Presentation on theme: "CLIMATE AND LANDSCAPE CONTROLS ON SPATIAL (SUB-BASIN) VARIABILITY OF WATER BALANCE WITH CHANGING TIME SCALES: APPLICATION OF THE DOWNWARD APPROACH D. Farmer."— Presentation transcript:
CLIMATE AND LANDSCAPE CONTROLS ON SPATIAL (SUB-BASIN) VARIABILITY OF WATER BALANCE WITH CHANGING TIME SCALES: APPLICATION OF THE DOWNWARD APPROACH D. Farmer and M. Sivapalan Centre for Water Research, University of Western Australia
The south-west Western Australia has a semi-arid, Mediterranean climate with cold, wet winters and hot, dry summers. Both the climate and the landscape features are highly variable over relatively short distances (i.e.. 50 to 100km). Except for a few water resources catchments the landscape is sparsely gauged for rainfall and streamflow. This poster deals with the application of a “downward” or top-down approach to a systematic analysis of the space-time variability of observed streamflows around a 7,000 km2 basin. The method is used to discover the key climatic and landscape (soils, topography and vegetation) controls on this variability, with a view to eventually making predictions for ungauged sub-catchments within the basin. These predictions are ultimately used to describe patterns of within-basin spatial and temporal variability. First the systematic analysis is carried out for a number of representative, gauged catchments in and around the basin of interest. In the first instance this involves the analysis of rainfall-runoff data to assess general behaviour differences in signatures of streamflow variability at various timescales (i.e., inter-annual, intra-annual or mean monthly, flow duration curves).. A series of conceptual, storage based, water balance models of increasing complexity with decreasing timescales (annual, monthly daily etc.) are then used to investigate the ability of the models to reproduce the various signatures. The required parameters for the models are estimated from available landscape information and recession data. The combination of model complexity and sensitivity analysis on key parameters provides feedback that is subsequently used to conduct “first pass” distributed modelling. Results demonstrate significant east-west trends driven by climate variability, and systematic and random variations caused by soils, topography and vegetation, including land use. Abstract
Murray Basin, Western Australia Perth Avon Blackwood Murray Collie Albany Soils / landscapeVegetation Murray River basin (SW-WA) at Dwellingup 6840 km2 Two major upstream catchments: Hotham 4500 km 2, Williams 1700 km 2 Many smaller tributaries ~60-200 km 2 Dwellingup area – high rainfall, natural Jarrah landscapes, deep laterite profiles, mature pine plantations. P= 1000 mm/yr Ep = 1500 mm/yr Eastern portion – low rainfall, shallower soils, less relief, large scale clearing for agriculture, small pockets of remnant veg. P = <600mm/yr Ep = 1600-1800 mm/yr Notional east-west changes in soils, landscape, rainfall and vegetation.
Introduction Initial analysis of river gauging stations on the Hotham, Williams and Murray Rivers identified behaviours that were not easily explained at catchment scales. Why does the drier Williams catchment have greater yield than the wetter Murray catchment (Fig A and B) ? How does increasing dryness to the east impact upon the 4,000 km 2 Hotham catchment (Fig D) ? What role does the decreasing soil storage capacity play (Fig E) ? It was also found that the catchments within the Murray Basin were generally drier than catchments from Western Australia and around Australia where the downward approach had been previously applied (Fig C). A B CDE
A BC ED Analysis of key signatures Observed data from nine representative catchments inside of and adjacent to the Murray Basin (Fig A) were used to produce derivative datasets. Data for five of these are shown. Name colours correspond to colours used in the figures. The remaining four had behaviours similar to ‘Dwellingup’ and ‘South’. Comparisons were made of annual (Inter-Annual variability, Fig B), average monthly (Intra-annual variability, Fig C), flow duration (Fig D) and recession (Fig E) behaviours. The data identified subtle differences between the test catchments, these included: Yield differences between catchments North and South, and Dwellingup and East. Some variability in Cuballing catchment (Figs B, C and D). A notable lag in the Dwellingup catchment (Fig C). Recession analysis identified a general trend in ‘a’ values (Fig E). Higher values for Dwellingup were confirmed by nearby catchments (purple line in Fig. E). Conclusion: some general trends existed but data alone could not explain all variability.
Representative catchment modelling analysis Farmer et al (2003) used the model framework described below on two catchments within the Collie Basin immediately south of the Murray. Catchment ‘South’ was one of these. The approach had also been previously run on Catchment ‘North’ for another project. The model was first run using mean basin values for soil depth (pink lines). These showed that local variability could not be explained by climate alone. Models were then run using ‘best estimates’ parameter sets. Model sensitivity analysis (green lines) indicated that the prediction of general catchment behaviour was not substantially impacted for small variations to realistic model parameters estimates. Independent testing on remaining gauged catchments showed that while predicted data may not be perfect they did capture the local behaviour. Predictive uncertainty was smallest for longer timescale signatures and this gave confidence in the ability of the models to predict mean annual and seasonal patterns using interpolated parameter sets. Comparison of results from Yarragil (‘Dwellingup’ catchment) and Falls farm (‘Cuballing’ catchment) identified differences. The deeper soils for ‘Dwellingup’ tended to promote a long wetting up and late winter flows. Summer depletion was exacerbated by evaporation from the native forest. The shallower soils of ‘Cuballing’ and lack of substantive vegetation meant that less rainfall was needed before parts of the catchment became wetted and runoff began to occur. Baseflow contributions were required before low flows were captured. Realistic parameter estimates were needed to capture these behaviours. Similar outcomes were identified for ‘North’ and ‘South’.
