National Assessments and Information for Managing Water Dependant Sectors in a Changing Climate

Activity Rationale

Canada's forest, agriculture, and energy sectors rely heavily on freshwater resources. Information required by these sectors to manage these resources under changing water regimes is not consistently available at a regional to national scale. This activity builds on existing work that has developed observing systems and modelling approaches to determine national water quantity trends.

Leader: Richard Fernandes

The Topic

Canada is a water rich country, containing approximately 6% of the Earth’s renewable freshwater.  Our ecosystems and economy rely on freshwater.  The freshwater cycle includes precipitation (rain and snow), evapotranspiration and condensation (water vapour moving from the land to the atmosphere), sublimation (ice and snow moving to the atmosphere as water vapour), runoff to streams, and the net change in groundwater storage.

Precipitation and runoff are consistently measured by the Water Survey of Canada, allowing us to quantify their climatic interactions and trends over time.  For example, from 1950 to 1998, precipitation has increased by 5-35% and the ratio of snowfall to total precipitation has also increased (Zhang et al. 2000).  In contrast, evapotranspiration, condensation, and sublimation (lumped together as evapotranspiration) are not consistently measured even though they constitute, on average, 55% of precipitation over North America. 

In the absence of soil moisture limitations, evapotranspiration will increase as temperature increases (Maidment, 1993).  Temperature has increased by 0.5-1.5ºC across Southern Canada and faster over northern Canada between 1950 and 1988 (Zhang et al. 2008).  Intergovernmental Panel on Climate Change (IPCC) Global Circulation Model (GCM) projections indicate a likely range of 21st century temperature increases between 2 to 6 ºC over Canada. Therefore, freshwater availability is constrained by evapotranspiration losses that may change with changing climate.

The goal of this research is to improve our knowledge of Canada’s freshwater cycle, focusing on trends in evapotranspiration.  To do so the following questions were asked:

1. What are the historical temporal and spatial patterns in evapotranspiration across Canada between 1960 and 2000 and are they related to surface temperature trends?
2. What are the projected regional trends in evapotranspiration in regions of Canada where we have confidence in our historical estimates and where water issues are at the forefront?

Results

Question 1 – Historical Trends in Evapotranspiration

To answer this question, a numerical land surface model (EALCO- Ecological Assimilation of Land and Climate Observations) developed in the ERCC activity Developing Earth Observation-based Ecosystem Modelling Tools for the Assessment of Climate Change Impacts was used at stations with climate data and then with gridded climate inputs (climate variables from spatial grids of data). Sensitivity analysis with the model over field sites in Southern Ontario showed that evapotranspiration is sensitive to vegetation density (as defined by leaf area index) and land cover (see figure below).

Sensitivity of evapotranspiration to vegetation density and land cover over the Oak Ridges Moraine, Toronto, based on the EALCO model and calibrated with in-situ measurements (Simic et al. 2004).  Bare regions have high evaporation due to exposed soil.  Vegetated regions have similar increasing transpiration as vegetation density (leaf area index) increases. Larger image

Land cover and leaf area index maps of Canada, produced by the ERCC project National scale satellite climate data records of Canada's landmass and ecosystems, were used to parameterize the EALCO model (a parameterized model has all of the information required to simulate the dynamic response of the modelled system). Based on this framework, historical trends (1960-2000) in evapotranspiration at climate station locations were determined (see figures below). Increases in annual evapotranspiration of up to 25% were identified at 81 locations and decreases of up to 10% were found at the remaining 20 stations. Statistically significant increasing evapotranspiration trends of up to 25% were detected in 30 of the locations with the majority corresponding to Atlantic and Pacific coastal regions. Increasing trends were generally related to increasing temperature except in the Prairies. Annual evapotranspiration trends in the Prairies were mixed in terms of increases and decreases with no locations showing statistically significant trends.

Historical trends (1960-2000) in evapotranspiration at climate station locations. Upper panel indicates historical trends in actual evapotranspiration at locations across Canada. Lower panel identifies statistically significant trends as red triangles.  Other trends may have large magnitude (e.g. over Prairies) but may be statistically insignificant due to large annual variability unrelated to climate warming (Fernandes et al. 2006). Larger image

The EALCO model was then applied to all of Canada using climate inputs from the US National Centers for Environmental Prediction (NCEP) Regional Reanalysis 2 Model; however, it was found that the climate inputs had biases in precipitation variables. While NCEP was addressing these issues, the historical gridded climate datasets were used to produce Canada-wide annual evapotranspiration estimates. The estimates are presented at Environment Canada’s RésEau website. The information is useful for general reference but requires local validation before it can be used.

Question 2 – Projected Trends in Evapotranspiration

Locations of study areas used for the regional assessment of projected trends in evapotranspiration during the 21st century. Larger image

To quantify the likely range in annual trends in evapotranspiration and water availability (precipitation minus evapotranspiration), an approach recommended by the IPCC was followed. The approach begins with a validated historical (baseline) estimate for the 1960-1990 period. Two regions that have significant water supply issues and that represent extremes in the ratio of evapotranspiration to precipitation were chosen: the western Prairie region and the Annapolis Valley, Nova Scotia (see figure to the right). The model was validated by comparing annual estimates of observed precipitation less runoff. Decadal average differences between modelled and water budget-based evapotranspiration were less than 10% in both cases.

