || 0.003296723 ||
 * [[image:March29_Compare_bds.png width="640" height="640" caption="Original"]] || [[image:March29_InversePC_bds.png width="640" height="640" caption="Inverse PC"]] ||
 * 1) The brown spots in southern central portion of the inverse PC image.
 * 2) The shift from green to blue lakes (or craters). I find this interesting as only some of them changed.
 * 3) Although there is still a lot of whiteness in the inverse PC image, it seems more centralized, allowing ground features to be more defined.

Two Color Multiview NDVI