3. HYDROSPHERE

3.1 Introduction

This chapter is divided into three parts: (1) hydrological observations required to understand the impact of climate change on hydrological regimes' feedback to climate; (2) hydrological observations required for GCMs; and (3) hydrological observations required for biospheric modelling. The first three sections are not meant to be mutually exclusive and the reader will notice many variables are repeated in all three sections. These three sections lead to the selection of a minimum set of variables that are to be included in the Initial Operational System (IOS). The reader is referred to Annex I for details on each variable, including a rationale, and the required temporal and spatial resolutions.

3.2 Understanding Climate Impact on Hydrological Regimes and Feedbacks to Climate

The hydrologic cycle is intrinsic to the Earth's climate system. The processes by which water, in all its phases, moves through the atmosphere, and moves to and from the various repositories on the Earth's surface are inter-meshed with those governing the Earth's energy budget (Brutsaert, 1984). However, the processes are not adequately understood and are the subject of current research, e.g., in various Global Energy and Water Cycle Experiment (GEWEX) activities. The quantity and quality of water affects human society, both directly and indirectly through its control of the biosphere functioning. The availability of water is a major controlling factor in the distribution and abundance of vegetation and of biological productivity.

One of the most significant consequences of climate change will be a geographical shift of regional hydrological regimes, and associated changes in the availability of water resources which are a critical factor in planning for economic and sustainable development. In turn, changes in the hydrological cycle will create feedbacks to the global climate system. Our ability to characterize the current state of the hydrological cycle, globally and regionally, and to make quantitative predictions about the nature of potential changes depends on consistent, quality information with adequate spatial and temporal coverage. Such hydrologic data sets are currently lacking for many regions of the Earth.

Availability of adequate water of appropriate quality is a major factor in planning for sustainable development. Consequently, studies providing evidence that both water quality and quantity can be reduced by climate change raise valid social concerns (Revelle and Waggoner, 1983). Gaining an understanding and knowledge of the water and energy budgets are critical first steps in detecting, assessing and predicting the global climate change. Unlike data relating to radiance or surface thermal regimes such as albedo and surface temperature which are fairly readily obtainable from satellites, many observations relating to the hydrological cycles require in situ measurements. Further, many hydrological data are collected locally and/or nationally and are not adequately shared internationally.

To understand the interactions between climate and the hydrological cycle, it is necessary to understand the relative distribution as well as fluxes of water among its various repositories, as well as the basic composition of the atmospheric and terrestrial water balance (Rhodes, et al., 1994). The approximate distribution of the available water is as follows. About 97.3 % of the Earth's water supply is in the oceans. Most fresh water is captured in ice sheets, ice caps and glaciers (2.05% of the total). The 0.7% remaining includes 0.68% as groundwater, 0.01% in surface lakes, ponds and man-made reservoirs, 0.005% characterized as "soil moisture", 0.001% in the atmosphere, 0.0001% in rivers and approximately 0.00004% bound up in the biosphere (Berner and Berner, 1987). The rate of exchange within these reservoirs ranges from thousands of years for the oceans and ice sheets, to hundreds to thousands of years for groundwater, to years for surface impoundments, to months for soil moisture, and days for atmospheric, river channel and biospheric reservoirs (Nance, 1971; L'Vovich, 1979). It is significant that the smallest and most quickly exchanged repositories are those of most importance to humans and potentially most responsive to changes of climate.

Whereas a regional atmospheric water vapour balance can be summarized as:

E - P = Q + D

The fundamental relationship for the terrestrial water balance is:

P - E = R + D S_GW + D S_SM + D S_SS + D S_B

where:

E represents evapotranspiration;

P represents precipitation;

Q represents the precipitable water in the water column;

D represents horizontal divergence (convergence), that is, water transported out of (into) the regional water column;

R represents runoff (in-channel flows due to atmospheric contribution);

D S_GW represents the change in ground water storage;

D S_SM represents the change in soil moisture storage;

D S_SS represents the change in surface storage, such as lakes, reservoirs and wetlands (technically, D S_SS can also include storage of water in its solid phase, such as glaciers or seasonal snow mass, but these will be treated here as components of the cryosphere); and

D S_B represents the change in water in biomass storage.

There is a complementarity between the atmospheric and terrestrial phases of the hydrological cycle. It should be noted that for the terrestrial phase, the global water balance has sometimes been written as simply:

P - E = R

by assuming that in the long term, an equilibrium condition exists and there is no net change in the other terrestrial reservoirs (Baumgartner and Reichel, 1975; Korzun, 1978). This abbreviated relationship draws attention to the fact that discharge, a quintessentially hydrologic phenomena, can serve as an integrating phenomenon, summarizing the behaviour of two highly variable (in space and time) parameters, precipitation and evapotranspiration. However, if one allows for the possibility of climate change rather than just long-term climate variation, this simplified depiction of the water balance is inadequate to study hydrospheric balance.

