Science of the Total Environment 360 (2006) 158 179 www.elsevier.

com/locate/scitotenv

Modelling of recharge and pollutant fluxes to urban groundwaters

Abraham Thomas a, John Tellam b,*
b

Department of Earth Sciences, University of the Western Cape, P Bag X17, Bellville 7535, Cape Town, South Africa Hydrogeology Research Group, Earth Sciences, School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham B15 2TT, UK Available online 1 December 2005

Abstract Urban groundwater resources are of considerable importance to the long-term viability of many cities world-wide, yet prediction of the quantity and quality of recharge is only rarely attempted at anything other than a very basic level. This paper describes the development of UGIf, a simple model written within a GIS, designed to provide estimates of spatially distributed recharge and recharge water quality in unconfined but covered aquifers. The following processes (with their calculation method indicated) are included: runoff and interception (curve number method); evapotranspiration (PenmanGrindley); interflow (empirical index approach); volatilization (Henrys law); sorption (distribution coefficient); and degradation (first order decay). The input data required are: meteorological data, landuse/cover map with event mean concentration attributes, geological maps with hydraulic and geochemical attributes, and topographic and water table elevation data in grid form. Standard outputs include distributions of: surface runoff, infiltration, potential recharge, ground level slope, interflow, actual recharge, pollutant fluxes in surface runoff, travel times of each pollutant through the unsaturated zone, and the pollutant fluxes and concentrations at the water table. The process of validation has commenced with a study of the Triassic Sandstone aquifer underlying Birmingham, UK. UGIf predicts a similar average recharge rate for the aquifer as previous groundwater flow modelling studies, but with significantly more spatial detail: in particular the results indicate that recharge through paved areas may be more important than previously thought. The results also highlight the need for more knowledge/data on the following: runoff estimation; interflow (including the effects of lateral flow and channelling on flow times and therefore chemistry); evapotranspiration in paved areas; the nature of unsaturated zone flow below paved areas; and the role of the pipe network. Although considerably more verification is needed, UGIf shows promise for use: in providing input for regional groundwater solute transport models; in identifying gaps in knowledge and data; in determining which processes are the most important influences on urban groundwater quantity and quality; in evaluating existing recharge models; in planning, for example in investigation of the effects of landuse or climate change; and in assessing groundwater vulnerability. D 2005 Elsevier B.V. All rights reserved.
Keywords: GIS; Urban; Aquifer; Recharge; Pollutants; Groundwater; Birmingham; Urgent

1. Introduction Urban aquifers are of considerable importance to the long-term viability of many cities across the world (e.g.,

* Corresponding author. Tel.: +44 121 414 6138; fax: +44 121 414 4942. E-mail address: J.H.Tellam@bham.ac.uk (J. Tellam). 0048-9697/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2005.08.050

Chilton, 1997; Howard and Israfilov, 2002). A major issue is the sustainability of supplies of sufficient quantities of sufficient quality groundwater, given the potentially major effects that urbanization has on rates and quality of infiltrating water. Despite the importance of recharge in urban development, research is still at a relatively early stage, and there are no generally accepted methods for assessing the rates and quality of urban recharge.

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

159

This paper describes the development of a computer model, UGIf, for use in urban recharge and pollution studies. It is intended that the model be appropriate for the following purposes: (i) to provide input for regional groundwater solute transport models; (ii) to identify gaps in knowledge and data; (iii) to determine which processes are the most important for influencing urban groundwater quantity and quality; and (iv) to evaluate current recharge models. Once developed, the code could be used as a planning tool to investigate the effects, for example, of landuse or climate change, and as a tool for assessing groundwater vulnerability. For most urban areas, a central issue is the heterogeneity of sources, and hence the model has been developed using a GIS. The first part of this paper describes the model design. The process of dverificationT has begun, using data for 19802000 from the Birmingham urban Triassic Sandstone aquifer in the UK: this work is reported in the second half of the paper.

2. Overview of model The model has been developed for the case of an unconfined aquifer overlain by superficial deposits of various permeabilities. It deals with non-point source pollution only, though related models have been developed to deal with point source pollution (petrol stations, including non-aqueous phase liquids, and sewers) (Thomas, 2001). The main calculations evaluate recharge and pollutant flux rates at the water table: the latter include the effects of sorption and first order degradation. The main data sources are: geological maps; ground level and groundwater level maps; land cover maps; and meteorological data. The geological maps have associated attributes of hydraulic and chemical properties, and the land cover maps associated attributes of hydraulic properties and runoff water quality data. The basic structure of the model is shown in Fig. 1. A landuse/land cover classification allows the production of a land cover map. Each land cover class is assigned attributes relating to permeability/runoff and water quality. With meteorological data and the land cover-related runoff characteristics, an estimate of dpotential rechargeT for each of the landuse classes can be made, where potential recharge is defined here

Land Use Classification *1 Leakage Recharge Source Standard Soil Moisture Balance

*2 Chemical Data Map Attribute Table Data and /or Calculation Chemical Concentration in each Land use

*1

Potential Recharge from each Land Use Class Land Use Map *2

Rivers /Canals Mains / Sewers Septic Tanks Landfill Leachate Fuel Tank Spillage Horticulture Industrial Landfills/ Dumps Domestic Road River /Canal Sewer / Mains

Potential Recharge Map

Drift Classes

Drift Reaction

Potential Mass Flux

Interflow Indices Drift Map Drift Reaction Inter Flow Indices Actual Recharge Mass Flux In Recharge Pollutant Mass Flux From Drift Reaction Term

Fig. 1. The structure of UGIf (drift = superficial deposits).

160

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

as dactual rechargeT (i.e., the water reaching the water table) plus interflow. An estimate of potential mass flux (i.e., flux before interflow, evapotranspiration, and reaction have been taken into account) at the water table can also be made using runoff water quality data associated with the land cover classes. Interflow is estimated from the geology maps and used to convert the potential recharge estimates into dactual rechargeT estimates. Time taken to cross the unsaturated zone can be estimated using the actual recharge estimates, hydraulic properties related to the geological units, (reversible) sorption properties related to each geological unit, and unsaturated zone thickness as calculated from land surface and water table maps. Using pollutant-related reaction properties and the estimates of times taken to cross the unsaturated zone, the decay of degrading pollutants can be calculated, and thence the pollutant mass flux at the water table. Solute concentrations in recharge waters, corrected for evapotranspiration where necessary, are also calculated. Although the model is thus not very sophisticated, it does take into account the principal processes involved, and, as it is incorporated in a GIS, it allows the complexities of spatial heterogeneity to be investigated. A major limitation is in the way that time is dealt with. It is assumed that landuse and landuse-related properties do not vary within the dtime-sliceT or period being considered by the model. Daily recharge estimations are undertaken, and summed over the user-specified period. Within this period, steady-state conditions are assumed for the movement of water and solutes through the unsaturated zone. Thus individual recharge pulses are not tracked: residence time in the unsaturated zone is calculated on the basis of the averaged recharge rate, but it is only used, with a delay arising from any sorption, to estimate degradation/decay of the pollutant concentration. Without incurring very considerable computer run times, it would be difficult to track individual recharge pulses simultaneously. Another weakness is the lack of feedback between water table elevation and recharge rate: this could conceivably be added by iteration with a regional groundwater flow model, though again there are computer run time implications. In many cities, recharge from water supply networks is significant. This recharge source is not included in the model described here, but is dealt with by a separate code as described by Thomas (2001). 3. Code and code organization The model has been developed using ArcView GIS version 3.2, with Spatial Analyst (Environmental Sys-

