APPLICATION OF HYDROLOGIC MODELS TO ASSESS THE EFFECTS OF CLOUD SEEDING ON AGRICULTURE IN THE WALKER RIVER BASIN OF NEVADA

A set of existing hydrologic models of the headwater and agricultural areas of the Walker River basin are used to estimate the effects of cloud seeding activities by the Desert Research Institute on the amount of water delivered to the agricultural areas in the lower part of the basin over the water years 2004 through 2013. A monthly water balance model is used to simulate the accumulation and melt of the snowpack and the associated runoff in the headwater areas. A river basin management model is used to simulate the spatial and temporal complexity of the movement and use of surface and groundwater below the mountain front in the streams, reservoirs, irrigation ditches, and crop areas which are completely driven by water right priorities and water supply. Two case studies are presented, and the effects of cloud seeding on two different target areas are identified and compared in terms of: (1) changes to the amount of surface water delivered to the crop areas for irrigation purposes; (2) changes in the amount of supplemental groundwater pumping; and (3) changes in crop shortage for each year over the study period. Estimates of crop irrigation water requirements and crop value per acre were used to assess the financial benefits of the cloud seeding activities under a range of seeding efficiency estimates for each case. The results indicate that, over the study period, the benefits to seeding the target areas in the first case, characterized by relatively large long-term annual precipitation, are significantly greater than seeding the target area in the second case, characterized by less annual precipitation. The results from both studies provide useful information for planning and operational decisions related to cloud seeding in the Walker River basin.


INTRODUCTION AND SCOPE
A cooperative cloud seeding research effort between the Desert Research Institute and the U.S. Bureau of Reclamation (e.g, Hunter et al., 2005), is one of more than a dozen wintertime cloud seeding programs in the western U.S. whose primary goal is to increase snowfall over specific drainage basins in order to subsequently increase stream runoff in the spring and summer months.An accurate assessment of the impacts of any cloud seeding operation on streamflow requires detailed knowledge of the spatial and temporal increases in precipitation due to the cloud seeding activities as well as the watershed response to the additional precipitation.
Hydrologic modeling has been previously applied (e.g,.Boyle et al., 2006) to predict how changes in the snowpack from cloud seeding will alter runoff in the affected streams of targeted basins.Initial studies (e.g,.Boyle et al., 2006) for the Water Year (WY), defined as the period between October 1 st of one year and September 30 th of the next, 2004(i.e., 1 October 2003through 30 September 2004) focused on the use of a hydrologic model to simulate the hydrologic response of the target areas to estimated additional precipitation derived from cloud seeding activities.The model accomplished this through simulating the changes in hydrologic processes (snowpack evolution, evaporation and transpiration, infiltration, soil moisture movement, runoff, and streamflow) that be attributed to the seeding activities.Results indicated that additional precipitation from seeding activities resulted in increases in runoff and evapotranspiration from the target areas and that the fraction of the additional precipitation that resulted in streamflow varied from 49% to 89% among the different target areas.
In this study, a set of existing hydrologic models of the Walker River basin is used to estimate the effects of wintertime cloud seeding occurring within the basin headwater areas on the amount of water delivered to the agricultural areas in the lower part of the basin during WYs 2004-2013.The headwater region was targeted during the study period with different combinations of ground seeding generators and aircraft seeding operations.Hydrologic models calibrated in previous studies (Boyle et al., 2013;Barth, 2013), were used for a baseline "cloud seeding" simulation and two case study simulations during this study period.
For the water years of interest it was assumed that the effects of cloud seeding are incorporated in the precipitation that is measured in the Walker Basin target areas, so the effects of seeding were estimated by reducing observed amounts.In the first case study, the models were used to simulate the effects of reducing the precipitation over two target areas in the headwaters by 5%, 10%, and 15% to represent a range of possible impacts by the ground and/or aircraft cloud seeding activities.In the second case study, the same reductions in precipitation were investigated over a different target area in the headwaters to investigate the influence of average annual precipitation on the effectiveness of the seeding efforts.Changes in the amount of surface water delivered to the crop areas for irrigation purposes, supplemental groundwater pumping, and crop shortage for each year over the study period were estimated and compared for both cases.Finally, estimates of crop irrigation water requirements and crop value per acre were used to estimate the financial benefits associated with each case study.
The paper is organized as follows: section 2 contains a description of the Walker River Basin study area, the available data, and a discussion of the cloud seeding operations conducted by the Desert Research Institute.The hydrologic modeling approach is described in section 3. Results of the model case studies are presented in section 4, and the results and future extensions of the research are discussed in section 5.

