Investigation of Climate Change Impacts on Water Resources in the California region

Investigators: D. Cayan¹, M. Dettinger², R. Hanson ², T. Brown ³, A. Westerling¹

---

ACPI Progress report  1/19/01  

1.  PCM simulations:  extraction and preliminary analyses
 
      a. Data extracted (monthly and daily average)
          Historical 7 B06.22 (greenhouse CO2+aerosols forcing)  1870-2000
          Climate Change B06.37 (BAU6, future scenario forcing) 2000-2076
          Climate Change B06.40 (BAU6, future sceanrio forcing) 1995-2099
          Climate Change B06.42 (BAU6, future scenario forcing) 1995-2099
      b. Analysis

2.  Effects upon snowmelt-driven watersheds in the Sierra Nevada

      a. statistical downscaling to Sierra watersheds
      b. historical runs, Carson, American, Merced Rivers

3.  Effects upon water resources in California coastal Aquifers

      a. sea level change sensitivity runs

4.  Summer Aridity--Wild Fire impacts

      a. fire danger code
      b. historical wildfire dataset
      c. analysis of variability fire starts vs. climate patterns

Appendix A:  Model Simulation Archive

---

1. PCM simulations: Data extraction and preliminary analyses

Our efforts towards extracting data from new PCM simulations continues. Currently three new PCM simulations have been completed: B06.40, B06.42 and B06.43. All three of these simulations are Business-As-Usual future scenario forcing simulations. We have extracted the monthly data for B06.40 and B06.42 and are currently extracting the daily average data from B06.40 (Appendix A). We will extract monthly and daily average data from all three simulations. Tables of the variables we are extracting can be found below (Monthly, Daily and 6-Hourly Variable Table in Appendix A).

Our preliminary analyses of the PCM data have focused on the amount of variability and change in the monthly data. We extracted the PCM data along the west coast of the Americas for 20-year periods. Looking at 2m air temperature along the coastal land strip, values continue to warm during the climate change (B06.37) simulation and are generally 1.5 to 2K warmer into the later part of the simulation as compared to the historical (B06.22, years 1870-1969) simulation (Figure 1).

Total winter (ONDJFM) precipitation anomalies are more variable along the coastal strip (Figure 2).
 

Some 20-year periods during the climate change simulation appear to be significantly drier than the historical simulation along the US coast which is of concern for our regional investigations. Interestingly, this pattern tends to be symmetrical about the equator, indicating that a global circulation change is involved. The global ONDJFM total precipitation for the period 2059-2075 reveals a split ITCZ with an interesting dipole south of the equator (Figure 3).

Figure 3

 

2.  Effects upon snowmelt-driven watersheds in the Sierra Nevada

Much of the water supply in California and the western United States is derived from snow melt runoff from mountainous watersheds. To investigate impacts of climate change, we will use the relevant output from ACPI global and regional-downscaled model simulations to drive well-calibrated hydrologic models of selected mountainous watersheds in the Sierra Nevada. These models, which have been run to elucidate variability under present day historical conditions and a range of climate change scenarios, include the Merced River, the North Fork American River, and the Carson River (e.g., Jeton et al., 1996; Dettinger et al., 1998; Wilby and Dettinger, 2000; Miller et al., 1999), and cover a broad range of elevation, exposure to storms, and latitude. Tests to be performed using these models will include:

• How well do the atmospheric models simulate present day mean and variable conditions, as applies to Sierra Nevada hydrology?
• What, why, and how significant are long-term changes in the volume and timing of runoff (snowmelt vs. rainfall) and other hydrologic properties (snow pack,  soil moisture) that might occur from global change runs?
• How well do the simulations provide input necessary to reproduce present day occurence of extremes (floods and drought), and how, why, and how significant are changes over the course of the climate change runs?
• In a changing climate of the sort simulated by the global and regional models, how well can flood and drought frequencies be estimated? Can engineering thresholds, such as "one exceedence in fifty years", be estimated or replaced in any practical way, to accomodate the statistical nonstationarities implicit in global-change projections?
• To what extent do the regional atmospheric model simulation outputs provide improved inputs for the hydrologic models relative to those from the global model, and how much do the global-change sensitivities of the hydrologic models depend on the use of regional vs. global climate models?

In order to begin to address these questions and those ennumerated under the other two thrusts (3. and 4. below), we have begun to extract a GCM regional to global model data subset from the PCM historical and climate change simulations archived at the NERSC facility in Berkeley. This set of California-related vaiables is broad enough that it may serve other ACPI application needs as well, and is available upon request.