tems Research Institute (ESRI), Redlands, USA). It can be divided into two main sections: the recharge model, and the non-point source (NPS) pollutant flux model. The recharge model includes estimation of both direct (rainfall) recharge and indirect (surface water) recharge. The NPS pollutant flux model is concerned with estimating fluxes and concentrations in recharge (and runoff), for both unreactive and reactive pollutants. The input data required are: (1) land use/land cover map with runoff and water quality attributes; (2) geological map of both bedrock and superficial deposits with hydraulic and chemical property attributes; (3) meteorological data; (4) ground surface elevation data in grid form; and (5) water table elevation map. The main outputs from the present model are distributions of: surface runoff, infiltration, potential recharge, ground level slope, interflow, actual recharge, pollutant fluxes in surface runoff, travel times of each pollutant through the unsaturated zone, and the pollutant fluxes and concentrations at the water table. 4. Recharge models 4.1. Direct recharge model 4.1.1. Conceptual model and definitions The conceptual basis for the direct recharge model is shown in Fig. 2. The term dinitial lossT signifies intercepted rainfall which ultimately does not infiltrate (vegetation and building interception/depression or detention storage). Formal definitions of infiltration, potential recharge, and actual recharge as used here are: Infiltration Rainfall Initial Losses Runoff Potential Recharge Infiltration Actual Evapotranspiration Actual Recharge Potential Recharge Interflow 3 where actual evapotranspiration is the evapotranspiration taking into account soil moisture availability. 4.1.2. Infiltration Infiltration requires estimates of rainfall, initial losses, and runoff (Eq. (1)). Rainfall estimates are 2 1

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

161

Precipitation
Initial Abstraction (Ia)

Ia

S, also called the retention parameter, represents how easy it is for the water to runoff, and is estimated using a dcurve numberT (CN). A curve number is a numerical description of how low the permeability of the land surface in a catchment is. It varies from 0 (100% rainfall infiltration) to 100 (0% infiltration). S is related to CN by:
Ia

Runoff

1000 10 CN

Infiltration Interflow

Recharge
Fig. 2. The conceptual model for the direct recharge calculations.

assumed to be available, but estimating runoff and initial losses is less straightforward. After considering various available methods for runoff estimation, the United States Department of Agriculture Soil Conservation Service (SCS, now known as the Natural Resources Conservation Service, or NRCS) Curve Number (CN) method was chosen. This empirical method is applicable in situations in which daily rainfall is available (US EPA, 1998b). It is simple, requires few input parameters, and has been widely applied in the field, including in urban catchments (US EPA, 1998a). However, it is accepted that the method, especially in the form applied here, is a considerable over-simplification of a complex set of processes, and further work will be necessary to justify its use and, when applied, parameter assignments. Nevertheless, it will suffice in the present context as an example runoff-estimating formulation. The method estimates surface runoff using: Q P Ia 2 P Ia S 4

CN is determined by several factors. The most important are the hydrologic soil group (HSG), the ground cover type, ground treatment, the hydrological condition of the ground, the antecedent runoff condition (ARC), and whether low permeability areas are connected directly to drainage systems, or whether they first discharge to a permeable area before entering the drainage system. HSG is particularly important in determining the runoff curve number. Soils are generally classified into four HSGs, named A, B, C, and D, according to how well the soil absorbs water after a period of prolonged wetting. In the current model, HSGs are assigned to each of the geological units: this is obviously a simplification, and if local data on soil properties exist, they should be used in preference. The resulting HSG theme is combined with the land cover map to produce a map showing areas under various landuses on different HSGs. A curve number value is assigned for each unit of this map, which leads to the preparation a runoff curve number map. Further details are given in US EPA (1998a). The term dinitial lossT incorporates rainfall loss due to interception, depression, and detention storage. The value of I a depends greatly on the vegetation or other (e.g., paved) cover types, the kind of soil (HSG, its treatment, and hydrologic condition), and the antecedent soil moisture of the area being studied. For a given drainage basin, the values of I a are highly variable. Ideally the values of the parameter I a should be evaluated with field data for each specific site. 4.1.3. Potential recharge For non-vegetated areas, the potential recharge is estimated using Eqs. (1) and (2), with (post infiltration) evapotranspiration set to zero, i.e.: Potential Recharge non vegetated areas Rainfall Initial Loss Runoff 6

in which Q P Ia S total rainfall excess (runoff) for a storm event [L]; total rainfall for a storm event [L]; initial loss [L]; potential maximum retention capacity of soil at beginning of storm (i.e., maximum amount of water that will be absorbed after runoff begins) [L].

This assumes that once into the ground, the water is unlikely to be removed by evapotranspiration. This may not be true where roads are tree-lined.

162

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

In vegetated areas, the calculation procedure takes into account the soil moisture deficit parameter and actual evapotranspiration. When the soil moisture deficit (SMD) (i.e., water required to bring the soil to maximum water holding capacity) at the end of the day is greater than zero, Potential Recharge 0 7

When the soil moisture deficit at the end of the day is zero: Potential Recharge vegetated areas Rainfall Initial Loss Runof f Actual Evapotranspiration 8

For convenience in the current version of the model, actual evapotranspiration is taken from the UK Meteorological Offices MORECS (Meteorological Office Rainfall and Evaporation Calculation System) data set. 4.1.4. Actual recharge Actual recharge is calculated from potential recharge by subtracting interflow (Eq. (3)). Interflow depends largely on the following characteristics: slope, porosity, storage capacity, HSG, average moisture content, effective permeability, permeability anisotropy, and lateral continuity of a perching horizon. Given the complexity of the process, rigorous calculation of interflow was not attempted. Instead, it was decided to develop a simple model based on indices reflecting what were viewed as the four most important parameters: (i) slope; (ii) specific retention (soil storage capacity); (iii) permeability anisotropy ratio of the formation (K h / K v); and (iv) the presence or absence of low permeability deposits underlying the more permeable deposits at the surface. The relationships are userdefined, but during the verification study (see below), the relationships were taken to be linear, with the indices taking a value between 0 and 1. The fourth index, the clay presence index, is based on the absence or presence of a low permeability layer underneath the outcropping deposits, and its form (sheet or lenses). When calculating interflow in this model, a weighting factor is assigned by the user to each of the four interflow indexes, so that the relative importance of each can be assigned. The lateral flow Q L is thus estimated using the following equation. h i QL PR IdF SL IdF SR IdF Kh=Kv IdF CP 9

where PR = potential recharge, I is the index value (0 1), and F is the weighting factor (01). Subscripts refer to the four main variables: SL= slope gradient; SR = specific retention; Kh / Kv = anisotropy ratio; and CP= clay presence. The sum F SL + F SR + F Kh/Kv + F CP is 1. Thus Q L = PR if each I value were at a maximum: alternatively, if clay layering was thought to be the predominant process, F for CP could be set to 1, the other weighting factors to 0, and Q L would then depend directly on the presence of clay layers. The input data for the actual recharge calculation are elevation grid, geological maps with attributes of specific retention, permeability anisotropy ratio, and low permeability deposit presence, and a potential recharge grid (preferably having an attribute of landuse types for summarizing interflow based on landuse). All calculations in this model use grid data. First a slope map is generated from the elevation grid theme. The potential recharge is also input as a grid. The interflow indices grids are generated from the respective attribute values of the vector format geological map. The output from this model is a map of interflow rates resulting from recharge at each pixel. As interflow is not tracked from one cell to the next, but assumed to discharge eventually into the drainage system, it is very important to check the summed predicted interflow against whatever drainage flow data are available. A priority in future versions of the model is to address the routing of interflow (and surface runoff): this will make the model substantially larger. The actual direct recharge is calculated finally by using the potential recharge grid and the interflow grid. 4.2. Indirect recharge model In the current model, recharge from surface water bodies is assumed to be at a constant rate. It would be simple to include a conductance term, but until the model is connected to a regional groundwater flow model, the non-linear connection between groundwater heads and surface water recharge cannot be included properly. 4.3. Non-point source pollution flux model 4.3.1. Flux rates for non-reactive pollutants in surface runoff The most common method of approximating NPS pollution uses long-term average contaminant loadings for common land uses. An early user of this approach was the US National Urban Runoff Program (US EPA, 1983), and it has been followed in many other countries. In this approach, Event Mean Concentrations are defined for each landuse. These EMCs are assumed