Description of Walker River Basin
The Walker River basin area is approximately 10,200 km 2 , and ranges in elevation from near 1,200 m near Walker Lake to near 3,800 m in the headwater areas of the Sierra Nevada (Figure 1).It straddles the California-Nevada border and is bounded by the Sierra Nevada mountain range to the west and southwest, the Pine Nut Range to the northwest, the Desert and Terrill Mountains to the north, the Gillis Range and Excelsior Mountains to the east, and the Bodie Hills to the south.Within the basin, a series of valleys are separated by the Sweetwater Range in the west, the Pine Grove Hills near the middle, and the Wassuk Range to the east.The basin is a closed watershed whose perennial river, the Walker River, originates in the Sierra Nevada as West and East forks.Both streams flow north, until they merge in the southern end of Mason Valley.At the northern end of the valley the river takes a sharp turn toward the east before it turns south and finally terminates in Walker Lake, the lowest part of the basin.
Since the late 1800s, the Walker Basin has been strongly influenced by agriculture development, irrigation (particularly in Smith and Mason Valley), and, more recently, reservoir operations (Bridgeport and Webber Reservoirs, Twin Lakes and Topaz Lake).Groundwater pumping has increased dramatically over the last decade as the demand for water has continued to exceed the surface supply due to overallocation of water resources by governmental authorities and a general decrease in annual precipitation throughout the region.

Climate and Hydrologic Data
The Walker River basin is located within a semiarid climate regime though the precipitation and temperature vary considerably among with latitude, elevation, topography, and season (e.g., Houghton et al., 1975).In this study, a 800-m resolution Parameter elevation Regressions on Independent Slopes Model (PRISM) estimates of monthly precipitation and temperature values, available from the Spatial Climate Analysis Service at Oregon State University (Daly et al., 2001) were used.The 30-year annual average  estimate of precipitation ranges from approximately 100 mm near Walker Lake to over 1,200 mm in the highest elevations of the Walker River headwater region within the Sierra Nevada (Figure 2).The majority of the precipitation in the mountainous southwestern part of the basin falls as snow during the winter months.Water is stored in the seasonal snowpack and released as snowmelt during the warmer, drier spring and summer months.
The United States Geological Survey (USGS) maintains surface water stations on the West and East Forks of the Walker River (USGS 10296500 and 10293000, respectively) that have provided continuous daily streamflow estimates from 1902 and 1921, respectively.The watershed area con-to the flow at these stations is hereafter termed the "headwater" area of the Walker River basin for the purposes of this study.
The SNOw Data Assimilation System (SNO-DAS) product (Barrett, 2003) available from the National Operational Hydrologic Remote Sensing Center (NOHRSC) is a model/data assimilation system product that provides a continuous time series of daily hydrologic variables related to snow cover (e.g., snow water equivalent [SWE], Figure 2: Spatial distribution of average annual precipitation based on PRISM estimates .and 20g (AgI mass content) ejectable flares manufactured by Ice Crystal Engineering.The mass of AgI seeding material released per hour into the clouds by aircraft is, in contrast to that from the ground generators, highly variable based upon the number of each type of flare that is released during a given flight.Such releases are determined by the pilot while on station based upon ambient conditions.
The focus of this study was on three high altitude areas within the Walker River headwaters that are targeted by cloud seeding operations: Sierra Crest, Sweetwater Range, and Bodie Hills (Figure 1).These are mountainous, snow-dominated areas that range in spatial area from 92 km 2 for the Bodie Hills, to 147 km 2 for the Sweetwater Range, to over 679 km 2 for the Sierra Crest with average annual precipitation (Figure 2) of 494 mm, 650 mm, and 1,020 mm, respectively.The vegetation is predominantly coniferous forest in the Sierra Crest, with a mix of pinion pine forest, desert shrub, and grasses in the other areas.The soils range from mostly sand for the Sierra Crest and Sweetwater Range to a mix of sand and loam for the Bodie Hills (61% sand and 39% loam).The Sierra Crest and Sweetwater Range form Case 1 and the Bodie Hills forms Case 2.
The collection and trace chemical evaluation of snow samples at four sites in the Walker Basin during 2003-04 found a high percentage of silver in snow profiles at the Sierra Crest, Sweetwater Range, and Bodie Hills sites, indicating that they were routinely targeted by seeding operations (Huggins et al., 2005).Although lacking a targetcontrol evaluation to quantify seeding results, the trace chemical assessment supports the distribution of estimated seeding results that are assumed to apply in the simulations involving the hydrologic model for the entire study period, which is the subject of the following section.
A strict quantitative estimate of the actual impact of seeding operations on precipitation is unfortunately not available due to the paucity of precipitation measurement sites over the Walker Basin.For the purposes of this study, without such a strict quantitative estimate of the impact of seed-ing operations on precipitation, we assume that when seeding is occurring, a base 10% increase in the amount of precipitation occurs regardless of whether aircraft or ground-based seeding is occurring.This 10% figure is well within the range of seeding effects claimed by studies undertaken in the Walker River Basin since 2000.This 10% increase is modulated by a "Seedability Factor (SF)", which, for ground seeding relates the amount of hours seeded to the amount of possible hours that could have been seeded in each storm.For aircraft seeding, the SF is directly related to reports of aircraft icing intensity from both the seeding aircraft and other aircraft in the general vicinity, while an aircraft seeding flight is in operation.
The SF statistics are available (e.g., Huggins, 2011Huggins, , 2012;;Tilley and Huggins 2013;Tilley and McDonough 2014) on a seasonal basis only for the entire study period.In this study, we used the seasonal values of the SF to modulate the base 10% amount of precipitation for each water year as shown in column 3 of Table 1.Note that no seeding occurred during WY 2010.Also note that we simply multiplied the values for the 10% base case by 0.5 and 1.5 to estimate the base effects for the 5% and 15% cases, respectively, to explore the impacts for a range of 5% above and below the 10% value.

