Scenarios of the Future

drought_wwf

What might the flows of Espirito Santo look like, under different conditions of climate change?

     Increasing temperature and precipitation changes are expected to bring fundamental changes to the country of Brazil as predicted by Global Climate Models (GCMs) (IPCC, 2007). Potential effects of climate change in Brazil suggest changes of 4-4.5°C in surface temperature as a result of increased CO2 concentrations. In plantation states such as Minas Gerais and Espírito Santo, a much drier climate is predicted to be a result of global warming and/or reduced water vapor transport from Amazônia. As water quality and availability become increasingly stressed with climate change, water managers in Espírito Santo must plan for the ensuing uncertainties,  but currently have little information to work with.

     This section seeks to address how changes in climate will affect streamflow runoff across the region. The issues include:

  • Is it possible to identify (construct future scenarios of) the impacts of climate change on agricultural activities and the management of water in urban areas?
  • If the hydrologic modeling is based on predictions of regional climate models, could the potential conditions of a near-term future be anticipated?
  • Could early warning signals be adopted?

     There are different ways to approach this problem. The strategy here is to build on the application of the VIC land surface hydrology model and the development of the indices of flow sensitivities described in previous sections, with future projections of climate that are obtained from the Coupled Model Intercomparison Project (CMIP5). We then extend the sensitivity-based approach to gain a better understanding on future land-atmosphere scenarios.

     This approach can produce first-degree approximations of long-term hydrologic responses due to climate changes. The sensitivity-based approach is based on measures of precipitation elasticity and temperature sensitivities (whether streamflow change is more sensitive to precipitation (P) or temperature (T) change in a particular river basin) as defined in many previous studies. This provides a simplified way of incorporating climate change information into planning by making the process less computationally intensive and more accessible, yet still based on physical processes.

Runoff Sensitivities in Espirito Santo and the Rio Doce

    Spatial maps for precipitation elasticity, ε and temperature sensitivity, S were produced and ε values ranged from 1.5 to 8 (Fig. 1). Maps of sensitivities provide an overview of the interactions between precipitation, temperature and streamflow. For example, if ε is 3, a 10% increase in precipitation would result in a 30% increase in streamflow.  When precipitation is perturbed up to 30%, a streamflow increase of about 150% is predicted using the sensitivity analysis (ε = 5.5). This is plausible if the upper soil layer is relatively thin, and the water holding capacity of the upper soil is thin. Very wet, tropical areas that receive large annual amount of precipitation are likely to see huge increases in streamflow as precipitation increases since runoff and baseflow are limited by soil saturation and evaporation.

    For temperature sensitivity, S remains relatively unchanged relative to the reference T for all the gauges i.e. sensitivities < -0.1%. Sensitivity increases slightly after a 3°C perturbation. The magnitude and direction of S demonstrates how land-surface hydrology can both exacerbate, and more rarely modulate, regional scale sensitivities to global climate change. Generally, as T increases, ET increases and runoff decreases (resulting in negative S). The low sensitivities suggest that the streamflow regime in Espirito Santo is influenced mostly by precipitation.

 

Figure 1: Spatial maps of precipitation elasticity (ε) and temperature sensitivity (S)


Scenarios of the Future

     The next problem is to evaluate potential future conditions. The process is to examine downscaled global model results, in terms of the sensitivities derived from the historical records. The downscaled and biased corrected monthly global precipitation and temperature from atmosphere-ocean general circulation models (AOGCMs) participating in CMIP5 were obtained from the "Downscaled CMIP3 and CMIP5 Climate and Hydrology Projections" archive at http://gdo-dcp.ucllnl.org/downscaled_cmip_projections/. Since the sensitivity-based streamflow estimation only accounts for the long-term annual average discharge, the monthly product was sufficient. Six AOGCMs (GFDL-CM3, GISS-E2R, HadGEM2-CC, HadGEM2-ES, INM-CM4 and MPI-ESM) that performed best in terms of precipitation simulations and its related processes over the tropics and for southeastern Brazil were selected (Lei, 2013). The Representative Concentration Pathways (RCP) 4.5 and 8.5 scenarios were chosen since these scenarios take into account modest and extreme emission scenarios.

     To test for linearity and correlation between the UW dataset and the AOGCM historical precipitation and temperature (both maximum and minimum), the seasonal mean and seasonal bias of both datasets were calculated; three-month blocks beginning in January were taken as one season. The correlation coefficients were then calculated between the UW dataset and the AOGCMs. All AOGCMs used in this study yield the same seasonal precipitation amounts since the dataset has been bias corrected spatially and downscaled according to methods outlined in Maurer et al. (2007). The differences between the UW dataset and the AOGCM historical precipitation were calculated and found that the UW dataset precipitation is lower in low-lying basins such as Cachoeira, Rive and Itabapoana. Seasonally, the highest biases between the UW dataset and the AOGCMs tend to occur in the rainy seasons, which may explain the lower magnitudes in VIC simulated streamflows.