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

163

to be determined only by the landuse, and to remain constant independent of the duration and intensity of the rainfall events (Naranjo, 1998). EMC values are available in the literature (Lopes and Dionne, 1998; Delzer et al., 1996; Shepp, 1996). The EMC approach is simplistic, but until quantitative relationships between rainfall quantity and quality properties become available for urban areas, there is little point in making the model more sophisticated. The input data for the calculation of pollutant flux rates in runoff are therefore: (a) the landuse grid; (b) the grid of average annual runoff volume calculated by the recharge model; and (c) the associated EMC values for each landuse for each of the selected pollutants. An Avenue script was written linking the EMC value of various pollutants to the land use types. The EMC grid is then multiplied by the grid of average annual rainfall runoff. The result is the annual surface runoff loading for the NPS pollutant in each grid cell, i.e.: Load M=T Flow L3 =T4 Concentration M=L3 10 4.3.2. Flux rates for non-reactive pollutants at the water table EMC values can also be used for estimating the NPS pollutant fluxes in the infiltrating water (= rainfall iniinitial loss runoff):
Pollutant flux f or non-reacting solutes in inf iltrating water 2 M= L T EMC M=L3 Infiltration rate L=T 11

first order decay are taken into account. Pollutant fluxes are estimated through four stages, viz.: (1) Estimation of volumetric water content in the unsaturated zone; (2) Calculation of soilwater partitioning coefficients; (3) Calculation of retardation factors; and (4) Calculation of concentration and mass fluxes at the water table. 4.3.3.2. Estimation of volumetric water content. The volumetric water content in the unsaturated zone, h w, is calculated using the method of Clapp and Hornberger (1978): ! Vd 1=2b3 hw hs 13 Ks where h s is the saturated water content of the soil (total porosity), V d is the recharge rate, K s is the saturated hydraulic conductivity at the saturated water content h s, and b is the Clapp and Hornberger constant for the soil []. 4.3.3.3. Velocity of aqueous phase contaminant migration. The average linear velocity, v i, of a pollutant subject to linear sorption as it passes through the unsaturated zone is given by: q vi 14 hRf where q is the recharge rate [L/T], and R f is the retardation factor. The retardation factor is calculated using: Rf 1 qKd hs hKH = h 15

The infiltrating water is subject to evaporative loss across the soilair interface, and hence the concentration of the recharge water will be higher than that of the infiltrated water: Recharge concentration for non-reacting solutes M=L3 Pollutant flux in infiltrating water 2 M= L T =Recharge rateL=T 12

where q is the dry bulk density of the soil and K H is the Henrys law constant (gas concentration/aqueous concentration, []). The soilwater partitioning coefficient, K d, is estimated using: Kd Koc foc 16 where K d = soilwater partitioning coefficient, K oc = organic carbon partitioning coefficient, f oc = fraction of organic carbon within the soil matrix. 4.3.3.4. Calculating the time to reach the water table. The leading edge of the contaminated pulse will reach the water table at a time T that can be calculated using: T ZhRf q 17

4.3.3. Flux rates for reactive pollutants at the water table 4.3.3.1. Outline of calculation. As they pass through the unsaturated zone, pollutants in the infiltrating recharge water may be subjected to various processes including sorption, volatilization, and biodegradation. In the current model, volatilization, linear sorption and

164

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

where Z is the thickness of the unsaturated zone and q is the recharge rate. 4.3.3.5. Estimation of the effect of degradation. The model includes a first order rate equation to describe removal of a contaminant: C2 C1 exp T k 18

portion of the Birmingham Triassic Sandstone aquifer as the first step in verification. The main purpose is to determine how predictions of total recharge and runoff compare with other estimates, and whether the dstyleT of the results makes hydrogeological sense. To this end, input parameter values have been set a priori rather than being calibrated to produce the dbestT outcome. 5.2. The Birmingham aquifer Birmingham is located in central England (Fig. 3) and is Englands second largest city with a population of around 1.2 million people. It is a major industrial centre. The unconfined Triassic Sandstone aquifer (Fig. 3) underlies part of the city, and occupies an area of ~111 km2. It was chosen because of the considerable amount of hydrogeological data available. The southeastern boundary of the unconfined aquifer is defined by the Birmingham Fault which downthrows the aquifer, the Sherwood Sandstone Group, to the southeast so that it is confined below the Triassic Mercia Mudstone Group. The western and south-western boundaries have irregular shapes that correspond to the limit of the outcrops of sandstones as they thin out over older (Carboniferous) formations. The northern boundary of the study area is defined by a groundwater divide. The sandstones are covered by a lithologically variable sequence of Quaternary deposits of glacial, lacustrine, aeolian, and fluvial origin. These deposits are up to 20 to 30 m thick in places, and can seriously limit recharge locally. The hydrogeology and contamination of the aquifer has been described by Jackson and Lloyd (1983), Knipe et al. (1993), Rivett et al. (1990), Ford et al. (1992), Ford and Tellam (1994), and Shepherd et al. (2006-this issue). 5.3. Approach 5.3.1. Outline The following tasks need to be considered before the model can be implemented: ! ! ! ! ! ! development of a landuse/land cover map; collation of geological data; assignment of hydrological parameters; preparation of meteorological data; collation of EMC data; and collation of geochemical parameter values.

where C 1 is the initial concentration of chemical applied at the ground surface, C 2 is the concentration of chemical at the water table, T is the travel time through the unsaturated zone, and k is the first order degradation rate coefficient for the chemical. Substituting the expression for travel time in the above equation, and rewriting in terms of half-life (T 1 / 2), the concentration at the water table becomes: ! 0:693Rf Zh C2 C1 exp 19 qT1=2 The pollutant mass flux to the water table is given by: Pollutant Flux Net Recharge Rate Concentration of Pollutant Reaching the Water Table qC2 i.e., Recharge Pollutant Flux qC1 exp 0:693Rf Zh qT1=2 ! 21 20

4.4. Code validation The code was checked wherever possible against manual calculations representing a range of conditions. Details are given by Thomas (2001). 5. Preliminary verification 5.1. Introduction Verification of the model is not straightforward. Ultimately, it is intended that the results be compared first with field data on surface water flows and quality, and subsequently, once the model has been extended to other periods and incorporated into a regional groundwater solute transport model, with historical groundwater levels and quality. However, before this is attempted, the model has been applied to the case of the unconfined

These tasks are considered in turn in the following paragraphs. For each variable, a value was chosen based on literature values or on local data. The model

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

165

Fig. 3. The geology of the unconfined Birmingham aquifer area. Digital data from British Geological Survey (2000). The main aquifer, the Sherwood Sandstone Group, has three units, the Kidderminster, Wildmoor Sandstone, and the Bromsgrove Sandstone Formations. Head = weathered material redistributed by solifluxion. (For the colour version of the maps in this paper, the reader is referred to the web version of this article.)

was then run, and the predictions examined to determine any obvious inconsistencies. The only inconsistency found was that recharge rates for winter were greater in some paved areas than for some vegetated areas. As a result, CN values for the industrial, commercial, and transport landuses were all increased by a small amount (see below for values). No other modifications were made to the initial parameter values. A 10 m by 10 m grid size has been used throughout.