Parallel Climate Model Historical Simulations of Sierra Nevada Streamflow

Daily precipitation and reference-height temperatures from PCM Historical Simulation #7 (1871 through 1972) have been used as input to precipitation-runoff models of the Merced, Carson, and American Rivers in the Sierra Nevada. These simulations demonstrate the linkages available to use climate-model simulations to project possible hydrologic responses to historical and future climate scenarios.

In order to make the connection between the global-scale climate simulations and the watershed-scale (on order ot 1000 km**2) hydrologic models for this initial demonstration, a simple statistical correction of the grid-scale daily climate series from Historical Run #7 was performed. Because the coupled climate model's historical simulation represents just one realization among the infinite number of possible climate histories in the last century (under historical radiative forcings)--whereas the actual historical climate has represented another, no one-to-one relation can be expected between the simulated and historical precipitation and temperature series over California. Therefore, the simulated climate series were not be expected to directly mimic the weather inputs to the watershed models (which are designed to be driven by time series from local weather stations). Instead, the monthly-mean probability distributions of daily temperatures (maximum and minimum), frequency of wet vs. dry days, and wet-day amounts for the 1951-1972 period from the simulation were mapped onto the observed distributions from the same period. This time period is one in which all three river models had historical observations to allow the proper mappings to be developed.

Figure 4 shows an example of the mapping for one of the weather stations that drives the Carson River model, with black circles corresponding to the observed temperature and precipitation exceedences and the red circles corresponding to the mapped PCM values from a grid cell over Northern California. Notice that the observed and simulated temperatures compare very closely; the mapped, simulated precipitation amounts differ somewhat among the wettest 10 days but otherwise agree very well. The number of wet days in the mapped, simulated series and in the observations are exactly equal. Upon mapping, the lag-one-day correlations of simulated temperatures and precipitation (not shown) are very similar to the observed series over all; this simulated autocorrelation is a direct result of the dynamics in the climate model, as no attempt was made to correct the mappings for antecedent conditions. 

Figure 4

The mappings developed to relate the simulated and observed climate series are then applied--without change--to other parts of the historical simulation and, at least initially, to the subsequent climate-change simulations. One question that arises in this mapping procedure, when applied to COUPLED climate model outputs, is whether the large-scale climate of the model and observations were sufficiently similar during the training period. That is, suppose that the coupled model during the 1950-72 period had been in a different Pacific decadal climate regime than was historically observed. The mappings might then have the effect of making an El Nino-rich model climatology of temperatures and precipitation look like an observed La Nina-rich climatology. Then in other parts of the simulation, when the model became more La Nina-rich, the mappings would possibly overcompensate, making the mapped climatology resemble a super-La Nina decadal period.

As it happened, in the Historical Simulation #7, the 1950-70 period experienced remarkably Pacific decadal climate regimes, at least as measured by the simple index plotted in Figure 5. Consequently, the 1950-70 training period for the Historical Simulation #7 may well be the most appropriate period available.

Figure 5

The daily temperatures and precipitation from the entire Historical Simulation #7, 1871-72, period were mapped to resemble the distributions from the watershed-model input sites using the relations from 1950-72. If during earlier periods, or in later simulations, the character of the daily climate changes, the mappings will ensure that the inputs to the watershed models change accordingly. Upon running the three Sierra Nevada river models with Historical Simulation #7, the resulting streamflow simulations were compared to observed flows (during periods when flow observations were available, and to flows simulated using observed meteorology as inputs.

Figure 6 compares the annual-flood series from observations, simulations with observed meteorology, and simulations with the PCM climate history (#7). The annual-flood series is a series comprising the annual maximum-daily flows at each river. The figure compares exceedence probabilities estimated from the three sources.

Figure 6

The correspondences between observed and simulated flood probabilities are best for the Carson River model. The Merced River model yields relatively similar simulated flood probabilities, but both simulations tend to overestimate the probabilities of the largest floods.Simulated changes in flood frequencies in both the Carson and Merced Rivers, due to greenhouse gases, will probably fall outside the range of differences shown above and thus will be useful. The American River flood probabilities estimated from the PCM history is least like the observed probabilities or the probabilities from simulations with observed meteorology. [The probabilities plotted are based on the entire periods of "record" available in each case; that is, they are based on 100 years of PCM history and anywhere from 12 years (for Carson) to 22 years (for American) to 85 years (for Merced) of observed flows and meteorology.]

