Improving Landsat predictions of rangeland fractional cover with multitask learning and uncertainty

Authors: ALLRED, BRADY - University Of Montana; Bestelmeyer, Brandon; Boyd, Chad; BROWN, CHRISTOPHER - Google; Davies, Kirk; DUNIWAY, MICHEAL - (NCE, CECR)networks Of Centres Of Exellence Of Canada, Centres Of Excellence For Commercilization A; ELLSWORTH, LISA - Oregon State University; ERICKSON, TYLER - Google; FUHLENDORF, SAMUEL - Oklahoma State University; GRIFFITHS, TIMOTHY - Natural Resources Conservation Service (NRCS, USDA); JANSEN, VINCENT - University Of Idaho; JONES, MATTHEW - University Of Montana; KARL, JASON - University Of Idaho; KNIGHT, ANNA - Us Geological Survey (USGS); MAESTAS, JEREMY - Natural Resources Conservation Service (NRCS, USDA); MAYNARD, JONATHAN - University Of Colorado; McCord, Sarah; NAUGLE, DAVID - University Of Montana; STARNS, HEATH - West Texas A & M University; TWIDWELL, DIRAC - University Of Nebraska; UDEN, DANIEL - University Of Nebraska

Submitted to: Methods in Ecology and Evolution
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/15/2021
Publication Date: 5/20/2021
Citation: Allred, B.W., Bestelmeyer, B.T., Boyd, C.S., Brown, C., Davies, K.W., Duniway, M.C., Ellsworth, L.M., Erickson, T.A., Fuhlendorf, S.D., Griffiths, T.V., Jansen, V., Jones, M.O., Karl, J., Knight, A., Maestas, J.D., Maynard, J.J., McCord, S.E., Naugle, D.E., Starns, H.D., Twidwell, D., Uden, D.R. 2021. Improving Landsat predictions of rangeland fractional cover with multitask learning and uncertainty . Methods in Ecology and Evolution. 12(5):841-849. https://doi.org/10.1111/2041-210X.13564.
DOI: https://doi.org/10.1111/2041-210X.13564

Interpretive Summary: We used advanced computational procedures to produce a vegetation cover product that spans western United States rangelands. We trained the model with 52,012 on-the-ground vegetation plots to simultaneously predict fractional cover for annual forbs and grasses, perennial forbs and grasses, shrubs, trees, litter, and bare ground. To assist interpretation and to provide a measure of prediction confidence, we also produced spatially-explicit, pixel-level estimates of uncertainty We evaluated the model with 5,780 on-the-ground vegetation plots removed from the training data. The advancements achieved with the new rangeland fractional cover product expand the management toolbox with improved predictions of fractional cover and pixel-level uncertainty. The new product is available on the Rangeland Analysis Platform (https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Frangelands.app%2F&data=02%7C01%7C%7Cb6aedae4faa445f24aac08d83956278b%7Ced5b36e701ee4ebc867ee03cfa0d4697%7C0%7C0%7C637322389800163489&sdata=2dOw6cwqGIuBhL8TMt8FR%2BsOE5i38McfJHrInb3FMv0%3D&reserved=0), an interactive web application that tracks rangeland vegetation through time. This product is intended to be used alongside local on-the-ground data, expert knowledge, land use history, scientific literature, and other sources of information when making interpretations. When being used to inform decision-making, remotely sensed products should be evaluated and utilized according to the context of the decision and not be used in isolation.

Technical Abstract: 1. Operational satellite remote sensing products are transforming rangeland management and science. Advancements in computation, data storage, and processing have removed barriers that previously blocked or hindered the development and use of remote sensing products. When combined with local data and knowledge, remote sensing products can inform decision making at multiple scales. 2. We used temporal convolutional networks to produce a fractional cover product that spans western United States rangelands. We trained the model with 52,012 on-the-ground vegetation plots to simultaneously predict fractional cover for annual forbs and grasses, perennial forbs and grasses, shrubs, trees, litter, and bare ground. To assist interpretation and to provide a measure of prediction confidence, we also produced spatially-explicit, pixel-level estimates of uncertainty. We evaluated the model with 5,780 on-the-ground vegetation plots removed from the training data. 3. Model evaluation averaged 6.3% mean absolute error and 9.6% root mean squared error. Model performance increased across all functional groups compared to the previously produced fractional product. 4. The advancements achieved with the new rangeland fractional cover product expand the management toolbox with improved predictions of fractional cover and pixel-level uncertainty. The new product is available on the Rangeland Analysis Platform (https://rangelands.app/), an interactive web application that tracks rangeland vegetation through time. This product is intended to be used alongside local on-the-ground data, expert knowledge, land use history, scientific literature, and other sources of information when making interpretations. When being used to inform decision-making, remotely sensed products should be evaluated and utilized according to the context of the decision and not be used in isolation.