5.3.2. Landuse/land cover mapping Although there are existing landuse/land cover maps for Birmingham, none proved suitable for recharge estimation purposes, being compiled at an inappropriate scale, or not including features important to recharge. Hence a new scheme was set up. Table 1 shows the landuse classification used. The classification is an a priori hierarchical one with three levels. In a priori classification schemes, the land

166

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179 Table 1 (continued) 1. Urban built up area and developed land 232. Barren land (Sterile, non profitable land with or without vegetation) 24. Other Open Urban Areas 241. Verges and Lawn 242. Landfill/Waste Dumps 243. Open Derelict Land (Open dBrown FieldsT) 244. Cemeteries 245. Construction Sites 3. Shrub /scrub lands 31. Moor/Herbaceous Cover 311. Tall/Short Grass 312. Bracken 313. Grass Heath 32. Shrub Heath 321. Open Shrub Heath Land 322. Dense Shrub Heath Land 33. Scrub Land 331. Deciduous Scrub Land 4. Forest/woodlands 41. Deciduous forest 411. Mature Forest (Closed Canopy Woods) 412. Young Forest/Open Forest (Open Woods/Felled/Degraded) 42. Coniferous forest 421. Mature 422. Young 43. Mixed forest 431. Mature 432. Young 5. Water bodies 51. Inland water bodies 511. River/stream 512. Lakes/pond 52. Artificial water bodies 521. Canal 522. Reservoirs

Table 1 Urban landuse classification scheme 1. Urban built up area and developed land 11. Residential 111. High density residential (Terraced/multi-storied flats) 112. Medium density residential (Semi detached) 113. Low Density Residential (Detached) 12. Commercial/Business and Services 121. Commercial Retail (Supermarkets, Gasoline Stations, Building Materials etc.) 123. Commercial Wholesale/Warehouses/Depots 124. Institutional (Governmental/Educational/Medical/ Religious) 125. Commercial Services (Finance/Real Estate/Insurance/ Repairs/Automotives) 126. Water Treatment/Sewage Treatment 127. Hotels/Lodging 13. Industrial/Manufacturing 131. Industrial Heavy (Chemical, Metal, Electrical, Automotive) 132. Industrial Medium (Raw material processing and preparation) 133. Industrial Light (Food and Drink Processing, Furniture/ Wood Processing) 14. Transportation 141. Railway Transport facilities 142. Road Transport Facilities 143. Air Transport facilities 144. Water Transport Facilities 15. Parking/Car Parks 151. Industrial Car Parks 152. Commercial Car Parks 153. Office Car Parks/Institutional Car parks 16. Roads 161. Motorways 162. A Road 163. B Road 164. Minor Road 17. Pavement/Pedestrian Footpath 171. Pavement Brick 172. Pavement Concrete 173. Pavement Asphalt 18. Derelict Built Up Land (dBrown FieldsT) 181. Industrial Derelict Built up Land 182. Commercial Derelict Built up Land 2. Open urban area 21. Recreation 211. Parks/Gardens 212. Golf Courses 213. Playgrounds/Sports fields 214. Zoos 22. Agriculture/Horticulture 221. Pasture Land/Meadow 222. Allotment Gardens 223. Farms 224. Orchards/Vineyards/Nurseries 225. Animal Operations (Horses, Poultry, Livestock) 23. Bare Land 231. Public Open Ground (Hard ground used for cultural/ social purposes)

use/land cover classes are defined prior to the collection of data, the main advantage being that the classes are standardized independently of the area to be described and the means/techniques used in the mapping (De Bie et al., 1996; AFRICOVER, 1998). Initially, a preliminary dlevel IT classification system was constructed, which consisted of five possible major classes: (1) urban built up areas and developed land; (2) open urban areas; (3) shrub/scrub land; (4) forest/woodlands; and (5) water bodies. These broad level I classes are subdivided further into dlevel IIT and dlevel IIIT classes which have direct significance for recharge and pollution activities. Using this classification, a landuse map was prepared for the area of the unconfined Birmingham aquifer for the period 1980 to 2000: twenty landuse classes from Table 1 were eventually included, the

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

167

Fig. 4. Landuse map for the Birmingham unconfined aquifer for 19802000. This map is based in part on Ordnance Survey data: n Crown copyright Ordnance Survey. All rights reserved.

others not being found in significant proportions (Fig. 4). The information sources were: digital topographic maps (1 : 1250 scale, Ordnance Survey (OS)); satellite imagery (LANDSAT TM images); aerial photographs (Cities Revealed n orthorectified digital images having a resolution of 2 m). The satellite images were rectified and enhanced using the ArcView Image Analysis extension, and then used as the basis for the maps, with the OS data being used to identify most landuses. Where a feature was not clear on satellite images or OS maps, visual interpretation of

the aerial photographs was undertaken. In addition, local knowledge of the area was used where other sources of information were not adequate. Details are given by Thomas (2001). Fig. 5 shows the proportion of land overlying the Birmingham unconfined aquifer associated with each of the 20 landuse classes. Residential areas dominate, covering N50% of the area in aggregate, with commercial, recreation grounds, minor roads, and woodlands being present in 5% to 10% of the area: all other classes of land use are present at less than 3%.

168

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

30.00

Area Percentage

20.00

10.00

0.00
Lo w D ig en s h D ity en R s i es R ec t y id re R en es t i at a io i n Co den l M G ed ro m m t i a iu un e l m d rc i a D en M (Gr l in as si ty or s) O pe W Re Ro n o o si ad d G r o d l a en un nd tia d/ /S l G hr ra ub A ssl gr an ic u d In ltu R es du re er s vo 'A tri i r / ' al L a Ro C ke ad em /P et o er 'B nd y/ ' R G ra oad R a i vey lw ay ard M Ya ot rd or C wa ar y Pa r C k an Tr al an sp Ri v or er ta tio n

Fig. 5. Landuse in the unconfined area of the Birmingham aquifer, 19802000, as a percentage of total land area.

5.3.3. Geological data ArcView GIS compatible pre-Quaternary geology and Quaternary geology maps on 1 : 10000 scale were procured from the British Geological Survey British Geological Survey, 2000). These maps were reconciled and merged to generate a combined map with all geological units (Fig. 3). The area of outcrop of each geology unit was calculated within the GIS and added to its attribute table. Finally, the

total area of each geological unit was also estimated (Table 2). 5.3.4. Hydrologic soil group HSGs are needed to estimate curve number for runoff estimation. As no soil map was available, it was decided to use the geology map to prepare a map of HSG distributions. Table 3 lists the HSG assigned to each geological unit.