Figure 7 compares the observed and simulated monthly flow climatologies of the three rivers. [As before, the full periods of "record" are used in each case.] Comparisons are encouraging, although the Carson simulation with PCM history is generally too wet and the simulated American flows are too early. This tendency may reflect a tendency for wet days to be cooler than dry days overall in the Sierra Nevada, a tendency that has not been explicitly accomodated in the simple weather mappings used here.

Figure 7

These initial simulations encourage us to believe that we will be able to identify and (broadly) quantify the influences of future climate changes simulated by the PCM with the Sierra Nevada river models. The simple "correction" procedure used here may be limiting the results here and certainly may be suspect in the subsequent future-climate investigations. Thus we await eagerly the next-generation downscaled climate scenarios that other parts of ACPI are developing.
 
It turns out that the period from about 1950 to 1970 (when our copy of the simulated data runs out) the pdo states are in remarkable agreement. This is strictly fortuitous; in the coupled ocean-atmosphere model, there is no particular reason for the actual pdo transitions to line up with the real world, just their amplitudes and periodicities should be similar.

References:

California Applications Program. Web site location:  http://meteora.ucsd.edu/cap/

Dettinger, M.D., Mo, K., Cayan, D.R., and Jeton, A.E., 1999. Global to local scale simulations of streamflow in the Merced, American, and Carson Rivers, Sierra Nevada, California. Preprints, American Meteorological Society's 14th Conference on Hydrology, Dallas, January 1999, 80-82.

Hall, B., T. Brown and J. Roads, 2000. Utilization of Scripps ECPC Forecasts for Regional Monthly Assessments. Proceedings AMS Fireweather Conference. Long Beach, CA.

Jeton, A.E., Dettinger, M.D., and Smith, J.L., 1996. Potential effects of climate change on streamflow, eastern and western slopes of the Sierra Nevada, California and Nevada. U.S. Geological Survey Water Resources Investigations Report 95-4260, 44 p.

Knowles, N., 2000. Modeling the Hydroclimatology of the San Francisco Bay-Delta Estuary and Watershed. Ph.D. dissertation, University of California - San Diego, Scripps Institution of Oceanography, 277pp.

Miller, N., Kim, J.W., and Dettinger, M.D., 1999. California streamflow evaluation based on a dynamically downscaled 8-year hindcast  (1988-1995), observations, and physically based hydrologic models. Eos, American Geophysical Union Fall Meeting supplement, v. 80, p. F406.

Wilby, R.L., and Dettinger, M.D., 2000. Streamflow changes in the Sierra Nevada, California, simulated using statistically downscaled general circulation model output. Chapter in "Linking Climate Change to Land Surface Change", S. McLaren and D. Kniveton (eds.), Kluwer Academic Publishers (Netherlands), in press, 20 p.

 

3.  Effects upon water resources in California coastal Aquifers

Coastal aquifers in California supply water to several of the nation's largest cities and population centers, and are strongly affected by fluctuations in precipitation and evaporation. These aquifers are being exploited at, or beyond, the natural limits of their capacities as renewable resources. One consequence of this kind of overdraft is that the aquifers are threatened by salt water intrusion and land subsidence when/if ground-water levels fall below certain threshhold levels. Another consequence is that the aquifers are currently being managed by close coupling with various local and imported sources of surface water, in order to replenish some of the overdrafts of water. Using the Ventura basin north of Los Angeles as a prototype of such surface-water/ground-water systems, a fully calibrated, combined surface and ground water model of this basin will be used to address the following questions:
 

• How well does the combination of atmospheric models and the ground-water/surface-water flow model (e.g., Hanson and Dettinger, 1999) simulate present-day mean and variable conditions in surface water flow and ground water levels?
• What, why, and how significant are changes in streamflow, ground water levels, and other hydrologic properties (soil moisture, evaporation) that might occur from global change runs?
• What changes in frequency of extreme conditions, such as floods, drought, and severe low ground water levels which exceed salt water intrusion threshholds, are implicit in the global-change scenarios?
• To what extent does the use of the regional atmospheric model simulation outputs improve the present-day hydrologic model over that from the global model, and how different are results from the global change runs from the regional vs. global models?
Climate Change and Water Resources in the Ventura Basin  
In southern California, ACPI members Randy Hanson and Mike Dettinger (USGS) are carrying out simulations of ground-water/surface-water responses and management to selected PCM climate simulations. The Santa Clara-Calleguas region with its many forms of water-resources management are depicted schematically in Figure 8. At the heart of this management system is the Santa Clara-Calleguas aquifer system.
 
