*2.2. Quantifying Benefits of Remote Sensing in Watershed Monitoring*

We modeled the potential for remote sensing using Equation (1) and surrogates for accessibility and acquisition. However, these maps in themselves provide only one piece of the puzzle. It is only when the potential for watershed monitoring using remote sensing is combined with risk maps and lack of coverage by conventional monitoring that areas can be identified where remote sensing can play a crucial role. To do this, we developed maps of five metrics of risk using published GIS data from the EnviroAtlas [32] (Table 2): (i) the human footprint, (ii) anthropogenic water demand, (iii) ecosystem vulnerability, (iv) impairment, and (v) conventional monitoring and assessment coverage gaps. As the units of these metrics are different from one another, we simply overlaid these maps and determined the intersecting areas to determine where the remote sensing role would be most useful.

*Human footprint*: We collected the human population in each subwatershed. The conventional wisdom is that impairment and watershed management matter most where human settlements are prevalent. The subwatersheds populations were then ranked from 0 to 6 with decadal increase from less than 10 to more than one million people.

*Anthropogenic water demand*: We combined the domestic, agricultural, industrial and thermoelectric water demand in each subwatershed to estimate the total water demand from each subwatershed. Watersheds with greater demand are likely to experience larger ecological deficit flows and a higher propensity for critical conditions of impairment. The subwatershed demands were then ranked from 0 to 6 with decadal increase from less than 12,000 Gallons Per Day (GPD) [the per capital daily water demand is 1200 GPD estimated by dividing the total water demand by the total population].

*Ecosystem vulnerability*: We combined three key ecosystem services with equal weightage to each to estimate an overall vulnerability index. First, we summed the average number of days in a year that hunting, fishing, and recreational activities are typically engaged in within a subwatershed to obtain a fractional recreation time, *f*Recreation. Large values of this number indicate greater pressure for recreational activities. Second, we combined the total fraction of land cover earmarked for conservation by both the International Union for Nature Conservation (IUCN) and the United States Government, *f*Conservation. Large values of this number indicate that additional monitoring, assessment, and restoration effort must be expended in these subwatersheds. Third, we collected the native aquatic species vulnerability index, *f*Vulnerability, a key measure of how many endangered, native

species are threatened [33]. This index is also a proxy for ecosystem health in general. Higher values of this index imply generally poorer ecosystem health. All these numbers range from 0 to 1. The overall ecosystem vulnerability of the subwatershed is then

$$f = w \left( f\_{\text{Receration}} \mathbb{I}\_{\text{Receration}} + f\_{\text{Conversion}} \mathbb{I}\_{\text{Conversion}} + f\_{\text{Valnerability}} \mathbb{I}\_{\text{Valnerability}} \right) \tag{2}$$

where I*<sup>i</sup>* is an indicator that can either be 0 or 1 depending on whether the service *<sup>i</sup>*, i.e., recreation, conservation or species vulnerability, is being provided by the subwatershed or not, and *w* is 1, 0.5 or 0.33 if there are one, two or all three of the services being provided by the subwatershed. The subwatersheds ecosystem vulnerability indicator was then ranked 0 for no vulnerability, and then from 1 to 4 in increments of 0.25.

*Impairment risk*: We developed subindex curves of the total wastewater discharge in Million Gallons per Day (MGD), the total permitted pollutant load in pounds per year, and the total agricultural overland, tile and non-tile subsurface runoff in mm as described in Walsh and Wheeler [34]. We then combined these subindices using a maximum subindex measure as the overall index of impairment within a subwatershed. This index was deemed to represent the impairment risk in the subwatershed. To develop these subindex curves the following approach was adopted: first, we summed the total load in each class, i.e., wastewater, permitted sources, and agricultural runoff, and divided it by the total surface area of waterbodies within the watershed to normalize loads across subwatersheds. Then we omitted zero values and obtained the 25th, 50th and 75th quantiles of these normalized loads across all the subwatersheds in the country and assigned rank values from 0 to 4 depending on whether there was no impairment from that loading class to whatever quantile range the normalized loads fell into. This effectively positioned the normalized loads in each load class onto a subindex curve. Then, as the nature of impairment is likely to vary depending on the dominant economic sector, human activity, and land surface processes in each subwatershed, we took the overall impairment risk to be the maximum of the three rank values. Thus, higher values of this index (ranging from 0 to 4) represents higher risk of impairment.

*Conventional monitoring and assessment coverage gaps*: For each subwatershed, we estimated the total length of assessed stream segments, *L*Stream, the total area of assessed ponds, lakes, and estuaries, *A*Body, and the total number of monitoring stations, *N*Station, from the Assessment and Total Maximum Daily Load Tracking and Implementation System (ATTAINS) database [4]. A subwatershed was determined to have a monitoring gap if *L*Stream + *A*Body + *N*Station = 0.

Finally, we also produced a remote sensing monitoring potential map using the normal cost-payoff estimates from Table 3 to indicate where remote sensing would be likely feasible. Subwatersheds where the role of remote sensing is likely to be most crucial were then classified as those that met the decision tree in Figure 3, that is, where all the layers after suitable thresholding to represent various risks intersect with at least good potential for remote sensing.

**Figure 3.** Flowchart for estimating where the role of remote sensing will be crucial in watershed management. The operation represents the intersection of geospatial layers. The numbers indicate threshold values of various index levels.