Table 2 The hydraulic and interflow index parameters assigned to each of the geological units in the study area Names of geological formations Lithology Total area (km2) Alluvium Bromsgrove sandstone formation Glaciofluvial deposits Glaciolacustrine deposits Head Hopwas breccia Kidderminster formation Lacustrine deposits River terrace deposits, First terrace River Terrace deposits, Second terrace Salop Formation Sandy till Till Wildmoor sandstone Formation Silty clay with sand and gravel Sandstone Sand and gravel Clay and silt Clay with rock fragments and pebbles Breccia and sandstone Sandstone Clay, silt and sand Sand and gravel Sand and gravel Mudstone Sandy, pebbly clay with lenses of sand and gravel Pebbly clay Sandstone Total 5.72 8.25 44.49 1.33 1.56 0.92 22.25 0.14 3.58 1.62 0.44 0.36 17.21 3.16 111.02 5.16 7.43 40.07 1.2 1.4 0.83 20.04 0.13 3.23 1.46 0.4 0.32 15.5 2.85 100 Area % Porosity () 0.40 0.27 0.35 0.40 0.42 0.24 0.25 0.35 0.35 0.25 0.25 0.26 Clapp and Hornberger constant () 1.26 2.12 1.90 1.40 1.35 2.50 2.38 1.80 1.80 1.40 1.40 2.21 Clay index () 0.8 0.5 0 0.9 0.95 0.05 0.3 0.8 0.1 0.1 1 0.8 0.95 0.3 Permeability anisotropy Ratio(K h / K v) 75 1.37 1.54 166.36 348.84 2 1.81 95.19 4.26 4.26 4.92 90.19 184.84 1.88 Specific retention () 0.35 0.17 0.09 0.34 0.39 0.04 0.1 0.29 0.09 0.09 0.28 0.09 0.26 0.14

[Data from: Allen et al. (1997), Bridge et al. (1997), Freeze and Cherry (1979), Shepherd (2003), and Clapp and Hornberger (1978)].

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179 Table 3 Lithological and textural descriptions of geological units of the Birmingham area Name of formation Glaciofluvial deposits River terrace deposits, first terrace River terrace deposits, second terrace Bromsgrove sandstone formation Lithology Yellow and orange sand and gravel Mainly sand and gravel A Mainly sand and gravel Sandstone, red brown, micaceous, pebbly in part (interbedded with mudstone in upper part Sandstone, orange-red, feldspathic, with sparse thin mudstone beds Sandstone, red, and pebbly sandstone, pebblecobble conglomerate in lower part Breccia and pebbly sandstone Silty clay with sand and gravel and, locally, peat Yellow to brown stoneless clay and silt Brown or red brown, sandy, pebbly clay with lenses of sand and gravel Yellow, red or brown clay with rock fragments and pebbles Clay, silt and sand Brown or red brown pebbly clay Mudstone, red, and sandstone, red brown with subordinate lenticular beds of conglomerate and thin beds of limestone Hydrologic soil group A A A A

169

tion, and values are likely to be altered as new field data become available. 5.3.6. Initial losses Since no field data on initial loss in Birmingham are currently available, a literature survey was carried out to provide possible values (see, e.g., Browne (1990) and NCSPA (1999)). The initial loss values tentatively chosen for the land uses in Birmingham are shown in Table 4. Ultimately it may be possible to refine CN and I a values by testing against field runoff and groundwater data. For example, Durr (2003) has recently suggested that runoff is not seen until at least 0.5 to 2 mm rainfall has occurred, and if this were the case in Birmingham, it is possible that the figures listed in Table 4 will be a little too large. 5.3.7. Preparation of meteorological data Daily rainfall and MORECS data for 19801999 were procured from the UK Meteorological Office. Apart from direct use in the recharge calculation, these data allow assignment of antecedent moisture conditions for the curve number runoff method. Wet conditions (antecedent moisture condition III, see below) were assumed when SMD values were close to zero (010 mm) for a period of 1015 consecutive days; otherwise, normal conditions (antecedent moisture condition II) were assumed.
Table 4 Values of initial loss for each landuse/land cover class in Birmingham Land use/land cover Initial surface loss (mm) 3.0 3.0 4.1 4.5 5.1 2.0 3.0 5.5 5.5 8.0 5.5 5.5 0 0 0 2.0 2.0 2.0 2.0 2.5 4.4

Wildmoor sandstone formation Kidderminster formation

Hopwas breccia Alluvium Glaciolacustrine deposits Sandy till

A B C C

Head

Lacustrine Deposits Till Salop Formation

D D D

5.3.5. Interflow indices There are four interflow indices, requiring data on: slope angle, clay presence below surface deposits, permeability anisotropy, and specific retention. Slope angle is calculated by the model from the topographic elevation data. For each outcropping geological unit, a clay presence index ranging from zero (no clay) to one (sheet clay) was assigned, based on interpretation of local geological information. The permeability and specific retention data were obtained from Allen et al. (1997), Bridge et al. (1997), and Buddemeier (1996). This approach is an approxima-

Commercial/business Industrial High density residential Medium density residential Low density residential Car parks Transportation Recreation ground Agricultural/horticultural/farm Woodland/shrub Cemetery/graveyard Open ground/grassland Reservoir/lake/pond River Canal Motorway dAT Road dBT Road Minor road Railway yard Area-weighted average

NB: The term dinitial surface lossT incorporates rainfall loss due to interception, depression and detention storage.

170

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179 Table 6 Event mean concentrations assigned for each landuse class Land use/land cover Commercial/business Industrial High density residential Medium density residential Low density residential Car parks Transportation Recreation ground Agricultural/horticultural/farm Woodland/shrub Cemetery/graveyard Open ground/grassland Reservoir/lake/pond River Canal Motorway dAT road dBT road Minor road Railway Yard Nitrate (mg N/l) 1.2 1.9 2.1 1.8 1.8 0.8* 0.8 0.9 4.1 0.8 1.2 1.8 0.6 0.6 1.5*** 0.8 0.5* 0.5* 0.6* 0.2 Chloride (mg/l) 148 148 148 50*** 15*** 42* 148*** 0.9** 0.9** 0.9** 0.9** 0.9** 0.9** 0.9*** 10*** 149*** 149* 125* 15.4* 0.9 Toluene (Ag/l) 2.4 1.8 1.8 1.8*** 0.9*** 2.4*** 2.4*** 0 0 0 0 0 0 0 0 1.2*** 1.2*** 1.2*** 0.6*** 0.1***

5.3.8. NRCS curve numbers The curve numbers for normal antecedent moisture conditions (AMC II) are calculated as explained above. For wet conditions (AMC III), curve numbers are calculated by (US EPA, 1998a): CN III 23CN II 10 0:13CN II 22

Using this relationship, the NRCS curve numbers for Birmingham landuse classes during wet conditions were calculated, and these are listed in Table 5. 5.3.9. Surface water infiltration rates The infiltration rates for surface waters were set at 2 mm/day for reservoirs and 12.5 mm/day for unlined canals. The latter figure was obtained from information supplied by British Waterways. The River Tame is effluent (Ellis, 2003). 5.3.10. Collation of geochemical parameter values For this investigation of UGIf, the pollutants chloride, nitrate, and toluene have been considered. Chloride will be considered inert. Nitrate will also be considered inert, and, although this is clearly untrue,

* Runoff chemistry data obtained from Harris, 2000, Unpublished data and Antonio, 1999, Unpublished data. ** Chloride measured in rainfall samples collected at Winterbourne, Birmingham University. *** Estimates based on Birmingham University campus measurements and literature data.