Predicted long-term average yield Individual model runs were made on the 70 sub-catchments. Soil and recession parameters were interpolated from key data points. Measured vegetation cover was used. Rainfall data for each sub-catchment was derived from 31 well distributed daily rainfall gauges. Evaporation data was derived from long term pan estimates (see Farmer et al 2003). Model data was produced for a 10 year period from 1980-1989. The resulting data highlighted the role of vegetation in evaporating water and reducing net runoff. Cleared catchments in the wetter western portion of the Williams catchment were found to produce the highest yields (Q/P). Catchments in the eastern areas were found to be drier. Since all sub-catchments are of equal size it is possible to assess net contribution to total basin yield. The long-term averages tend to reflect the eastward decline in annual rainfall. While they provided some indication of spatial variability they did not serve to explain observed differences at smaller temporal scales.
Seasonal behaviour and runoff generation Aggregation of sub-catchment predicted data permitted water balance signatures to be generated for the Murray, Hotham and Williams catchments. These showed a reasonable reproduction of the observed behaviours, including flow (Fig A and Fig B). Comparison of intra-annual behaviour between the western and eastern parts of the basin shows the lag seen in the earlier analysis (Fig C and Fig D). This lag reflects the longer period of wetting up associated with the deeper soil profiles of the Darling Range. In the shallower soil profiles runoff occurs more closely to winter rainfall patterns. Bi-monthly analysis of rainfall and yield over the critical June- July wetting up period provided a visual summary of the situation (Fig E and Fig F). While rainfall totals are typically higher in the western parts of the basin, yields are low. A B E F Darling Range Eastern Murray Basin CD Blue = observed data, red = predicted Predominantly cleared catchments immediately east of the forested catchments and in the eastern basin areas are quite active, while areas of remnant vegetation are predicted to be lower yielding. Further investigation has shown that due to lower soil capacities, moderate rainfall and slightly lower evaporation the Williams River is most likely to respond to a larger storm event. This has implications for flood risk assessment.
Background to the Method Farmer et al (2003) introduced a model framework that can be used in a parsimonious manner to assess the impact of climate and landscape interactions upon key water balance signatures (Figure). The parameters associated with the model framework could be related to identifiable catchment characteristics (see Table right). Values could therefore be subjectively estimated from available information and spatially distributed to reflect landscape transition. This study uses the model framework to examine sub-basin impacts resulting from spatial variability of key parameters. Interaction between climate inputs and gradually changing parameters across the basin causes subtle changes in process dominance and thus an observable variation in behaviour. Parameters were estimated for reference sub-catchments from soil- landscape reports for the Murray Basin and some catchments. Vegetation cover was estimated from satellite imagery. Recession parameters were adopted from the recession analysis. This knowledge was then spatially distributed in order to derive indicative parameter sets for each of the 70 sub-catchments. Model combinations are then run for each sub-catchment in order to obtain output data that can be used to examine spatial and temporal variation across the basin. Reference: Farmer, D., M.Sivapalan and C.Jothityangkoon, 2003, Climate, Soil and Vegetation Controls upon the Variability of Water Balance in Temperate and Semi-Arid Landscapes: Downward Approach to Hydrological Prediction, Water Resources Research, 39(2)
. Summary Data and model analysis of the various temporal signatures at five sub-catchment locations established key differences between landscapes within the basin. Independent testing on remaining catchments showed that while results may not be perfect local behaviour was adequately captured. Subsequent distributed modelling made it possible to predict the general (large scale) trends that cause detectable basin scale impacts. The study identified that variability within the Murray Basin resulted from changing interactions between soil water storage capacity, rainfall and evaporation. Though model structure remained essentially similar, small changes in parameters and process significance contributed to observed spatial and temporal variability. Responses were significantly altered in the presence of vegetation. Variations in behaviour were identified between the vegetated deeper soil catchments in the west, the shallow drier catchments in the east and the shallow active catchments in the south-east. Of note is the way that different parts of the basin transition from dry hot summer states to peak wetness during winter-spring. These trends impact upon runoff generation, streamflow contribution and space-time patterns of catchment wetness. Using the downward approach provided: 1.Confidence that parsimonious model configurations could capture the essential behaviour. 2.Insight into local process and parameter significance. 3.A means to assess model sensitivity to landscape derived parameters and parameter change. 4.The ability to incorporate understanding from outside-of-basin gauges. 5.The means to derive the maximum benefit from limited basin datasets before making predictions. 6.Support for decisions needed to make spatially distributed predictions for ungauged catchment areas.