Overview of Modeling Approach
This section presents a description of the hydrologic modeling approach used to estimate the impacts of cloud seeding in select target areas of the headwaters of the Walker River basin on the irrigation water delivery and crop production in the agricultural system below the mountain front during WYs 2004-2013.An overview of the existing monthly water balance and river basin management models and the methods in which they were applied in this study for the two cases are provided below.

Water Balance Model of Headwater Areas
In this study we used a model based on the Thornthwaite scheme, a monthly water balance model (WBM) employed extensively in studies of water-resource management and climate change (Alley, 1984;Wolock et al., 1993 andWolock andMcCabe, 1999).We implemented our WBM in MATLAB based on algorithms presented by McCabe and Markstrom (2007).This method simulates the basic components of the hydrologic cycle (snow accumulation and melt, soil-water storage, evapotranspiration, and runoff) for each watershed grid cell (chosen to be the same as the PRISM cells) using the input monthly precipitation and temperature values from PRISM for each grid cell.
The WBM parameter values used in this study (Table 2) were taken directly from those determined by Barth (2013) in a study aimed at simulating the response of the Walker River headwaters to changes in precipitation and temperature associated with past and possible future climate conditions.In that study, Barth (2013) used a manual multi-step calibration approach to fit average monthly values of potential evapotranspiration (PET) estimates, SNODAS SWE estimates, and the streamflow observations on the West Fork of the Walker River basin over the period WY 2004 through 2011.The parameterization procedure used in that study was based directly on the procedures presented by Hay et al. (2006) and Boyle et al., (2014).In general, the procedure involves adjusting different groups of WBM parameters based on their association with specific hydrologic processes until a reasonable fit to the observations (e.g., PET, SNODAS SWE, and streamflow) are obtained.
The first step in the procedure is to estimate the value for the Hamon Equation Coefficient (ham-onCoef), which directly controls the amount of ET and indirectly controls soil moisture storage, evapotranspiration (AET), and surface runoff (Q).The hamonCoef has a default value of 0.55 based on studies by Hamon (1961) for areas in the humid southeastern U.S. In the arid climate of the Great Basin, however, Barth (2013) found that a hamonCoef of 0.9 was needed in order to produce reasonable fits to long-term monthly average estimates of PET from Farnsworth et al. (1982).The monthly PET WBM simulations using the default and calibrated parameter values are compared with the long-term average monthly estimates from Farnsworth et al. (1982) in Figure 4.Note that a further increase in the hamonCoef was found to provide a better fit to the long-term estimates of PET.However, the increased ham-onCoef values also significantly reduced the fit to the observed streamflow and SNODAS SWE values.The final two steps in the procedure involve the manual adjustment of the snow parameters (Train, Tsnow, and meltmax) to fit the SNODAS SWE values and then a manual adjustment of the runoff parameters (drofrac and rfactor) to fit the streamflow values.Evaluation of the model's ability to simulate the PET, SNODAS SWE, and observed streamflow was based on a subjective visual inspection of the closeness of each fit and the computation and consideration of the objective measures of root mean squared error and bias for each of the three variables.These three steps were repeated until a compromise solution to this multi-objective optimization problem was determined.The final parameter values found by Barth (2013) and used in this study are found in Table Table 2: Thornthwaite model parameter values.
2 and the resulting WBM fits are shown in Figure 5.The same procedure (except for the final step which could not be performed due to the impacts of reservoir operations on the observed streamflow) was used to estimate the parameters for the East Fork of the Walker River basin.Farnsworth, et al., (1982).

