1. Introduction
The Regional Hydro-Ecological Simulation System (RHESSys) is a hydroecological model designed to simulate integrated water, carbon, and nutrient cycling and transport over spatially variable terrain at small (first-order streams) to medium (fourth- and fifth-order streams) scales. The model is structured as a spatially nested hierarchical representation of the landscape with a range of hydrological, microclimate, and ecosystem processes associated with specific landscape objects at different levels of the hierarchy. This approach is designed to facilitate environmental analysis that requires an understanding of within-watershed processes as well as aggregate fluxes of water, carbon, and nitrogen.
As a hydrologic model, RHESSys is intermediate in terms of complexity. Unlike lumped parameter hydrologic models such as the Identification of unit Hydrographs and Component flows from Rainfall, Evaporation, and Streamflow data (IHACRES; Evans and Jakeman 1998) or empirical curve number approaches, RHESSys explicitly models connectivity and lateral hydrologic fluxes between landscape units within a watershed. The representation of the vertical soil profile, however, is based on a fairly simple two-layer model with a single unsaturated and saturated zone. Additional hydrologic stores include a litter layer, surface detention store, multiple canopy interception layers, and a snowpack. Process-based hydrologic models such as MIKE-SHE (Refsgaard and Storm 1995) are at the more complex end of the continuum and provide a 1D Richard's equation solution to drainage through multiple soil layers down to the saturated zone. In addition to hydrology, RHESSys is able to model feedbacks between hydrology and ecosystem carbon and nutrient cycling, including the growth of vegetation. Other process-based eco-hydrologic models such as Macaque (Watson et al. 1999) and Topog (Vertessy et al. 1993) also provide this capability. In each of these models, representation of specific processes may differ but these models are relatively similar in terms of overall complexity.
The version of RHESSys described in this paper has evolved from the integration of stand-level ecosystem models with methods to compute the distribution and flux of soil water at the landscape level. These earlier versions of RHESSys were designed to explicitly couple the Forest Biogeochemical Cycles (FOREST-BGC) canopy model (Running and Coughlan 1988) with landscape-level patterns of critical meteorological forcing (Running et al. 1987) and later with hydrologic processes using the TOPMODEL (Beven and Kirkby 1979) hydrologic model. The first approach to distribute ecosystem processes at the landscape level involved gridding the “topclimatic” logic of Mountain Climate Simulator (MTN-CLIM) with FOREST-BGC for a 1200 km2 watershed in western Montana (Running et al. 1987). Later versions of RHESSys followed a generalization of FOREST-BGC to multiple biomes as Biome biogeochemical cycles Model (BIOME-BGC) (Running and Hunt 1993). Hydrologic modeling studies using RHESSys have included analysis of model sensitivity to landscape representation (Band 1993; Band et al. 1993) and using the model to explore the sensitivity of hydrologic response to climate change (Baron et al. 1998).
This current version of RHESSys continues to follow the basic BIOME-BGC framework. As discussed in this paper, however, many submodels used for specific processes have been altered and/or extended, largely to improve the soil biogeochemical process representation and expand canopy representation to consider both understory and overstory layers. Representation of soil organic matter decomposition in both RHESSys and BIOME-BGC is based largely on the CENTURY model. RHESSys also uses the CENTURYNGAS (Parton et al. 1996) approach to model (nitrogen) N-cycling processes such as nitrification and denitrification.
The current version of RHESSys has extended these hydrologic studies through the incorporation of an explicit hydrologic routing model in place of TOPMODEL. This current version has been used to explore the combined impact of roads and forest harvesting for several catchments in the Pacific Northwest (Tague and Band 2001; Krezek 2001), as well as in the Canadian boreal plain (Creed et al. 2000).
