REMOTE SENSING

Remote Sensing

HINT: Clicking on any map on this page will open a higher resolution PDF version, which can then be zoomed, printed, or downloaded. PDF map files posted to this website are designed to reveal excellent detail when zoomed to 400%. This is particularly useful for viewing our various leafy spurge maps, including the remote sensing maps available from this page.

Introduction

Key to Remote Sensing Maps

Dinosaur to Maybell Series

Axial Basin/Little Yampa Canyon Series

Craig to Hayden Series

Publication

We are very excited to report that an article on Chloe Mattilio's Remote Sensing project has now been published in the January 2023 edition of Frontiers in Remote Sensing. A link to Remote mapping of leafy spurge (Euphorbia esula, L.) in Northwest Colorado can be found on our Resources page, or it can be downloaded directly by clicking on the title in this sentence.

Remote Intro

Introduction

Researcher Chloe Mattilio, while a graduate student at the University of Wyoming (under the direction of Dr. Dan Tekiela, invasive weed scientist with the Department of Plant Science) developed a remote sensing mapping application that adds another tool to the toolbox for the control of leafy spurge in the Yampa River Basin. Chloe's application seeks to detect the presence of leafy spurge infestations using high spatial resolution, multispectral satellite imagery.

Chloe’s work began in 2019 with searching out appropriate satellite photography (recent, little or no cloud cover, inclusive of our areas of interest in the Yampa River corridor, covering a variety of spectral bands . . . and reasonably priced!). Ultimately the purchased photography covered the length of the Yampa River from Deerlodge Park to Hayden, and extended away from the main river channel approximately 1.5 miles on either side—for a total area of roughly 174,706 acres.

Chloe then applied a number of sophisticated digital tools to further refine the pixel resolution of the photography, before processing it into multiple spectral band combinations. These were further tweaked by applying different contrast, brightness and gamma values—all in an attempt to tease out the subtle spectral differences recorded in each photographic pixel.

The end goal was to develop a set of algorithms that can effectively classify each enhanced photographic pixel as to whether it recorded light that was, or was not, reflected from a patch of leafy spurge on the ground. Chloe "trained" the classification algorithms by running them on the multispectral photography, tweaking them to enhance the results, and then running them again—multiple times. (To see a brief, but more technical description of this stage of Chloe's work, click HERE.)

Chloe's development of the remote sensing application also included correlation of the results with the Field Mapping data collected by YRLSP during 2019 and 2020. Chloe then conducted additional groundtruthing of her own during the summer of 2021, with the goal of further refining the accuracy of the application, as described in this PowerPoint presentation to the YRLSP Working Group.

The final result was a comprehensive dataset for remotely sensed leafy spurge infestations in the Yampa River corridor. The final raw dataset delivered to the YRLSP is a raster (digital image) file with a pixel resolution of approximately 2.25 square meters per pixel (or roughly 1,800 pixels per acre). Each pixel is encoded with a unique value somewhere on the spectrum between 0 and 1, representing the probability that the spectral value recorded in that location by the original satellite imagery represented light reflected from a leafy spurge canopy. A zero value represents the lowest probability of leafy spurge presence, and a one value represents the highest probability.

To facilitate interpretation of the raw data, the YRLSP then classified the entire spectrum of pixel values into four discrete categories, representing "no probability," "low probability," "moderate probability," and "high probability." For display purposes in our interpretive maps, pixels corresponding to each of these categories can then be represented by a unique color. However, the YRLSP’s maps give the “no probability” category a “no color” value, rendering it invisible—and thus limiting the actual values displayed in each map to just the "low," "moderate," and "high" categories.

The break points delimiting the range of values for each probability category were deliberately adjusted so that the resulting map display roughly parallels the field mappers’ actual on-the-ground observations of leafy spurge presence or absence. This is in many respects an arbitrary decision, and the results are not always satisfactory (more on this, below). Note that further adjustments of these break points would, of course, change the relative acreage calculated for each category.

For the three remote sensing map series published on the YRLSP website, the break points and acreage represented are as follows:

Field Maps

Finally, it must be stressed that the "Remote Sensing" maps published on this website are not intended to represent the documented presence of leafy spurge on the ground. Only field mapping can accomplish this. However, "high probability" (red) pixels do suggest that field mapping efforts might be maximized by prioritizing visits to those locations.

Inevitably, what is on the ground will not always correspond to what is suggested by the maps. In particular, some irrigated fields where leafy spurge is shown on the maps to be a "high probability" are known to have little or no leafy spurge actually present. Nevertheless, used with their limitations in mind, these maps should be a useful tool for facilitating the identification of areas of concern.

Note that the original, unclassified raster provided by Chloe, as well as the YRLSP's shapefiles of the processed, categorized data, are both available for download on our GIS Data page.

Remote Key