Tue, Aug 03, 2021:On Demand
Background/Question/Methods
With ongoing anthropogenic climate change, arid grasslands may catastrophically bifurcate into deserts. Theoretical models predict that preceding this transition vegetation should organize itself into periodic patterns. We study vegetation transitions in a shortgrass steppe in southwestern Colorado. We ask: (1) Is there a periodic pattern? (2) Does it change over time? (3) What are the mechanisms that drive pattern formation? To answer these questions, we acquired historical images via Google Earth and took multispectral drone images. In the field, we used a Micasense RedEdge multispectral camera on a custom-built drone and flew at an elevation of 30 meters with 75% overlap between adjacent images. Images were processed into NDVI orthomosaics using Pix4D, which were then clipped to 90x90 meter areas and resampled to 0.5 meter resolution. We used R to analyze the images by making 2D periodograms and radial and angular spectra. In the field, we also experimentally measured the rate of infiltration at 12 bare soil and 12 grassy patches using 24.5 cm diameter plastic rings with 500 mL of water. Climate data were obtained from a National Historical Climatology Network Station located 38 km northwest of the field site at Colorado Springs Airport.
Results/Conclusions Over the last 50 years, average annual temperature increased significantly (R^2=0.194, n=72, p=0.0001) and precipitation did not change. Inspection of Google Earth images of our field site from before 2004 showed no visible vegetation periodic pattern however images after 2011 did contain a clear pattern. To explore this contrast, we have statistically analyzed a 2003 Google Earth image and our own 2020 NDVI image. Vegetation structure in 2003 was not random, and the radial spectrum detected a dominant wavelength of 90 meters. Given that the size of the image was 90 meters, it is hard to argue that this constitutes a periodic pattern. Analysis of our 2020 NDVI aerial image shows the dominant pattern wavelengths were 60, 26, 20, 16, and 12 meters, and the pattern was strongly oriented in the north-south direction, regardless of the direction of slope of the terrain. The pattern was composed of dense patches of Buffalo and Blue grama grass alternating with bare soil covered in a thin crust ( <1 cm). Infiltration experiments showed that soil infiltration rates in grassy patches was more than 7 times that for bare soil (t=3.231, df=23, p=0.0037).
Results/Conclusions Over the last 50 years, average annual temperature increased significantly (R^2=0.194, n=72, p=0.0001) and precipitation did not change. Inspection of Google Earth images of our field site from before 2004 showed no visible vegetation periodic pattern however images after 2011 did contain a clear pattern. To explore this contrast, we have statistically analyzed a 2003 Google Earth image and our own 2020 NDVI image. Vegetation structure in 2003 was not random, and the radial spectrum detected a dominant wavelength of 90 meters. Given that the size of the image was 90 meters, it is hard to argue that this constitutes a periodic pattern. Analysis of our 2020 NDVI aerial image shows the dominant pattern wavelengths were 60, 26, 20, 16, and 12 meters, and the pattern was strongly oriented in the north-south direction, regardless of the direction of slope of the terrain. The pattern was composed of dense patches of Buffalo and Blue grama grass alternating with bare soil covered in a thin crust ( <1 cm). Infiltration experiments showed that soil infiltration rates in grassy patches was more than 7 times that for bare soil (t=3.231, df=23, p=0.0037).