Table 5 NRCS curve numbers for various landuse/land cover types in the Birmingham area (Antecedent Moisture Condition III) Land use/land cover Average % of impermeable area 90 90 75 65 4550 99 90 5 4 2 25 2 Hydrologic condition and vegetative ground cover Curve numbers for hydrologic soil group A 98 98 94 92 86 99 98 59 59 56 69 59 100 100 100 99 99 99 99 73 B 99 99 97 95 92 99 99 78 78 78 84 78 100 100 100 99 99 99 99 85 C 99 99 98 97 95 99 99 88 88 86 91 88 100 100 100 99 99 99 99 91 D 99 99 98 97 96 99 99 91 91 91 93 91 100 100 100 99 99 99 99 94

Commercial Industrial High density residential Medium density residential Low density residential Car parks Transportation Recreation ground Agricultural field Woodland/Scrub Cemetery/graveyard Open ground/grassland Reservoir/lake/pond River Canal Motorway A road (paved) B road (paved) Minor road (paved) Railway yard

Good*; N75% Ground Cover Good*; N75% Ground Cover Good*; 50% Woods and 50% Grass Poor*; Grass b50% Good*; Grass N75%

99 99 99 99 35 40

Poor*: factors impair infiltration and tend to increase runoff. Good*: factors encourage average and better than average infiltration and tend to decrease runoff.

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

2000 1800 1600 1400 1200 1000 800 600 400 200 0 120 Recharge (mm/y) 100
Rainfall Recharge

171

Rainfall (mm/y)

80 60 40 20 0

Fig. 6. Calculated actual recharge as a function of time in the unconfined area of the Birmingham aquifer.

140.0 120.0 100.0 80.0 60.0 40.0 20.0 0.0 1975 1980

Recharge (mm/y)

Recharge (mm/y)

19 8 19 0 8 19 1 8 19 2 8 19 3 8 19 4 8 19 5 8 19 6 8 19 7 8 19 8 8 19 9 9 19 0 9 19 1 9 19 2 9 19 3 9 19 4 9 19 5 9 19 6 9 19 7 9 19 8 99

Year

100.0 80.0 60.0 40.0 20.0 0.0 1975 1980 1985 1990 Year 1995 2000 HDR MDR LDR

Commercial Industrial 1985 1990 Year 1995 2000

200.0 Recharge (mm/y) 150.0 100.0 50.0 0.0 1975 Car Park 1980 Transport 1985 1990 Year 1995 2000

120.0 Recharge (mm/y) 100.0 80.0 60.0 40.0 20.0 0.0 1975 1980 1985 Recreation Ground (Grass) 1990 Agriculture Woodland/Shrub Cemetery/Graveyard Open Ground/Grassland 1995 Year 2000

450.0 400.0 350.0 300.0 250.0 200.0 150.0 100.0 50.0 Motorway 0.0 1975 'A' Road 1980 1985 1990 1995 2000 Year 'B' Road Minor Road Railway Yard

4500.0 4000.0 3500.0 3000.0 2500.0 2000.0 1500.0 1000.0 500.0 0.0

Recharge (mm/y)

Recharge (mm/y)

Reservoir/Lake/Pond River Canal

1975 1980 1985 1990 1995 2000 Year

Fig. 7. Variation of calculated recharge for the unconfined area of the Birmingham aquifer as a function of landuse type over the time period 1980 2000.

172

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

the results allow the influence of source on recharge concentrations to be investigated. Toluene will be considered to be sorbing and degrading. EMC data are not available for all the landuse classes in the Birmingham area. However, some unpublished data are available (Antonio, 1999; Harris, 2000) that provide some information on concentrations of nitrate and chloride. Using the latter information, together with other data from the literature, the EMC values listed in Table 6 were assembled.

The Henrys Law constant for toluene was taken as 643 kPa m3 mol 1, and the decay half-life as 28 days. The K d values were calculated using the f oc values (Shepherd, 2003; Shepherd et al., 2006-this issue) listed in Table 2. Table 2 also lists the porosity and Clapp and Hornberger constants used. The latter were estimated from the textural descriptions of each of the geological units, but some measurements of the former are available (e.g., Allen et al., 1997).

Fig. 8. Calculated recharge in the unconfined area of the Birmingham aquifer for autumn and winter 1980 (cm). This map is based in part on Ordnance Survey data: n Crown copyright Ordnance Survey. All rights reserved.

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

173

6. Model results 6.1. Recharge Due to limitations on the size of ArcView grid files, only ~6 months of daily data could be run at one time. Hence the data had to be split for calculation purposes. This was usually done by considering, for each year, data for winter and autumn, and data for spring and summer. Winter and autumn were defined on the basis that SMD b 10 mm. For example, for 1980, the first run would comprise the data for January to March and October to December, and the second run would comprise the data for April to September. Some of the data sets had to be further split to conform to the ArcView size limitations. In total 51 model runs were needed to cover the twenty year period investigated. The recharge model runs generate grids of potential recharge, interflow, and actual recharge. Each direct recharge calculation took (in 2001) ~2.5 h using a Pentium III 450 MHz 128 Mb RAM PC (with solute transport, 67 h). The total predicted actual recharge to the aquifer is illustrated in Fig. 6, and predicted recharge for each landuse class in Fig. 7. The highest recharge rate calculated was 113 mm/year or 34 ML/day for 1993. The total rainfall in this year was 755 mm. Examination of rainfall data shows that the highest rainfall recorded during the 20 years was 820 mm in the year 1998. However, the predicted actual recharge in this year (33 ML/day or 110 mm/year) is slightly less than that in 1993. This is explained by higher evapotranspiration rates and SMD values in 1998. The lowest predicted recharge rate is 60 mm/year (18 ML/day) for 1996. The total rainfall (559 mm) was also the lowest over the 20 year period. In his regional groundwater flow modelling study, Greswell (1992) estimated recharge rates for 1980 1989 to be 31.2 ML/day on average excluding leakage from water supply pipelines (mains), and excluding recharge from surface water bodies (summarized in Knipe et al., 1993). Including surface water recharge, Greswells (1992) estimate is 34.9 ML/day. For 1980, the equivalent figures as predicted by UGIf are 28.6 ML/day and 33.5 ML/day (94 and 110 mm/year), which agrees well. Using the results of a baseflow analysis, Ellis (2003) estimates that recharge within the Tame catchment is ~28% of the rainfall, i.e. 196 to 224 mm/year (60 to 68 ML/day) for rainfall of 700 to 800 mm/year. This figure includes water supply (mains) leakage, and contributions from areas other than the sandstone. Using the estimate of Knipe et al. (1993) of 8.5 ML/day (28 mm/year) for water supply (mains)

leakage within the sandstone area of the catchment, and assuming that the part of the Tame catchment which overlies sandstone aquifer is similar to the rest of the catchment, the recharge estimated for the aquifer becomes 78 to 106 mm/year (24 to 32 ML/day) depending on the rainfall. This is probably an underestimate, as the baseflow contribution from the sandstone reach of the Tame may be greater than that from the non-sandstone part; however, the latter will have some baseflow from the Carboniferous bedrock, and in addition does contain disused mine systems and shallow Quaternary deposits which do contribute significantly to what would be interpreted as baseflow. Again, the agreement with the estimate of UGIf is encouraging. A full comparison with predicted surface water runoff is rather more complex, and has yet to be carried out. However, Ellis (2003) estimates that, at mean flows, non-groundwater inputs to the Tame are ~40 to 50 ML/day, and this compares with the UGIf estimate of ~40 ML/day. These comparisons do not constitute a rigorous test of the model, and the agreement may well be fortuitous. All that should be concluded is that the uncalibrated model is producing credible estimates of recharge or total runoff. A more interesting exercise is to examine the detail of the estimates. Fig. 8 shows an example potential recharge map produced for the 6 months of autumn and winter 1980. As expected, the predicted pattern is complex, reflecting both very heterogeneous landuse

% Mean Total Pot. Recharge

16.0 14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0

R LD M in Rd R HD al m an Co C cr Re ke La D M R d In A Rd

% Mean Total Pot. Recharge

14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0

Pk C em et er y Tr an sp gr ic ai lw ay oo ds G rd R rw ot o M ay d C ar A B

O pe n

Fig. 9. Calculated recharge as a function of landuse in the Birmingham unconfined aquifer.