Three IPCC GCM climate models were then selected that fell near the middle and at the boundary of the ‘likely’ (67%ile) range of changes in temperature and precipitation over the 21st century (detailed information on IPCC climate models can be found in IPCC Fourth Assessment Report, Working Group 1, chapter 8).  Projected changes in climate variables required for our land surface model were produced for each GCM for three different climate scenarios: A1, business as usual; A2, intensive fossil fuel use; and B1, aggressive mitigation. The projected climates were input into the land surface model over 50km x 50km regions with land cover and climate representative of the study area locations. Projected water budget components (including evapotranspiration) were derived for the 21st century (below).

Historical and projected trends in annual actual evapotranspiration and available water supply (precipitation less evapotranspiration) for the Annapolis Valley, Canada.  Symbols correspond to different IPCC GCM while colours correspond to different IPCC climate scenarios (blue= business as usual, red= intensive fossil fuel use, green= aggressive mitigation. Larger image

Results indicate that the IPCC GCM driven projections have substantial variability due to decadal oscillations (decade-long climate cycles) in the model inputs, model differences for a given scenario, and differences between the scenarios.  Results show that the former two sources of variability, which are not directly related to anthropogenic impacts on climate, far exceed the historical variability observed (compare the figures above for the period before and after 2000). Furthermore, for all of the Prairie regions tested, the likely range in evapotranspiration or precipitation less evapotranspiration due to between-model variability was the same or larger than the range due to different scenarios. In other words, the IPCC climate models do not constrain climate projections sufficiently to see the impact of policies related to mitigation on the freshwater cycle in these regions.

Concerns with the low precision of IPCC models are not unique (see chapter 8, “Climate Models and their Evaluation”, IPCC Fourth Assessment Report, Working Group 1). As expected, impact studies have more uncertainty in regions where, historically, climate was at the threshold between regimes (in the Prairies, for example). This leads to the third research question: can uncertainty be reduced in the IPCC GCM projections (especially temperature and precipitation) over northern land regions?

This is a difficult task since changes need to be made to the models to improve their performance under a changing climate for the next century. The simplest strategy is to determine if the sample of 18 IPCC GCMs used to define our ‘likely’ range are all equally valid. Eliminating models from the sample could potentially tighten the ‘likely’ range and reduce projection uncertainties.

Question 3 – Reducing uncertainty in IPCC GCM projections

The IPCC GCMs are an ensemble of models developed from various international and national groups. As such, differences between models are related to what and how physical processes are described and parameterized. It is reasonable to expect that not all models are equal for all tasks in all regions of the Earth. The IPCC recommends the use of relevant historical observations to constrain the models.

Ideally, one would be able to identify models that can predict both temperature and precipitation changes well over Canada. However, there is no guarantee that a model that predicts temperature and precipitation well historically will also do so for future climates. There are two reasons for this. Firstly, historical observations include both the impact of the global climate forcings (such as solar insolation and increased greenhouse gases) and local climate variability (such as El Niño).; Since we are using the GCMs to translate forcings into climate impacts, we need to find observations that are not sensitive to internal climate variability. For this reason, precipitation and snow cover are not good candidates. Secondly, we need historical observations that, if we constrain a GCM to match them, will actually improve the GCMs ability to predict the future climate.  Observed surface temperature is not a good constraint candidate because all GCMs match it within less than 0.5°C but they differ by much larger amounts in their future projections.

As the Earth warms, changes in snow cover can often lead to a decrease in surface albedo (brightness) which in turn increases absorption of solar radiation by the surface and completes a feedback loop by increasing local air temperature. The feedback loop is called the snow albedo feedback (SAF). We see it every year when snow melts - the rate of melt in open areas tends to increase with time. The same process plays out with warming during climate changes – the rate of snow melt will tend to increase every year as climate warms and this will increase warming. All GCMs include this feedback. GCM studies have shown that the SAF explains about half of the change in temperature over northern land surfaces predicted by typical GCMs and that the SAF is relatively insensitive to internal climate variability within the GCMs. The SAF computed from seasonal outputs of GCMs run over historical periods and from projection outputs are strongly linearly related. In other words, if we can observe the seasonal SAF, we could identify GCMs that are not likely to represent the changing SAF (and therefore air temperature) well in projections. This raised two questions:

• What is the historical SAF?
• How do IPCC GCMs compare with it?

Question a: What is the historical SAF?

The SAF can be quantified using daily observations of surface albedo (brightness) and temperature covering the Northern Hemisphere (Fernandes et al. 2009). Since the SAF is only non-zero in areas with snow melt, only the Northern Hemisphere during snow melt periods requires observation. The figures below show the SAF for the Northern Hemisphere during April-May from 1982-1999 (similar results are found for other periods) based on our analysis of satellite-based albedo and in-situ temperature observations extrapolated spatially using a reanalysis model.