All of the variables of the terrestrial water balance equation are affected by, and serve as, feedbacks to the climate system, and consequently should be monitored if an assessment of climate variation over some interval is to be made. However, several of the variables such as E, D S_SM and D S_B are very dynamic on short temporal scales and very variable on small spatial scales, so that the choice of a time step or spatial aggregate to be monitored is determined by the impact being studied. Information on these three very important hydrologic and climatic variables is generally not available on any long-term or spatially large aggregate basis. Whereas measurements of E and S_SM are frequently found in seasonal watershed or regional agricultural studies, information regarding S_B is generally infrequent and only an estimate. These three variables should be considered as potential enhancements in any long-term regional or global monitoring programme, although there is currently no consensus on how to make measurements or how to aggregate measurements once made.

Because of its rapid atmospheric exchange rate, precipitation is potentially a highly sensitive indicator of climate variation and change. It is the only input in the water balance of the Earth's surface (evapotranspiration and runoff represent the main losses). In addition, it plays an important role in the energy cycle of the atmosphere due to the release of latent heat associated with the condensation of water vapour. The 1990 IPCC assessment noted that precipitation appears to be increasing outside of tropical areas, with a tendency for decreases in the sub-tropics. However, the highly variable nature of precipitation and the inadequacy of global precipitation networks make it difficult at present to characterize long-term trends.

Evapotranspiration is the process by which water is returned to the atmosphere. Over land surfaces, it is difficult to distinguish contributions obtained by purely physical evaporation processes and those due to transpiration by living plants, hence the term evapotranspiration which combines the contribution from both sources. Because of its highly variable spatial and temporal nature, evapotranspiration is a difficult parameter to measure (Oliver, 1983). Methods of measurement range from the use of evaporating pans and lysimeters, to mass budget estimates based on terrestrial regional water balance, to calculations based on energy budget derivations (Brutsaert, 1984; Shuttleworth, 1993), to flux measurements above the canopy (Baldocchi, et al., 1988). Although collections of evaporation data have been published in the past (e.g., Farnsworth and Thompson, 1982) and in spite of recent intense international interest in evapotranspiration measurements, including remote sensing- based methods, (e.g., Brutsaert, et al., 1988), no one method of measurement or aggregation is currently agreed on for regional or global studies.

The net balance between precipitation and evaporation is reflected in runoff and change in the storage values. Because stream flow is an integrating phenomena, it may in some specific cases provide an amplified, early-warning signal of climate change. Thus, the climate change detection objective needs to emphasize large rivers draining major continental areas and long time series, preferably at sub-annual scales sufficient to study seasonal variation. It also stresses the need for data from smaller but near-pristine river basins because of the difficulties in separating climate-induced change from other simultaneously occurring changes due to land- and water-use developments (Slack and Landwehr, 1992).

One of the key terms in the global liquid water balance that has generally not been taken into account is the contribution of groundwater to the world's oceans, either directly at the land-ocean interface or indirectly, due to anthropogenic mining of groundwater from aquifers with minimal recharge. Indeed, it has been suggested that the contribution of groundwater to the world oceans over the last century has been a significant factor in sea level rise (Sahagian, et al., 1994), at least as large as that attributed to the melting of mountain glaciers.

Soil moisture is a critical variable not only in theoretical studies of global water balance (Milly, 1994) and global biogeochemical cycling (Raich, et al., 1991; Rastetter, et al., 1991) but also in analyses/studies related to commerce and farming (Sonka, et al., 1992). It is intimately related to evapotranspiration rates and, like evapotranspiration, is highly temporally and spatially variable. Some progress has been made in studies to develop remote sensing methods to measure soil moisture, but soil moisture data sets are generally only available for spatially and temporally limited studies.

The stage and extent of large lakes and reservoirs can serve as useful indicators of climate change. They have a tendency to filter out short-term variability and respond to long-term changes in the hydrological cycle. Man-made reservoirs may be constructed to diminish extreme climate effects on the society (Thomas, 1959). The growth in the numbers of reservoirs and incorporation of natural lakes probably has had a dampening effect on sea level rise (Newman and Fairbridge, 1986). Recently, it has been suggested that globally, reservoir surfaces may serve as sources or sinks for two greenhouse gases, carbon dioxide and methane (Kelly, et al, 1994). Hence, some accounting should be made of the temporal fluctuations of surface water pondings, especially in hydrologically closed areas. Remote sensing is able to provide information oh ponded areas at an appropriate spatial resolution but in situ observations are necessary to assist with issues such as distinguishing open water from wetlands of different types, estimating the volume of water involved, and tracking seasonal occurrences.