Figure 8


The USGS developed the ground-water/surface-water model for the United Water Agency of Ventura (depicted in Figure 9) during the 1990s and it is used to address management issues concerning the impacts of ground-water and surface-water developments and usage in various parts of the basin.
 

Figure 9


The overall extent of climatic influences in a coupled ground-water/surface-water system such as this can be difficult to guess and depends on how closely the ground-water and surface-water systems are coupled, on which system dominates the year-to-year hydrologic variations in a basin, and on the time scales and avenues by which the climate forcings enter the particular hydrologic systems. On the whole, however, ground-water systems are well known to respond most vigorously to long-term, decadal and longer climate variations, and thus may be most susceptible to long-term climate changes. Thus the extent to which climate projections will be useful in management of ground-water/surface-water systems is initially uncertain. As part of ACPI, the Santa Clara-Calleguas ground-water/surface-water model will be forced by a variety of PCM historical and future climate simulations to determine the sensitivities and responses of this heavily developed coastal basin to climate variability and change. (An indication of the location of the Santa Clara-Calleguas basin in comparison to the scales typical of global climate models is given by Figure 10.)
 

Figure 10


As a first step, several experiments have been made to test the sensitivity of the simulated ground-water/surface-water system to a consequence of global warming that is not strictly due to local weather changes: the potential influence of projected sea-level changes on this coastal aquifer. Figure 11 shows the simulated (and observed) water levels in a well near the coast under several idealized scenarios. First the model was used to simulate historical ground-water level variations with no changes in sea level (green curve). This is actually how the model was calibrated. The calibrated model (overall) is capable of reproducing the observed (blue curve) water-level variations well over their period of overlap. Second, the model was run for the same 1890-1994 period with a steadily rising sea-level that reflects regional estimates of historical sea-level rise along the Southern California coast, +20 cm/century (yellow curve). The model was also run with monthly sea-level changes observed near san Francisco (orange curve). Both of these historical sea-level rises result in a higher water levels as the end of the simulations approach; indeed, these runs with historical sea-level rises may offer improvements upon the original calibration results. Finally the mode was run with an imposed sea-level rise equal to rates that have been suggested under future greenhouse-warming scenarios, 50 cm/century (red curve). This rate of rise results in increased water levels that are quite comparable to the influences of historical climate and pumping variations during the 20th century. Thus, at first glance,, it appears that sea-level changes may be prove to as important as climate variations and changes in the next century in the coastal aquifers and water resources of California.
 

Figure 11


ACPI activities will include ground-water/surface-water simulations of aquifer and stream responses to simulated historical and future climate variations, with and without attendant sea-level changes. 

References:

California Applications Program. Web site location:  http://meteora.ucsd.edu/cap/

Hanson, R.T., and Dettinger, M.D., 1999. Ground-water/surface-water responses to ensembles of global climate simulations, Santa Clara-Calleguas Basin, Ventura County, California, 1950-93. Eos, American Geophysical Union Fall Meeting supplement, v. 80, p. F215.

4.  Effects on summer aridity and fire potential in the California region

CEFA has obtained from the U.S. Forest Service Missoula Fire Laboratory computer code for the National Fire Danger Rating System.  This code, written in Fortran and C languages, ingests a number of input parameters
(shown in boxes with red headings in Figure 12) and outputs several parameters which describe fire danger at the location of interest. CEFA has been in the process of adapting the code for analysis as an applications component
of ACPI. The goal is to assess change and variability in fire danger over North America for the 100-150 year time period (1960-2080).
 
Figure 12

 

The two primary NFDRS components of interest are the Energy Release Component (ERC) and Burning Index (BI).  The ERC is defined as the potential available energy per square foot of flaming fire at the head of the fire and
is expressed in units of BTUs per square foot. It is used operationally as an indicator of fire severity. Because of the inputs needed to compute the ERC (in particular the 1000-timelag dead fuel moisture), it may be considered in some regard as a longer-term (i.e., weekly, monthly) climate indicator. The BI is a number indicating the potential amount of effort needed to contain a single fire in a particular fuel type within a rating area. It is tied directly into determining the type and amount of suppression staffing and resources to put on stand-by (Figure 13).
 