River Basin Management Model
In this study, we used an existing river basin management model (RBMM) that was originally developed (Boyle et al., 2013) to inform decision makers of potential outcomes from proposed water right acquisitions in the Walker River basin.The RBMM captures the spatial and temporal complexity of important relationships among crop demand, river flows, groundwater-surface water exchange in the river and drains, irrigation practices, groundwater pumping, and all known existing water rights (e.g., surface decree, storage, and flood) in the Walker River system in Mason Valley and Smith Valley.The RBMM allows users to track water through the complicated deliveries and returns in the heavily irrigated Smith and Mason Valleys and provides a means for better understanding the complex hydrologic relationships within the real system.The primary input to the RBMM are the monthly streamflow hydrographs from the West and East Forks of the Walker River.The current version of the RBMM also requires observed monthly streamflow diversions for each of the many canals and river pumps throughout Mason Valley and Smith Valley.These historic observations were available for each month of this study and were used to define the demand for water on the crops associated with each diversion for the baseline and both cloud seeding cases described in the next section.A detailed survey of the distribution of crops was performed in 2007 and the total crop area and consumptive water requirements for the entire Mason Valley and Smith Valley area are shown in Table 3.While surveys of the spatial distribution of crops for the remaining years of the study were not available it is reasonable to assume that in this area there has been little variation in the crop types by farm; as a result, we assumed crop distribution to be identical to those in 2007.

Modelling Cases for Different Target Areas
While there have been different types (ground vs. aircraft vs. both) of cloud seeding operations, with varying parameters (number of generators, number of aircraft flights, etc.), performed in the Walker River Basin over the past several decades, this study is focused on understanding how cloud seeding operations over specific target areas during WYs 2004 -2013 may have impacted streamflow from the headwaters and the associated surface deliveries to agriculture.To accomplish this, a baseline and two cases were designed for two specific target areas that, when compared with each other, provide an estimate of how additional precipitation from cloud seeding on the target areas results in additional streamflow from the headwaters (based on the WBM) and is delivered to the Mason Valley and Smith Valley agricultural area as surface water irrigation (based on the RBMM).Several assumptions were made in the development of each case, as described in the next two paragraphs.
In this study, we assumed that monthly precipitation values obtained from the PRISM data set for each target area were augmented by 10% (see Table 1 for actual percentages for each year after SF considered) relative to the precipitation that would have occurred without cloud seeding.These conditions are considered the baseline and the impacts of cloud seeding are examined in this study in an inverse fashion, through the reduction of PRISM precipitation over cloud seeding target areas by a percentage of the observed (in order to obtain the estimated 'natural' precipitation hydrologic inputs) and then run through the WBM and RBMM.
The first case assumes that only the Sierra Crest and Sweetwater Range areas are targeted by the seeding operations.The PRISM precipitation values are reduced over the grid cells associated with these target areas by 5%, 10%, and 15% and then run through the WBM and RBMM to estimate a range of impacts of the seeding operations on streamflow and deliveries to agriculture.
The second case assumes that only the Bodie Hills area is targeted by the seeding operations.
The PRISM precipitation values are reduced over the grid cells associated with these target areas by 5%, 10%, and 15% and then run through the WBM and RBMM.The results of both cases are then compared to better understand the value of the seeding operations on these specific targets in the headwaters.