Northern Hemisphere spatial pattern of the springtime snow albedo feedback (a) and the standard deviation (variability) among years (d) observed between 1983 and 1999 (units are % per Kelvin). Regions with large feedback (red end of the spectrum) will typically exhibit larger sensitivity in terms of increasing surface air temperatures due to climate warming. Larger image

The GCMs could simply be compared with the observed SAF to identify which GCMs are not likely to work as well; however, this exercise would not allow an explanation of what part of the GCMs is flawed. Therefore, the SAF was broken into two components - one related to how albedo changes with warming when snow is present and the other related to how fast albedo changes with warming during the transition from snow-covered to snow-free conditions. Snow cover was mapped over the Northern Hemisphere (see below).

Right: The mean snowmelt date (Smtd) for 1982–2004 over the Northern Hemisphere shown by 20-day intervals. The white regions represent permanent snow; colours correspond to the average day of the year when the snow starts to melt. Left: Time series of snowmelt averaged over northern Eurasia (EAmtd), North America (NAmtd),and Northern Hemisphere (Gmtd). Dashed lines are the mean Smtd over the period of 1982–2004 as shown by the digital values on the right (Zhao and Fernandes, 2009). Larger image

Comparison of observed (red dot) versus modelled (black symbols) surface albedo feedback broken into components related to snow cover change and components related to snow pack metamorphosis as temperature increases. Shaded bars indicate 95% confidence intervals of observations. Models outside shaded bars are not very likely to match the observed feedback components (Fernandes et al. 2009). Larger image

Unlike the SAF, trends in snow cover exhibit large year-to-year variability; therefore, our short observational record cannot be directly used to constrain GCMs. The figure to the right plots Northern Hemisphere SAF estimates from IPCC GCM model projections versus our observations. It is evident that few models match both the SAF and its two components. This initial comparison used GCM model output from projections, so it is possible that some GCMs may have better agreement with the observed SAF over the historical period. Work is underway as part of an International Polar Year project to address this question.

What next?

• It is possible that some GCMs may have better agreement with the observed SAF over the historical period?  We are currently working within an International Polar Year project to address this question. 
• Can GCMs be adjusted to improve their agreement with the observed SAF? Work is underway with scientists at the University of Toronto and Environment Canada to address this issue.
• Will water cycle projections change if the likelihood is weighted towards models that better match the observed SAF? Knowledge gained from Questions 1 and 2 will be used to refine the modelling work.

Study Data

Canada-wide annual evapotranspiration estimates are available from 1990-2000 from Environment Canada’s RésEau website. The information is useful for general reference but requires local validation before it can be used.

Links

Canada's Water Budget

Publications and References

Please note that subscriptions may be required to access some articles. To request a copy of publications, or for any more information, please contact Richard Fernandes.

• Check for more recent publications in GEOSCAN, the publications database of the Geological Survey of Canada and the Canada Centre for Remote Sensing.

Fernandes, R.A., Butson, C.G., Leblanc, S. and Latifovic, R. 2003. Landsat 5 TM and Landsat-7 ETM+ based accuracy assessment of leaf area index products for Canada derived from SPOT-4 VEGETATION data, Canadian Journal of Remote Sensing, Vol. 29, No. 2, pp. 241–258.

Fernandes, R.A., Koroleivich, V. and Wang, S. 2009. Historical and projected trends in net precipitation within the Canadian Western Prairie Provinces– is there evidence of significant decreases in renewable water supply? Water Resources Research (in review as of April 2009).

Latifovic, R. and Pouliot, D. 2005. Multitemporal land cover mapping for Canada: methodology and products, Canadian Journal of Remote Sensing, 31,  pp. 347-363.

Maidment, D., 1993, Handbook of Hydrology, McGraw Hill, U.S.A.

Rivard, C., Fernandes, R.A. et al. 2009. Assessment of climate impacts to groundwater over Annapolis Valley Basin, Journal of Hydrogeology (in review as of April 2009).

Simic, A. , Fernandes, R.A., and Wang, S. Sensitivity assessment of simulated evapotranspiration and groundwater recharge across a shallow water region for diverse land cover and soil properties, Geoscience and Remote Sensing Symposium, 2004. IGARSS '04. Proceedings. 2004 IEEE International. 09/2004; 7:4881-4884 vol.

Wang, S., 2008, Simulation of evapotranspiration and its response to plant water and CO2 transfer dynamics. Journal of Hydrometeorology, 9: 426-443, DOI:1 0.1175/2007JHM918.1.

Zhang, X., Vincent, L., Hogg, W.D. and Nitsoo, A.  2000. Temperature and Precipitation Trends in Canada During the 20th Century, Atmosphere-Ocean, 38 (3), 395–429.

Zhao, H. and Fernandes, R.A. 2009. Daily snow cover estimation from Advanced Very High Resolution Radiometer Polar Pathfinder data over Northern Hemisphere land surfaces during 1982–2004. Journal of Geophysical Research, Vol. 114, D05113, doi:10.1029/2008JD011272.