**3. Results**

We show the maps of potential for remote sensing in monitoring within watersheds for various cost-payoff scenarios in Figure 4. These potential rankings represent an interplay between the ambient cloud cover (Figure 2a) and the maximum of the shortest travel time to a population center (Figure 2b). The model predicts that when cloud cover is low, and travel times are short, the potential is maximized. Conversely, when cloud cover is high and travel times are long, the potential is inhibited. Typically, cloud cover is lowest in the Southwest, and increases generally through the Central Plains towards the East and is highest over the North and the Rockies and Appalachian mountains (Figure 2a). Travel times are generally relatively low (less than three hours) on the Eastern and Western seaboards, and throughout the Central plains and in the South, and increase to more than six hours in the mountainous Rockies and the remote Midwest. These patterns are reflected in the model.

Going from left to right in Figure 4, the expense of effort to build pipelines for occluded images and changing light conditions is budgeted increasingly generously. This reflects in the improved potential for remote sensing in more parts of the Mideast and East. Going from top to bottom in Figure 4, the effort required to collect in-situ data for groundtruthing is budgeted increasingly generously. This reflects in the improved potential for remote sensing in more parts of the Southwest and the Midwest, until in the bottom right of Figure 4, there is more or less uniformly high potential throughout the country. Nonetheless, most of the California Central Valley, Southern California, and the Central Plains have uniformly good to excellent potential under all cost-payoff scenarios, owing to generally clear skies and short travel times.

**Figure 4.** Modeled maps of the potential of remote sensing for data collection and monitoring of watersheds under various cost-payoff assumptions. As the effect of cloud cover decreases (more cloudy days are acceptable), more cloudy areas display higher potential. As the effect of access decreases (longer travel times are acceptable), more remote areas display higher potential. Thin red lines are state boundaries.

In Figure 5, we show the various risk metrics (a through e) and the associated remote sensing utility map (f) by applying the flowchart in Figure 3 to each subwatershed in these layers. Population centers, as expected, are generally concentrated near the major cities on the Eastern and Western seaboards and in the Mideast and the Gulf of Mexico coast (Figure 5a). The water demand generally tracks the population centers, except in California, the Midwest, and the Central Plains subwatersheds from where water is diverted to other places (Figure 5b). For instance, in California, water is gravity fed from reservoirs in the North to consumers in the South. Almost everywhere in the United States, ecosystems are vulnerable (Figure 5c). In most of the Midwest, this is likely due to large tracts of lands being designated as protected zones, and in the Sierra Nevada mountains in the West, due to ecological disasters by sustained drought and forest fires. Most of the population centers on the Eastern and Western seaboard are polluted by wastewater and permitted discharges, while the Central plains and Midwest impairments reflect of agricultural runoff into the Mississippi River tributaries (Figure 5d). The ATTAINS database indicates that most of the nation's subwatersheds have at least some waterbodies that have been assessed. But many subwatersheds in the United States-Mexico border region, the remote Midwest, and the Rockies remain unassessed (Figure 5e). Based on a combination of these layers in conjunction with the remote sensing potential model under the assumption of a normal cost-payoff scenario (middle panel in Figure 4 and Table 3), several hotspots emerge banded predominantly in the South and the North where remote sensing could be crucial, and often perhaps the only source of assessment of the condition of these subwatersheds (Figure 5f).

**Figure 5.** Predicted subwatersheds where the role of remote sensing could be crucial as the intersection of greater than good remote sensing potential, high impairment risk indicators, and low coverage. Across subwatersheds, (**a**) population, (**b**) total water demand, (**c**) ecosystem vulnerability index, (**d**) pollution entering streams and waterbodies, (**e**) water quality monitoring gaps, and (**f**) intersection of these layers with normal accessibility and normal cloud cover cost-payoff model predicted higher than good remote sensing potential subwatersheds (blue polygons).

It is evident from these remote sensing potential maps and the combination of risk metrics and remote sensing potential that there are specific areas within the country where watershed researchers and managers can beneficially leverage remote sensing data at various spatial scales ranging from stream segment-reach to the subwatershed scale. By allocating resources preferentially to occluded and poorly-lit scene imagery usage pipelines, the potential for using remote sensing can be tremendously maximized. This is borne out by rapidly improving potential going left to right in Figure 4, than by going from top to bottom in Figure 4.
