«Water Balance Methodology for indirect assessment and prediction of basin water yield under human-induced land use changes N. V. PENKOVA & I. A. ...»
Predictions in Ungauged Basins: PUB Kick-off (Proceedings of the PUB Kick-off meeting held in Brasilia,
74 20–22 November 2002). IAHS Publ. 309, 2007.
Water Balance Methodology for indirect
assessment and prediction of basin water yield
under human-induced land use changes
N. V. PENKOVA & I. A. SHIKLOMANOV
State Hydrological Institute (SHI), 23 Second Line VO, 199053, St. Petersburg, Russia
The lastest versions of the Heat and Water Balance Models (HWBMs) are presented and their applicability to predict evapotranspiration losses from watersheds under changed land use patterns in rain-fed agriculture (crop allocation, crop yield, fertilizer application), and to predict soil moisture availability dynamics during the growing period (for plant productivity esti- mation) are discussed. The versions are based on the integrated “index” water balance approach which presupposes construction of nonlinear relationships between heat and water budget components under different natural and perturbed conditions (soil type, groundwater table position and salinity, plant species and productivity, etc.) using the multidimensional cubic spline technique. The examples on application of the methodology to different conditions are cited for several regions (Middle and Lower Volga River basin, piedmont and lowland parts of the North Caucassian economic region). A comparison is made of the models with several mechanistic heat and water balance models, and combination techniques. The advantages and shortcomings of the models under analysis in respect of their accuracy and practical applicability are highlighted.
Key words evapotranspiration; HWBM; soil moisture regime; Volga River; water balance
BACKGROUNDWith respect to the scale-dependent prediction of man-induced changes in basin water yield, there is a strong demand for studies, not only into river runoff itself, but into runoff generation processes and catchment modelling, since both quantity and quality of surface and groundwater resources depend largely on interaction between surface runoff, soil and groundwater. During the last decades, much has been learned about runoff generation processes and flows along different pathways, especially at hillslope to small catchments scales. Innovation using remotely sensed data, tracer and new hydrometric approaches, and GIS-based computation techniques greatly contributed to a more realistic representation of the underlying mechanisms within process-oriented rainfall–runoff models at the meso and macroscale, and to providing descriptions of the models and their updating procedures in a standard format for disseminating them to all concerned parties.
Modern hydrological models of the distributed system type subdivide drainage basins into a lot of area elements of different scales: the regional climatological scale and different hydrological scales—the choric ones (drainage basin or sub-basin, polygon, runoff forming complex) and the topic ones (sites, plots, hydropedotopes, Copyright © 2007 IAHS Press Water Balance Methodology for indirect assessment and prediction of basin water yield 75 ecotopes, flow strips, elementary areas, hillslope catenas, hydrological response unit, etc.). For the latter, Water Balance Models (WBMs) usually calculate one-dimensional heat and water fluxes with a high temporal resolution, predominately using the statements of thermodynamics of moist air and semi-empirical theory of turbulence.
The regionalization and upscaling may be achieved by different methods: using derivation of location-independent statistical distributions of representative real and “theoretical” elementary areas; by the introduction of the “effective” parameters, by hydrological similarity analysis or aggregation; by scale-dependent pre-processing of input data sets (Dyck, 1985; Vinogradov, 1987; Kavvas, 1989; Wood, 1995; Kalma & Sivaparan, 1995; Diekkruger et al., 1999). Nevertheless, the usefulness of the models as instruments for predicting the impact of man-induced climate and land-use changes remains questionable, due to a range of peculiarities of the hydrological cycle transformations, both within the spatial and temporal domains.
Within the spatial domain, the errors may arise from disregarding the dependencies among soil type, surface slope and aspect, and land use. In a majority of hydrological models, the area elements are considered as natural associations of geology, geomorphology and soils. Vegetation cover can also be derived from general thematic and special schematic maps, and by remote sensing. But the land use information from the latter is classified, usually into very rough classes (urban areas, coniferous forest, deciduous forest, lakes, grassland, agricultural areas), although a more differentiated picture is required. For example, the crop models, to be included into the general catchment models, at least require characteristics such as plant type, plant age and height, nutrients, fertilizer or pesticides availability, to be able to calculate evapotranspiration losses with desirable accuracy. In real watersheds, the spatial inconsistency is usually being observed between natural land surface elementary units and the land use patterns. The latter depends to a large extent on socio-historical conditions and on economic considerations that can cause additional problems to the interpretation of land use maps and images for hydrological modelling.
Within the temporal domain, no less great difficulties exist due to the necessity to incorporate interactions between the rates of land-use changes and degradation processes and the changing soil and vegetation environment responding to both the degradation and external forcing (climate). There exist several approaches in progress, which incorporate the dynamic interactions to simulate possible scenarios over periods of several decades, and to provide some forecasts of the impact of alternative policy options in different natural conditions (Kirkby et al., 1996; Rodda et al., 1996; etc.).
The most significant advances of the approaches consist in using mainly publicly available data in the development of hydrological models within Decision Support Systems (DSS), in order to be able to present results which can be used directly by policy makers and managers, and in making the DSS an adaptable tool, with the ability to incorporate additional models when they become available.