174

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

distributions and geology. Nevertheless, the influences of the landuse patterns can be discerned, and these appear therefore to be dominant. Fig. 9 shows the amounts of total potential recharge categorized according to landuse. Much of the aquifers recharge is predicted to come from areas of low density housing, minor roads, high density housing, commercial development, and canals, with relatively little coming from woods, open ground, agriculture, railways, dBT roads, motorways, car parks, cemeteries, and transport depots. This is to be expected given the relative areas of each type of

land cover (Fig. 5). However, Fig. 7 shows that the recharge rate for the vegetated areas is less than that for the paved areas when averaged over a year. One reason for this is that the vegetated areas in the city correspond with locations where the geological units are less permeable (cf. Fig. 3). Another factor arises through the assumption that the recharge in the paved areas is not dependent on evapotranspiration. When recharge occurs in the vegetated areas, i.e., when SMD~0, the rates are greater than for paved areas. But when SMD N 0 and recharge in the vegetated areas ceases, recharge continues to occur through

Fig. 10. Calculated volumes of interflow generated per 100 m2 pixel over 169 days in autumn and winter 1980 for the unconfined area of the Birmingham aquifer. This map is based in part on Ordnance Survey data: n Crown copyright Ordnance Survey. All rights reserved.

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

175

the paved areas. The model suggests that this is enough to make the overall year-averaged recharge rate greater than that for vegetated areas. In addition, the lack of dependence of recharge on evapotranspiration in the paved areas makes the variability of annual recharge rates in the paved areas less than that in the vegetated areas, as shown in Fig. 7. It may well be that these model predictions are incorrect (e.g., because initial losses are underestimated): however, if true, they have important landuse and water quality implications, and hence more research is needed especially on the topic of recharge through paved

areas. For example, is evapotranspiration really not important? What about moisture re-distribution following infiltration in restricted zones? Does fingered flow establish itself (e.g., Ritsema, 1999; Sililo and Tellam, 2000)? What implications are there for pollutant travel times and attenuation? Sensitivity analysis for the direct recharge calculation indicates the following relative importance for the main variables: Grassed areas with underlying wet sand: Rainfall N I a, CN H [actual evapotranspiration N Clay index N Storage index N Slope N Permeability anisotropy]

Fig. 11. Calculated chloride concentrations based on recharge rates for winter and autumn 1980 for the unconfined area of the Birmingham aquifer. This map is based in part on Ordnance Survey data: n Crown copyright Ordnance Survey. All rights reserved.

176

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

Grassed areas with underlying wet clay: CN N Rainfall N Clay index N I a N Permeability anisotropy N Actual evapotranspiration N Storage index H [Slope] Commercial areas with underlying wet sand: CN N Rainfall H [Clay index N Storage index N I a N Slope N Permeability anisotropy] Commercial areas with underlying wet clay: C N N Rainfall H Permeability anisotropy N Storage index H [I a N Slope]

Some variables, those indicated in square brackets above, have little effect on the recharge estimates in the case of Birmingham. Fig. 10 indicates that interflow, as calculated by the model, is likely to be of most importance in the southwest of the aquifer area. Overall, interflow is estimated by the model to be ~10% potential recharge. There are few data to test these predictions, but the predicted difference between the southwestern parts of the aquifer and the rest of the study area is so marked that this

Fig. 12. Calculated nitrate concentrations based on recharge rates for winter and autumn 1980 for the unconfined area of the Birmingham aquifer. This map is based in part on Ordnance Survey data: n Crown copyright Ordnance Survey. All rights reserved.

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

177

seems to be a real feature. This conclusion suggests that the assumption of vertical flow through the unsaturated zone used when calculating the reactive pollutant fluxes will, in much of the rest of the aquifer at least, be reasonable. 6.2. Pollutant fluxes Fig. 11 shows the predicted concentrations of chloride in the recharge waters. Concentrations range from mg/l level to over 100 mg/l, as might be expected given

the EMC values used. It is immediately clear that the distribution of concentrations is complex, with large local differences. However, the main landuse controls are also evident, with, for example, the high density/ low density residential areas showing up in the southern part of the region (cf. Fig. 4). The residence times for pumped groundwaters in the unconfined part of the aquifer can be anything up to at least hundreds of years (Tellam and Thomas, 2002), and this, and the complexities caused by pumping boreholes, make detailed comparisons inappropriate. However, the range

Fig. 13. Calculated toluene concentrations based on recharge rates for winter and autumn 1980 for the unconfined area of the Birmingham aquifer. This map is based in part on Ordnance Survey data: n Crown copyright Ordnance Survey. All rights reserved.

178

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179

of concentrations seen and the scale of variability are features of the real system: for example, at one single industrial site, roughly 1 km2 in area, boreholes with a great range of depths extract water with Cl concentrations ranging from b 15 to N100 mg/l (Ford and Tellam, 1994). Fig. 12 shows the predicted nitrate distributions. Concentrations between less than a milligram per liter and a few milligrams per liter are seen. Again a local scale complexity is seen to be superimposed over a larger-scale pattern. The concentrations of nitrate are much less than the maximum measured values, but these latter are probably remnants of sewage pollution, much of which may be from times before the city had a complete sewerage cover. Fig. 13 shows the predicted toluene concentrations. It is clear that, for the rather low EMC toluene values normally presented in the literature, groundwaters are at little risk. The present model predicts measurable concentrations only where the water table is close to ground surface (around the river Tame). Field data do not yet exist, but would in any case reflect not just diffuse toluene sources, but point free phase sources as well. 7. Conclusions Although there is much further scope to develop and verify the model, the early results suggest that UGIf will reproduce city-wide recharge rates which are compatible with different approaches adopted in previous work, and that rates for particular landuses do not appear to be unreasonable. Further work is required to test the model predictions against field runoff data, and ultimately, against groundwater balance and head data. The latter test would require combining several UGIf models and a mains water leakage model with a regional groundwater flow model. The model outputs indicate a system which shows great spatial complexity at small scales, but nevertheless still displays a recognizable larger-scale structure, the latter caused mainly by landuse and geological distributions. The importance of these larger-scale structures may explain why the total recharge predicted by the model is similar to that predicted by previous work which was based on much coarser spatial averaging: it also suggests that for overall recharge assessment this coarse spatial averaging is satisfactory, but that it may be much less satisfactory for assessments of pollutant movement, where small-scale variations are important. The present results indicate that the detailed landuse classification might be simplified for recharge