Impacts to Streamflow from Headwaters
The total annual precipitation averaged over the headwater area and the total annual volume of streamflow from the headwater area for the Baseline are shown in Table 4.The differences or changes (compared with the baseline values) in streamflow from the headwater area for the 5%, 10%, and 15% scenarios for each Case are also shown in 2004 through 2013) and, with the exception of WY 2010, the change in streamflow is clearly related to the wet or dry conditions before and during each year (i.e., wet years result in a greater change in streamflow from the headwaters than dry years).In 2010, a year with no seeding operations, there is still a change in streamflow for each of the 5%, 10%, and 15% scenarios in Case 1 and none for Case 2. This behavior is likely associated with the "carry over" of baseflow from the groundwater system from the previously seeded year.Also presented in Figure 6 are the annual estimates of changes in supplemental groundwater pumping and crop shortage (in acre-ft of water not delivered).Supplemental pumping in Mason Valley and Smith Valley is allowed for some of the junior surface water right holders in times when their surface water rights are not delivered.The turquoise bars in Figure 6 represent the additional supplemental pumping that the RBMM determined would occur in response to the decrease in surface water delivered (dark bars) for Case 1.The grey bar in Figure 6 represents the actual shortage (compared with the Baseline) that the RBMM estimated remained due to the inability of all users to use supplemental groundwater pumping to "make up" for the shortage in surface water delivered in the non-seeded 5%, 10%, and 15% cases.Note that the sum of the changes in supplemental pumping and crop shortage for each year is equal to the change in the surface water delivery for each year.As a result, the hydrologic and monetary impact of the cloud seeding can be presented with an upper and lower bound; 1) in terms of the crop shortage related to the total surface water not delivered (i.e., the black bar); and 2) in terms of the crop shortage remaining after supplemental pumping is used to mitigate the decrease in surface water delivered (i.e., the grey bar).This has become an important distinction with the last three years of drought which led to a mandatory curtailment of 50% of supplemental groundwater pumping in the Mason Valley and Smith Valley area.
results for the 5%, 10%, and 15% scenarios in Case 2 are shown for each year in Figure 7. Notice the dramatic difference in the magnitude of the changes in surface water, supplemental pumping, and crop shortages as compared with Case 1.In general, the pattern is the same with the two largest impacts associated with two of the wettest years.However, there is little response in 2005 and 2009 since these wet years followed dry years.The differences between the Case 1 and the Case 2 results are most likely related to the differences in annual precipitation over the target areas; even with the increases in precipitation associated with cloud seeding, there is simply very little runoff generated in the Bodie Hills target area compared with the Sierra Crest and Sweetwater Range.
Using existing estimates of crop values (Curtis et al., 2009) and estimates of total crop area in the Mason Valley and Smith Valley area (Boyle et al., 2013) an estimate of potential crop return and total irrigation water needed can be estimated (Table 5).The quotient of these two estimates provides a rough estimate of the value of surface water delivered to the entire agricultural area each year of approximately $76.41/acre-ft.Using this estimate, Table 5 presents an estimate of the value of the streamflow surface deliveries associated with the 5%, 10%, and 15% scenarios for each of the two Cases when supplemental groundwater pumping is allowed to occur to make up for the reductions in surface water deliveries.Table 6 presents the same information for the condition when supplemental groundwater pumping is not allowed to occur to make up the difference.