     To estimate future streamflow changes with the sensitivity-based approach, P and T changes from GCM output were multiplied by their related hydrologic sensitivity measures (ε, S; under section on Discharge Regime) to estimate the long-term average percent change in streamflow  (dQ) at a specific location and time period (Eq. 1), where dP is the long-term average percent change in P and dT is the difference in long-term average T between the future and historical GCM simulation. d(P,T)int is the interaction between P and T changes which was neglected, due to the additive nature of S and ε in the Colorado River basin reported by Vano et al. (2012). The future streamflows for two 30-year average time periods (2020 – 2049 and 2050 - 2079) relative to the 1980 - 2006 historical period was then calculated using the formula below:

Eq. 1

dQest = dPGCM*ε + dTGCM + d(P,TR)int

where

dPGCM = (PGCMfut – PCGMhist)/PGCMhist

dTGCM = TGCMfut - TGCMhist

     Adjustments to improve the performance of the sensitivity-based approach were applied as outlined in Vano et al., (2013). The adjustments account for variations in annual e and S values as a function of dPGCM and dTGCM respectively. The seasonal variations in temperature were accounted for by applying the monthly dTGCM and monthly Smon values.

     In addition to providing a “shortcut” method for estimating future flows, the sensitivity-based approach allows the influence of temperature and precipitation changes to be segregated, and in so doing encourages better understanding of the factors that will drive changes in the hydrologic system. For instance, maps of ε and S can be used as a preliminary guide to determine the targeted areas for further studies and future management plans.

Outcomes

     The results show that precipitation elasticity, e is generally higher in lower precipitation regimes and a decrease in precipitation in already dry regions is more sensitive to further decrease in precipitation. This includes areas in the interior of Minas Gerais and north of the Espírito Santo. Annual temperature sensitivities are marginal (less than 0.001) and on a spatial basis, all sensitivities are negative. The low sensitivities indicate that streamflow regime for Espírito Santo is largely sensitive to precipitation alone.

     Most AOGCMs predict precipitation increases of about 25% while temperatures will increase by about 4– 6ºC per 30-year period. The estimates here of future streamflow using the sensitivity-based approach suggest an average increase of about 10% across the region for middle emissions scenario. The most sensitive areas to changes in precipitation are the upper ES and coastal regions.

Figure 2: Projected changes in precipitation and temperature for all seven basins were calculated for three 30-year average time periods (2010-2039, 2040 and 2050-2089) relative to the 1980 - 2006 historical period. The data are from the CMIP5 AOGCM output.

    

   Using a combination of Eq. 1 and future changes in temperature (dTGCM) and precipitation (dPGCM), we calculate the estimated future streamflow for all the major gauges. We find that in general, streamflow is projected to increase although the difference between each thirty-year period (2010 – 2039, 2040 – 2069, 2070 – 2099) or scenario (RCP 4.5 – RCP8.5) is very small (<3%) which indicates that streamflow is limited by water availability and is dependent mostly on precipitation.

Figure 3: Precipitation elasticity (ε) for the seven gauging stations from VIC simulations

 

Figure 4: Temperature sensitivity (S) for the seven gauging stations from VIC simulations.

Figure 5: Predicted rate of change of streamflow, ΔQest for the seven basins by the different selected AOGCMs for RCP4.5 and RCP8.5 for each corresponding 30-year interval (2010 – 2039, 2040 – 2069 and 2070 – 2099

 

    The estimates here of future streamflow (ΔQest) for using the sensitivity-based approach suggest an average increase of 20% to 45% across the region. Differences between each scenario (Fig. 5) are small (<5%). The largest streamflow increase was calculated for Cachoeira, located in the Northern part of the state where it is the hottest and has the largest variation in precipitation predictions from the AOGCMs resulting in a large variation of ΔQest. Results from the sensitivity-based analysis are consistent with more robust studies of estimated global streamflow studies estimated from full GCM analyses (e.g. Koirala, 2014). Medium basins in the central part of the basin (Guandu and Itabapoana) have higher precipitation changes and subsequently higher ΔQest.


Focus issues/implementable strategies.

  • Is it possible to identify (construct future scenarios of) the impacts of climate change on agricultural activities and the management of water in urban areas?

Changes in the climate forcings and the sensitivity of (specified) landscapes can be readily computed in VIC. Determining the actual “impact” of these changes in forcing on agriculture and urban areas would be the next step in the process. The sensitivity of, for example coffee, to the changes in precipitation and temperature would have to be addressed specifically. But the environmental conditions to do so are provided.

  • If the hydrologic modeling is based on predictions of regional climate models, could the potential conditions of a near-term future be anticipated?

The process of the hydrologic modeling of scenario futures is based explicitly on the predictions of regional (including global downscaled results). Outputs from regional climate models can be easily coupled to VIC (as demonstrated above) by using either the sensitivity method for a back of the envelope calculations or the output from the RCM fed directly into VIC. The caveat for how far to “trust” such models is the uncertainty in the climate models themselves.

  • Could early warning signals be adopted?

The hydrologic models (VIC, DHSVM) can be used as the robust base model to provide early warnings. At the longer climate scenario time scale (several decades out), the most sensitive regions could be identified, and options for different crops considered. At shorter time scales, drought indices could be provided with a one-month lead-time.