Biomass water storage is a conceptual parameter for studying the impacts of climate change and is necessary to bring closure in regional and global water balances, albeit only a tiny fraction of the global water balance. While it is practically impossible to measure directly, it can be estimated through the water balance equations. It is closely tied to questions of land cover and vegetation dynamics over a region, and approximate estimates may also be obtained from the knowledge of biomass, leaf area index, or greenness (Tucker, et al., 1985). This variable is mentioned for the sake of completeness only, it is not proposed to measure this variable at this time due to the measurement difficulties involved.

Rivers are the major means of transport of nutrients such as carbon, nitrogen and phosphorus from land to ocean and play a significant role in the sulphur cycle as well. Sediment loads can affect climatically significant ocean processes, but are also indicative of terrestrial changes. Concerns pertaining to the quantification of pollutant effects and loss of biodiversity, whether or not directly reflective of climate change, make sediments a significant variable. Thus, an additional factor - movement of sediments and concomitant biogeochemical substances such as nutrients, carbon and heavy metals, transported from the land by rivers to be measured at the mouths of major rivers - has been added to the list of variables suggested for observation for two reasons: this variable is relevant both to water balance and to coastal zone ecology questions, and it is an integrating variable which is related to climatic conditions.

3.3 Hydrological Variables Required for GCMs

A distinction must be drawn between numerical weather prediction (NWP) models and GCMs. NWP models predict a future state of the atmosphere on short time scales (hours to days), based on a set of initial conditions. The initial conditions for the model runs are derived from parameter fields incorporating assimilated observed data at the initial time. Validation is carried out by comparing model predictions of the atmospheric state at the final time with observations.

GCMs, on the other hand, simulate the evolution of the atmosphere over longer time scales, years to decades. Because the atmosphere is a chaotic system, predictions of actual future instantaneous atmospheric states cannot be made with any confidence over periods longer than a few days. Validation is carried out by comparing model output fields with multiple year averages of observed fields (usually summarized as monthly averages). These must be provided on global grids, at resolutions equal to or greater than the model grid. At present, this is generally quite coarse for GCMs in the horizontal; however, in order to fulfil future needs such as regional scale modelling, observational data will probably be required within the decade on scales of the order of a few (5-50) km.

Large-scale features of the free atmosphere can be validated using NWP model outputs obtained by four-dimensional data assimilation of observed variables. However, the terrestrial and near-surface assimilated fields (surface fluxes, soil moisture, screen air temperature, etc.) are not useful for validation because they are heavily influenced by model parameterizations of unresolved, sub-grid scale processes and are therefore highly model-dependent. Validation of model-generated surface fields must therefore be undertaken on the basis of remotely-sensed data and surface observations. Although numerous variables are useful as inputs or for model validation, those of special significance for modelling are listed below.

Precipitation drives the land surface hydrology. It is, unfortunately, also one of the most difficult variables to measure in a spatially comprehensive manner. On an annual time scale, the correct simulation of discharge together with precipitation, provide indirect validation of the modelled surface evapotranspiration rate. The collection of discharge data also presents fewer problems than that of evapotranspiration, which is both difficult to measure and highly spatially variable. To complement the existing databases, however, there is a requirement for accurate runoff estimates by major river basins. Important steps would be the provision of a global digital map of basin boundaries for the catchments being monitored; and a series of corrections to the raw discharge data, describing the amount of water removed from the river for human use (irrigation, municipal water supplies, etc.).

The ongoing collection of snow extent data using visible satellite imagery provides useful first-order validation data for snow modelling in GCMs. However, the actual modelled variable is snow water equivalent. A rigorous validation of the performance of GCMs with regard to the simulation of snow requires the development of a long-term database of snow water equivalent.

Climate simulations, unless multi-decadal, can be markedly sensitive to initial soil moisture, particularly for deep soils. Ongoing efforts to produce global soil moisture climatologies are therefore important, despite the problems of spatial heterogeneity. This item is given low priority currently, not because it is unimportant, but due to the severe difficulties associated with its measurement. An alternative to direct observations might involve the modelling of soil moisture on a continuing basis, using a high quality soil database, updated meteorological observation (e.g., from the WWW) and constraints based on stream flow records.

To enable the realistic modelling of surface fluxes, the characteristics of the land surface must be specified in such a way as to reflect as accurately as possible ambient conditions. Information is therefore required on land cover, vegetation characteristics, extent of water bodies and wetlands, and soil parameters.

3.4 Hydrological Variables Required for Biospheric Models

Since biological systems are tightly coupled to water, most ecological models related to climate change require as input hydrological data or outputs from hydrological models. The models described below are intended to demonstrate the types of hydrological data required by these biological models. A more detailed discussion of the biological models themselves is contained in Chapter 2.