SUMMARY AND CONCLUSIONS
In this paper, a set of existing hydrologic models of the headwater and agricultural areas of the Walker River basin are used to estimate the effects of cloud seeding activities on the amount of water delivered to the agricultural areas in the lower part of the basin over the water years 2004 through 2013.Two case studies were presented that identified and compared the results associated with two specific target areas within the headwater area of the basin.
The results from the two modeling cases highlighted the importance of considering the close relationship between the amount of annual precipitation and runoff in selecting a target area for cloud seeding operations.The two target areas in this study with the highest average annual precipitation were found to produce significantly more streamflow at the outlet of the headwater area and VOLUME 47 ~ SCIENTIFIC PAPERS ãt the lower portion of the basin where surface diversions are made to deliver water to crop areas in Mason Valley and Smith Valley.Even when the differences in extent (size) of the target areas are considered, the differences in streamflow generation appear to be most closely related to the annual precipitation.A third case, with all three areas targeted by seeding operations, would need to be simulated to determine if the impacts from cloud seeding are additive.This was not performed in this study but could be done in the future.
The results and conclusions presented in this study are limited to the uncertainties related to; 1) the selection of the target area extent; 2) the relationship between cloud seeding activities and changes in precipitation on target areas; 3) the application of the WBM to each target area under the seeded and non-seeded conditions; 4) the application of the RBMM with historic diversion deliveries used as a surrogate for crop demand in Mason Valley and Smith Valley; and 5) the estimates of crop type, area, water requirements, and value.It is intended for the results to provide potential cloud seeding providers and local water managers in the Walker River Basin with a better understanding of how the different targeted regions of the watershed respond to additional precipitation, which in turn may promote more efficient cloud seeding and water management activities.
map of Walker River Basin and cloud seeding target areas for Case 1 (Sierra Crest and Sweetwater Range) and Case 2 (Bodie Hills).The 800 m grid for the hydrologic modeling of the headwaters is shown as black cells.
snowmelt runoff, snow depth, sublimation).The system combines a spatially distributed, multilayered snow mass and energy balance model with ground measured and remotely sensed snow-covered area, snow depth, and SWE.A continuous time series of these variablesis available from October 2003 to present for the contiguous United States and parts of Canada in high spatial (1km) and temporal (daily) resolution.VOLUME 47 ~ SCIENTIFIC PAPERS Figure 3: Locations of the most commonly used cloud seeding flight lines (straight dashed lines) and four most recently used ground based seeding generators (blue dots).The locations of ground based meteorologic observation stations in or near the study area are shown as red X's.2.3 Cloud Seeding -Operations in the Walker River Basin Cloud seeding in the Walker River Basin dates back to the 1980s.The DRI-run seeding program during the period of study includes four to eight ground-based seeding generators and 8-17 aircraft seeding flights per season, with the number varying with the specific storm tracks and intensities during a given season.The recent four ground generator sites and the most commonly used seeding flight tracks are shown in Figure 3.Both ground silver iodide (AgI) generators and aircraft equipped with flares currently use a mixture pro-ducing ice nuclei similar to those described by Feng and Finnegan (1989).At temperatures below -5 o C, these particles act as ice nuclei, with the efficiency of ice nucleation increasing rapidly at temperatures colder than -8 o C. Because natural ice nucleation processes begin to occur at greater rates at temperatures colder than -20 o C, seeding is usually conducted between a temperature range of -5 o C to -18 o C from ground generators and airborne at levels where the temperature is between -10 o C and -20 o C. Ground generators release about 25 g h -1 of AgI seeding material while the aircraft use both 150g (AgI mass content)

Figure 4 :
Figure 4: Water Balance Model simulation of Potentential Evapotranspiration (PET) on the West Walker River using the default (red line) and calibrated parameter values (black line) compared with the long-term monthly estimates (black dots with thin black line) fromFarnsworth, et al., (1982).
water equivalent (a) and streamflow (b) for the calibration period on the West Walker River.The black lines are the model simulations and the black dots are the observations.

4. 2
Impacts to the Agricultural AreasThe changes in surface water delivered to the agricultural areas in Mason Valley and Smith Valley area due to cloud seeding for the 5%, 10%, and 15% scenarios in Case 1 are shown as differences (Case 1 minus Baseline results) for each year as dark bars in Figure6.Notice that the greatest impacts are in the wet years (except WY 2010 when no seeding occurred over the October 2009 through March 2010 seeding season) compared with the dry years.

Table 1 :
Seeding percentages for each case.

Table 3 :
Estimates of crop water requirements and potential return per acre.

Table 4 .
These results clearly show that the additional streamflow from the Bodie Hills target area (Case 2) is significantly less than the amount generated from the Sierra Crest and Sweetwater Range (Case 1).The model runs for the baseline and each of the two cases were made continuously over the entire time series (i.e., WY VOLUME 47

Table 4 :
Baseline precipitation, streamflow and change in streamflow from headwater areas for the 5%, 10%, and 15% scenarios for Cases 1 and 2.