Figure 13


This has a cost component involved of which Tony will attempt to analyze and assess as part of the project. The economic objective will be to estimate costs associated with suppression activities as a result of climate change
and variability. The meteorological input elements include maximum and minimum temperature, maximum and minimum relative humidity, 1300 local time cloudiness, relative humidity and temperature, and precipitation duration. These elements are currently available from the NCEP/NCAR reanalysis. CEFA has obtained the
necessary data files for the period 1958-2000. The gridded data will be used as input to the NFDRS code, and ERC and BI values calculated and analyzed for the 40-year time period.

The necessary input elements are also available from the 6-hourly ACPI model runs and are being saved at NCAR.  It is proposed that Mary extract these elements from the global grids and provide to CEFA for analysis sometime
during the first quarter of calendar 2000. Only the first 20 years of the 80 year run will be acquired, though it will be desriable to eventually obtain the entire period for each of the various scenario runs.

In recent years, an estimated $2 billion has been spent annually for fire suppression. Figure 14 shows the record of natural fire starts from 1986-1998 as recorded by the Bureau of Land Management (BLM) and the United States Forest Service (USFS).
 

Figure 14


This figure indicates the tremendous spatial variability in fire starts while Figure 15 shows the interannual variability over California in natural fire starts from 1970 to 1996 (from the USFS data).
 

Figure 15


Climate model simulations can be used to better understand conditions associated with summer aridity in this region. Figure 16 shows the percentage of annual average fires for each month of May through October, 1980-1999 (based on 330,000 individual fire reports from the Forest Service, the Bureau of Land Management and Bureau of Indian Affairs). The seasonal cycle of local fire seasons is illustrated in Figure 16. (Westerling, 2000).
 

Figure 16


Summer dryness has a strong influence on the California water economy and upon the growing problem of wildfire. Daily temperature, precipitation, wind speed and solar radiation will be used to assess variability and change in two components of the National Fire Danger Rating System. The Ignition Component (IC) is a number which relates the probability that a fire will result if a firebrand is introduced into a fine fuel complex.  The Energy Release Component (ERC) is defined as the potential available energy per square foot of flaming fire at the head of the fire and is expressed in units of BTUs per square foot. The IC will provide a useful assessment of variability and changes in fire starts (will be particularly useful for assessing changes in intraseasonal variability). The ERC is more of a seasonal indicator of fire danger, and thus useful for assessing interannual variability. It is anticipated that the results will provide indications as to whether or not a significant trend may occur in fire danger over the ACPI 100-year time frame. Also, it is anticipated that multi-year periods will be revealed that can be related to preferred or non-preferred times for prescribed burning. Thus, the results can potentially have policy and long-term planning implications for the fire management agencies. In this proposed project, three primary components will be examined:
 

• How well do the present day simulations reproduce summer dryness (precipitation, atmospheric moisture, wind), and related wild fire potential, as measured by a relevant fire weather index?
• What is the response of these conditions to changes simulated in the climate change runs?
• What are the socio-economic impacts, as measured by estimated dollars spent on mitigation, suppression, and burn damage, resulting from a climate driven change of the fire potential?

References:

Brown, T., 1999. Development of a Southern California Fire Weather Severity Index. Final Report, The Wildfire Hazard Reduction Training and Certification Program, University of California - Berkeley, Forest Products Laboratory, May 1999, 13 pp.

Brown, T.J., and J.L. Betancourt, 2000. Effect of climate variability and forecasting on fuel treatment schedules in the western U.S. Proceedings Joint Fire Science Conference and Workshop, Vol. II, Boise, ID, 15-17 June 1999, University of Idaho and the International Association of Wildland Fire, 167-172.

California Applications Program. Web site location: http://meteora.ucsd.edu/cap/

Program for Climate, Ecosystem and Fire Applications (CEFA). Web site location:  http://www.dri.edu/Programs/CEFA/

Westerling, A., 2000. Western United States Fire Climatology. AGU Fall Meeting, December 15-19, San Francisco.

Note that even though the 6-hourly and daily data were saved under different simulation names, the simulations are identical bit-for-bit.

We have also archived data from run B05.03 (Historical 1 simulation; PCM1.1 Historical GHG + SO4 Scenario).

 

Note all data are in global T42 arrays. T42 global resolution is 128 longitudes by 64 latitudes.

Note all data are in T42 arrays over the region 54N to 10N and 160W to 90W.

Note all data are in global T42 arrays. T42 global resolution is 128 longitudes by 64 latitudes.