Climate change is likely to have significant impacts on stream flow and water resources (Kaczmarek, et al., 1996, Arnell 1996), and these in turn control biological processes. Global biospheric priorities for hydrological data are somewhat different from those which hydrologists define. For example, biospheric models need the areal gridded runoff estimates, not data from river mouths. Also river carbon and nutrient transport from land surfaces are needed by global ecologists. Biogeochemical data are not commonly taken at standard hydrological stations. Biological models also require accurately defined spatially gridded precipitation and runoff data.

While biological models need soil moisture to a depth of 1-2 metres, knowledge of surface wetness can be useful. Surface energy partitioning reacts most quickly to surface (as opposed to soil profile) wetness. Also, nutrient decomposition, trace gas flux and seed germination are ecological processes whose modelling would benefit from surface wetness if available at high temporal frequency. Surface wetness observations are required for various ecosystem types although it appears that satellite-borne sensors will not be able to detect surface wetness under thick or high-biomass vegetation canopies.

Most of the biological models described below and in Chapter 2 compute evapotranspiration at daily to monthly steps. Recent technological advancements now make the concept of a global tower flux network for continuous measurements of water and CO₂ fluxes possible. Although only a limited number of sites can presently be supported by this technology because of the costs involved, even 10 well-distributed sites of continuous water vapour flux measurements would be immensely valuable for calibrating and validating global biospheric models.

Relative humidity is another critical variable in many biogeochemical models because of its importance to carbon assimilation rates. The derivation of radiation and humidity data from the original precipitation and temperature observations needs to be done with consistent logic and algorithms. In several respects, the global four-dimensional data assimilation forecasting models are the best current methodology for ingesting daily observations, error checking, spatial extrapolation, and derivation of variables at the appropriate scales of resolution for use in biospheric models. The IGBP Biospheric Aspects of the Hydrological Cycle (BAHC) Project's Focus 4 (Weather Generator) is exploring these issues.

The importance of extreme but infrequent hydrological and meteorological events is critical to the ecosystem response. For example, minimum temperatures, particularly frost events, have profound control over vegetation. Major flooding events or particularly intense periods of precipitation are also important. It is crucial that derived and archived data must not be aggregated if such aggregation results in removing extreme events from the record.

Numerous applications of hydrologic models, based on empirical relationships between measured temperature/precipitation and catchment runoff, have been used to evaluate climate impacts on stream flow. Revelle and Waggoner (1983) used an empirical model to evaluate discharge from several catchments in Arizona, USA, and found that a 2K air temperature increase with a 10% decrease in precipitation would result in a 40% reduction of catchment runoff. Lettenmaier and Sheer (1991) used a conceptual hydro-logic model to evaluate climate change effects on stream flow in several California, USA, watersheds and found that snow melt would occur much earlier and that the shape of the annual hydrograph would be altered, increasing the chance of winter flooding and reducing spring and summer stream flow.

The models used in such analyses are based on a description of physical processes, and are calibrated to historical data. They do not account for, or predict, the distribution of soil moisture, infiltration, evaporation or runoff over the basin. However, they may estimate evaporation as a function of temperature and available precipitation. This latter type of model is calibrated to the climatic conditions of the data record, and should not be extended to conditions outside the range of the calibration data.

The most appropriate modelling approach for climate change analysis is the water balance model (Gleick, 1987) because it can be used to account for the effects that variation in climate parameters may have on the hydrologic processes represented in the water balance equation. It thus allows assessing the sensitivity to climate variation of hydrologic processes such as infiltration, evaporation, snow melt, and runoff. The water balance approach also provides a flexible basis for model development because it accommodates varying degrees of complexity in the terms of the equation representing individual hydrologic processes, from simple empirical parameterizations to fully mechanistic representations. This allows the staged development of a model depending on application and available data.

The basic approach to water balance modelling was developed by Thornwaite (1948) to evaluate the importance of different hydrologic parameters under a variety of hydro-climatic conditions. Standard methods for calculating water balance developed during the past 50 years (Dunne and Leopold, 1978) have been applied generally to point data, or used in a limited way in catchment (Gleick, 1987) or zonal models (Beven and Kirkby, 1979; Hornberger, et al, 1985). Beven and Kirkby (1979) used a simple water balance approach in their development of a zonal model, which was improved by Beven, et al., (1984) for catchment scale simulations, and for a more detailed treatment of soil moisture by Famiglietti, et al., (1992). However, these models are limited to catchment scale, and do not include vegetation interaction in the water balance calculation.

Table 3.1 summarizes the minimum set of hydrological variables that are required to detect climate change, assess the impacts of climate change, predict seasonal to interannual climate and to simulate long-term climate changes.

Table 3.1 - Summary list of the key hydrological observations required for climate purposes. (Details on each variable are given in Annex I)