assessment (though the simplification which works best for recharge quantity estimation may not work best when attempting to estimate recharge quality). The prediction that recharge in paved areas might be larger than commonly considered likely from vegetated areas may just be an artefact of the input parameters chosen: however, consideration of the reasons for the prediction suggests that further research here is needed. The model suggests that groundwater pollution in this aquifer from non-point sources of organics such as toluene is unlikely to be important. Finally, the model has highlighted the need for more knowledge/data on the following: runoff estimation; interflow (including the effect of lateral flow and channelling on flow times and water quality); evapotranspiration in paved areas; unsaturated zone flow below paved areas (fingering/funnelling? moisture redistribution?); and the role of the pipe network. Acknowledgments We would like to thank the Commonwealth Scholarships Commission, United Kingdom, the Punjab Remote Sensing Centre, Ludhiana, India, The British Council, and the Natural Environment Research Council (URGENT) for the financial and organizational support which made this project possible. We would like to acknowledge with gratitude the help provided by Alan Dean and Richard Greswell. References
AFRICOVER. Technical document on the AFRICOVER land cover classification scheme: a dichotomous, modular-hierarchical approach. Rome, Italy7 FAO Research, Extension and Training Division, Viale delle terme di caracalla, 00100; 1998 http://www. fao.org/WAICENT/FAOINFO/SUSTDEV/EIdirect/EIre0044.htm. Allen DJ, Brewerton LJ, Coleby LM, Gibbs BR, Lewis MA, MacDonald AM, et al. The physical properties of major aquifers in England and Wales. British geological survey technical report WD/97/34. Nottingham, UK7 Keyworth; 1997. Antonio, E. Stormwater drain chemistry and its sources on the University of Birmingham Main Campus. Unpublished BSc Project Report, School of Earth Sciences, University of Birmingham, 1999. Bridge DMcC, Brown MJ, Hooker PJ. Wolverhampton Urban Environmental Survey: an integrated geoscientific case study. British Geological Survey Technical Report WE/95/49. Nottingham, UK7 Keyworth; 1997. British Geological Survey. Digital drift and solid geology 1 : 10000 maps for the Birmingham area, 2000. Browne FX. Stormwater management. In: Corbit Robert A., editor. Standard handbook of environmental engineering. New York7 McGraw-Hill Publishing Company; 1990. Buddemeier RW. Groundwater flux to the ocean: definitions, data, applications, uncertainties. University of Kansas, Lawrence, KS

A. Thomas, J. Tellam / Science of the Total Environment 360 (2006) 158179 66047 USA7 Kansas Geological Survey; 1996 http://data.ecology. su.se/mnode/gwtable_1.htm. Groundwater in the urban environment, volume 1 problems, processes and management. In: Chilton PJ, editor. Proceedings of the XXVII international association of hydrogeologists conference, Nottingham, UK, 2127 September 1997. Rotterdam, The Netherlands7 A.A. Balkema; 1997. Clapp RB, Hornberger GM. Empirical equations for some soil hydraulic properties. Water Resour Res 1978;14;601 4. De Bie, CA, van Leeuwen, JA, Zuidema, PA, The land use database; a knowledge-based software program for structured storage and retrieval of user-defined land use datasets. ITC, FAO, WAU, June 1996. 1996. http://version0.neonet.nl/itc/education/larus/ landuse/manual.html. Delzer GC, Zogorski JS, Lopes TJ, Bosshart RL. Occurrence of the gasoline oxygenate MTBE and BTEX compounds in urban stormwater in the United States, 199195. Water-Resources InvestigationsU. S. Geological Survey http://sd.water.usgs.gov/nawqa/ pubs/wrir/wrir96.4145/wrir.doc.html. Durr, CS. The hydrological response of large urban catchments. Unpublished PhD Thesis, University of Birmingham, 2003. Ellis, PA. The impact of urban groundwater upon surface water quality: BirminghamRiver Tame study, UK. Unpublished PhD Thesis, University of Birmingham, Birmingham, UK, 2003. Ford M, Tellam JH. Source, type of extent of inorganic contamination within the Birmingham urban aquifer system, UK. J Hydrol 1994;156;101 35. Ford M, Tellam JH, Hughes M. Pollution-related acidification in the urban aquifer, Birmingham, UK. J Hydrol 1992;140;297 312. Freeze RA, Cherry JA, 1979. Groundwater. Englewood Cliffs, N.J.7 Prentice-Hall Inc.; 1979. p. Greswell, RB. The modelling of groundwater rise in the Birmingham area. Unpublished MSc Thesis, University of Birmingham, UK, 1992. Harris, JM. Urban Runoff Contribution to groundwater contamination. Unpublished Report. School of Earth Sciences, University of Birmingham, 2000. Howard KWF, Israfilov R, editors. Current problems of hydrogeology in urban areas, urban agglomerates and industrial centresNATO science series, IV Earth and environmental sciences vol. 8. Netherlands7 Kluwer Academic Publishers; 2002. Jackson D, Lloyd JW. Groundwater chemistry of the Birmingham Triassic sandstone and its relationship to structure. Q J Eng Geol 1983;16;135 42. Knipe CV, Lloyd JW, Lerner DN, Greswell R. Rising groundwater levels in Birmingham and the engineering implications. CIRIA (Construction Industry Research and Information Association), Special Publication 92, London; 1993 http://www.ciria.org.uk/ acatalog/SP092.html (ISBN: 0-86017-364-X).

179

Lopes TJ, Dionne SG. A review of semivolatile and volatile organic compounds in highway runoff and urban stormwater. US Department of the Interior US Geological Survey Open-File Report 98409 http://ma.water.usgs.gov/fhwa/products/ofr98-409.pdf. Naranjo E. A GIS based nonpoint pollution simulation model. Denmark7 VKI, Institute for the Water Environment; 1998 http://www.esri. com/library/userconf/europroc97/4environment/E2/e2.htm. NCSPA (National Corrugated Steel Pipe Association). Modern sewer design. Fourth edition. 1101 17th Street, NW, Suite 1300, Washington, DC 20036-47007 American Iron and Steel Institute; 1999. Ritsema CJ, editor. Preferential flow of water and solutes in soilsJournal of hydrology special issue 1999;vol. 215, p. 1 214. Rivett MO, Lerner DN, Lloyd JW, Clark L. Organic contamination of the Birmingham aquifer. J Hydrol 1990;113;307 23. Shepp DL. Petroleum hydrocarbon concentrations observed in runoff from discrete, urbanized automotive-intensive land uses. Washington, USA7 Metropolitan Washington Council of Governments; 1996 http://www.epa.gov/OWOW/watershed/Proceed/ shepp.html. Shepherd KA. Contamination and groundwater quality in the Birmingham Aquifer. Unpublished PhD Thesis, University of Birmingham, 2003. Shepherd KA, Ellis PA, Rivett MO. Integrated understanding of urban land, groundwater, baseflow and surface-water qualityThe City of Birmingham, UK. Sci Total Environ 2006;360;180 95. doi:10.1016/j.scitotenv.2005.08.052 [this issue]. Sililo OTN, Tellam JH. Fingering in unsaturated zone flow: a qualitative review with laboratory experiments on heterogeneous systems. Ground Water 2000;38;864 71. Tellam JH, Thomas A. Well water quality and pollutant source distributions in an urban aquifer. In: Howard KWF, Israfilov RG, editors. Current problems in hydrogeology in urban areas, urban agglomerates and industrial centresNATO science series, IV earth and environmental sciences 2002;vol. 8, p. 139 58. Thomas, AA. Geographic information system methodology for modelling urban groundwater recharge and pollution. Unpublished Ph D Thesis, University of Birmingham, Birmingham, UK, 2001. US EPA RG. Results of the nationwide urban runoff program, volume I. Final report, NTIS PB84-185552. Washington, USA, 204607 United States Environmental Protection Agency; 1983. US EPA. Estimation of infiltration rate in the vadose zone: compilation of simple mathematical models Volume I. United States Environmental Protection Agency EPA/600/R-87/128a, February 1998a. US EPA. Estimation of infiltration rate in the vadose zone: application of selected mathematical models Volume II. United States Environmental Protection Agency EPA/600/R-87/128b, February 1998b.