Matthew Tippett-Vannini

1. Overall research question:

How are the spatial and temporal patterns of land use change from grass seed to hazelnut, with and without vegetation suppression, related to infiltration rate via ecologically developed soil structure?

Questions from each exercise:

Ex. 1) What is the spatial and temporal (chrono sequence) distribution of the infiltration rates in fields transitioning from grass seed to hazelnut orchards?

Ex. 2) Of fields which have transitioned, can I identify and predict which fields have had vegetation suppression using imagery?

Ex. 3)

i. Can I detect vegetation suppression of intercropped rows on a higher spatial scale (i.e. orchard level)?

ii. Can I detect vegetation suppression of intercropped rows on a wider temporal scale (i.e. 2000-2020 level)?

What is meant by vegetation suppression?

Figure 1) a conceptual diagram showing full vegetation suppression (left) and partial vegetation (right). Note that the tree row always has vegetation suppression.

2. Data Set:

I used the infiltration data I gathered from August/September 2021 from 3 farms in the Willamette Valley (the northern extent near Albany and the southern extent near Harrisburg.

For the imagery data I used first the National Agricultural Imagery Program (NAIP) of Oregon from 2009 and 2016, again over the same spatial extent. Finally, I used Landsat data from 2000 and 2021.

3. Hypothesis:

For exercise 1, I expected that infiltration rates would increase by since transition.

For exercise 2, I expected that I would be able to find differences in the change in Normalized Difference Vegetation Index (NDVI) from 2009 to 2016. I expected NDVI to decrease (2009 subtracted by 2016) for areas with increasing greenness and to go up for areas with decreasing greenness.

For excercise 3, I expected that I would be able to find a similar trend at a larger spatial scale and longer temporal period.

• Analysis:

Ex 1) I first simply geolocated the values of the infiltration and then ran an autocorrelation function on those infiltration rate values.

Ex 2) I first learned how to calculate NDVI from NAIP data.

I then did a logistic regression of the NDVI values along a point and an autocorrelation of those points.

Ex 3) I did a logistic regression of the NDVI values within different polygons covering the hazelnut orchards.

5. Results:

Ex.1) I created a map which I hope to use for my first poster this month.

Figure 2 Showing the locations of the infiltration measurements and their values.

Figure 3 An example of the early boxplot I made for each of my sites. Later, we came to the conclusion that for the small sample size a box plot isn’t the best.

Figure 4 A more recent plot of my data, with log infiltration rate vs. years since transition for values with vegetation suppression. Note the very slight negative trend in the data.

Figure 5 A more recent plot of my data, with log infiltration rate vs. years since transition for values without vegetation suppression. Note the positive trend in the data, but the lack of replicates.

Ex. 2) I created a NDVI map and a statistical relationship between vegetation suppression on a small scale and the change in NDVI.

Figure 6 Upper showing an example of the change in NDVI from 2009 to 2016 in an area with vegetation suppression.  Lower showing the distribution of those values along the points, note the positive values.

Figure 7 Upper showing an example of the change in NDVI from 2009 to 2016 in an area without vegetation suppression.  Lower showing the distribution of those values along the points, note the negative values.

Figure 8 Autocorrelation of NDVI points along the line of the transect in the orchard.

Ex 3) I found the relationship between vegetation suppression on a larger scale and change in NDVI and created a map of the NDVI over a larger time period with a different imagery (Landsat). I was not able to detect vegetation suppression with the Landsat imagery however.

Figure 9 NDVI with a higher spatial scale (polygons representing orchards) . I used Zonal Statistics to calculate the mean NDVI for each polygon.

Figure 10 Regression analysis for the orchard scale change in NDVI

Figure 11 Change in NDVI on a longer temporal scale 2000-2021. Unfortunately, I wasn’t able to detect the change NDVI from vegetation suppression.

6. Analysis Learning:

Ex 1) I learned that my infiltration values were not spatially autocorrelated. I also learned that there were big differences in infiltration values with and without vegetation suppression.

Ex 2) I learned that the NDVI values were spatially autocorrelated and that there was a relationship between vegetation suppression and change in NDVI.

Ex 3) I learned that I could predict the NDVI values on the orchard scale as well as on the point scale. I had a harder time doing this on the

7. Significance:
   1. Scientifically, this project can provide an indication of how soil structure forms following a disturbance (i.e. tilling) and seems to show the importance that vegetation plays in the development of the that structure (fields with vegetation suppression do not seem to develop this structure).
   2. Practically, while the role of soil structure is still being understood, there are strong indications that it is important to soil carbon cycling, soil health, soil hydraulic properties, and the general regional hydrology. Increasing (or decreasing) infiltration can have a large impact on the water availability for a land.

8.  Software learning:

ArcPro- I learned how to make a layout and make a map I was proud of. I learned how to work with imagery and NDVI analysis. I learned how to do raster calculations.

R- I learned how to do an autocorrelation function and a regression analysis. I also just felt more comfortable working with my own data in R.

9. Statistics learning:

I learned about neighborhood analysis, spatial autocorrelations, and logistic regressions.

I also learned how to do a power analysis which will be helpful for resampling.

10. Evolving question.

The results first showed me that there was a large difference between fields with vegetation suppression and those without. This led me to use imagery to see if I could not only find the transition from grass seed to hazelnut, but also the farms suppressing vegetation and those not.

Original: How is the spatial pattern of infiltration capacity related to the land use change in the Willamette Valley via ecological accommodation?

Current: How are the spatial and temporal patterns of land use change from grass seed to hazelnut, with and without vegetation suppression, related to infiltration rate via ecologically developed soil structure?

11. Future Goals: I would like to do some neighborhood analyses to see if there was clustering of transition sites. I would also like to apply some of what I learned to predict possible field sites here in Oregon, as well as in Chile with the transition from wheat to forestry.