But in a majority of hydrological models the runs are carried out using timeindependent soil and vegetation conditions since the advantages, in simplicity, are thought to outweigh the rather slight advantages of accounting for the relatively small man-induced water balance transformations, compared to the uncertainty of rainfall– runoff modelling at the larger scales. At micro and mesoscales, the impact of the transformations may be rather significant, but the difficulties of model calibration N. V. Penkova & I. A. Shiklomanov 76 remain, because to specify the land use of agricultural areas, the statistics at the county or district level should be used but usually no spatial data allocation is available, only percentile data that need an extensive pre-processing effort under the modern crop rotation practice. Due to the uncertainties and inevitable assumptions and speculations, most models of river hydrology remain conceptual and “semi-quantitative”. They usually include modules of “black-box” or “grey-box” type.
Now the tendency is being observed to larger abstractions and to simplification of the model structure. In that approach, the within-basin areas, which are particularly important, are modelled in detail, whereas the less important ones are lumped together by a simplified statistical approach. The calculation of the input parameters for various consolidated areas takes place by weighting them so that dominant basin or regional structures are emphasized. At the present time, the approach seems promising, bearing in mind the necessity to join the very different paradigms of disciplines associated with the environmental domain, and those of social, economic and management disciplines, which can be characterized by holistic and problem-solving modes of thought aimed at directly influencing policy and practice. At present, the vulnerability of economic systems is furthermore strengthened due to the set in of the “epoch of limits”, i.e. the situation when effectiveness of human activities in the production sphere is being limited by natural potentials and therefore, the systems are more sensitive to natural perturbations. In general, the problem resolves itself into “making better use of what exists” in all spheres of human activity. In a number of the most advanced studies into natural, semi-natural and artificial environmental objects, the role of humans in their functioning and effectiveness is considered, and development of knowledge is focused on creation of sufficiently rigorous and relatively easy-to-use applied models which do not break down in the absence of an inordinate high degree of site-specific information. The main emphasis is on “subject–object” interactive methodology based on Information Society Technologies, on individualization (personalization) of research, and on wide participation of experts, decision makers and all other concerned parties in the knowledge development.
The Water Balance Models (WBMs) this paper deals with, may be qualified as deterministic ones based on the macroscopic version of the equation of continuity when heterogeneous processes of the hydrological cycle are lumped over finite time
intervals and areas, and related by the general equation:
P + Rsi + Rui = E + Rso + Ruo + Rs + Rg + ΔS + η (1) where P is precipitation; E is actual evapotranspiration; Rsi, Rui are net surface and subterranean natural and man-induced water diversions from other basins; Rso, Ruo are net surface and groundwater natural and man-induced transfer to other basins; Rs, Rg are surface and subterranean runoff; ΔS is change in total storage; η is an error term.
Various versions of the equation exist since at different scales the importance of different hydrological cycle processes may change significantly. The modelling usually concentrates on the most important processes and on the appropriate Water Balance Methodology for indirect assessment and prediction of basin water yield 77 representation of them, and depends on current understanding of physical and other processes, on the data available and the purpose chosen, so there is no WBM which is suitable for the whole range.
The central idea of the water balance methodology is to avoid difficulties regarding the inadequate process understanding both at finer and at coarser scales and the specific problems of upscaling, through fulfilment of analysis and linking balances for different basin zones and separate water objects, in order to reveal systematic errors and miscalculations in the components’ determination, and to define the values of separate non-explored components (as residual terms). The latter is of great importance, since with the exception of river runoff, which is an integrated measurement, for all the water balance components, the problem of areal representation of point measurements exists, and a majority of methods for their determination have been developed using findings of the “site science” (Andreyanov, 1977; Penkova, 1984;
Dyck, 1985). For example, the mass calculations of water balance for small and medium size watersheds in the Middle Volga region, even within the well nested models, have shown that η−values may be comparable to the balance components (Table 1). In this case, the three store WBM for small river basins was under
P + Rsi = Ep + Rs + ΔSs + Ia (2) Ia = Ew + ΔSa + Ig (3) Ig + Rui = Eg + Rg + ΔSg (4) where Ep, Ew are evapotranspiration from surface zone of basin and from vadose zone;
ΔSs, ΔSa, ΔSg are change in surface water storage (in snow cover, glaciers, in local depression areas, in lakes and reservoirs, in stream channels), change in water storage in vadoze and saturation zones; Ia, Ig are filtration into vadose zone and groundwater recharge (Andreyanov, 1977; Penkova, 1984). The P, Ep, Ew, Rsi, Rui, Rs, Rg, ΔSs, ΔSa, ΔSg-values are determined independently, Ia, Ig and η − as residual terms for a conformable zone.
According to the most generalized classification, the WBMs may be qualified as system models belonging to the “grey box” structural form, i.e. the intermediate one between the “black box” ones with multiple inputs and a single output (river runoff), and the mechanistic models with defined intrinsic model structure. In the basin WBMs, N. V. Penkova & I. A. Shiklomanov 78 usually the structure of the basin body (or its part) and the structure of the water cycle processes are discerned. In part these structures correspond to each other in a definite manner. For example, a term such as inflow refers to a part of the vadose zone above the first impermeable layer, while the baseflow is usually the saturated zone. It is possible to speak about compliance between the spatial structure of precipitation and evaporation fields, as well as of other components, and the basin structure (dependence of the hydrological cycle process on morphography and geology). But the intrinsic structures are proved in the spatio-temporal distribution of the water cycle elements (including diurnal, seasonal and multi-annual cycles).