Coupled hydrologic-ecosystem and biogeochemical cycling studies using RHESSys include Mackay and Band (Mackay and Band 1997) who applied an earlier version of RHESSys to estimate carbon cycling in a small forested watershed. Mitchell and Csillag (Mitchell and Csillag 2000) are using the current version of RHESSys to explore spatially varying soil moisture controls on grassland productivity. Creed and Band (Creed and Band 1998) illustrate the use of similarity indices, derived using an earlier version of RHESSys, to compute a nitrate-flushing index based on saturation and a source index based on nitrate availability and dynamics in the saturation zone for a set of forested catchments in Ontario, Canada. By using these similarity indices, Creed and Band (Creed and Band 1998) were able to explain a significant amount of the variation in nitrate export between catchments and over time within the region. These indices, however, did not explicitly model N cycling and transport dynamics. Explicit modeling of N cycling and transport dynamics is included in the current version of RHESSys and has been tested for a small forested watershed as part of the Baltimore Ecosystem Study (BES; Band et al. 2001). The current version is also being used to explore spatial variation in N cycling in the Rocky Mountains (Laundrum et al. 2002), N saturation in the Smokey Mountains (Webster et al. 2001), and to examine the alteration of N sources and sinks along hydrologic flowpaths along a gradient of urbanization as part of the BES (Tague et al. 2000).
2. Implementation and model structure
One of the unique features of RHESSys is its hierarchical landscape representation. This approach allows different processes to be modeled at different scales and allows basic modeling units to be of arbitrary shape rather than strictly grid based. RHESSys is also structured using an object-based design approach to facilitate algorithmic substitution. Additional details on model hierarchical structure and coding implementation can be found in Band et al. (Band et al. 2000).
RHESSys partitions the landscape such that each level of the spatial hierarchy fully covers the spatial extent of the landscape. Spatial levels define a containment hierarchy with progressively finer units. Each spatial level is associated with different processes modeled by RHESSys and with a particular scale. At the finest scale, patches are typically defined on the order of meters squared, while basins (km2) define the largest scale.
Within RHESSys, a given spatial level is defined as a particular object type with a set of state (storage) and flux variables, process representations (equation sets), and an associated set of model parameters. For example, the estimation of atmospheric flux variables such as radiation occurs at the zone level. Thus, in deriving the set of spatial objects for a given simulation, zones are chosen to represent areas of similar climate or atmospheric-forcing conditions. The advantage of this hierarchical approach is that it allows different processes, that is, climate versus canopy processes, to be modeled at different spatial and temporal scales. It also allows modeling to occur on ecologically meaningful units as opposed to arbitrarily defined grid cells.
The definition of modeling units is done by the user prior to running the simulation. Although the user is given considerable flexibility in choosing a partitioning strategy for the different levels, partitioning should be tailored to take advantage of the patterns of relevant variability within the landscape and, in the case of patches, to maintain a coherent and solvable flow network. This permits efficient parameterization and reduces the error associated with landscape partitioning. Band et al. (Band et al. 1991), Lammers et al. (Lammers et al. 1997), and Tague et al. (Tague et al. 2000) provide further justification for and discussion of partitioning strategies, although this is an area requiring additional development.
2.1. Basins
Basins are defined as hydrologically closed drainage areas and encompass a single stream network. Basins typically serve as aggregating units to determine net fluxes of carbon, water, and nitrogen over the entire study area.
At present, RHESSys accounts for routing (storage and flux) of water within hillslopes to the stream. Once water reaches the stream, however, it is assumed to exit the basin within a single (daily) time step. For larger basins or shorter time steps, in-stream channel processes can be important. Future versions of RHESSys will likely use basins to organize the combination of hillslopes and stream reaches as separate object types and include a channel-routing model for stream reaches within the basin stream network.
2.2. Hillslopes
Hillslopes define areas that drain into one side of a single stream reach. Explicit routing between patches is organized at the hillslope level to produce streamflow. Hillslopes will usually be derived based upon drainage patterns using (Geographic Information System) GIS-based terrain-partitioning algorithms such as r.watershed in GRASS (Geographic Resources Analysis Support System) GIS environment and as described in Lammers and Band (Lammers and Band 1990). Like basins, hillslopes can also be used to aggregate sublevel processes. The spatial redistribution of patch soil moisture is organized at the hillslope level and, for simulations that use TOPMODEL for lateral flow distribution, a hillslope-level base flow is defined.
2.3. Zones
Zones denote areas of similar climate. Zone objects contain meteorological variables and use topclimate extrapolation methods necessary to characterize spatial variation in these variables. Each zone is linked to a particular set of climate input files. Thus, a given landscape may use data from multiple meteorological stations or atmospheric model grid cells if this information is available. Data from a particular station are modified based on zone elevation, slope, and aspect relative to the input climate station. Zone processing also estimates additional climate variables that may not be available from base climate station information such as vapor pressure deficit. Numerous strategies exist to partition areas of similar climate. Elevation bands in a mountainous area, for example, are likely to denote areas of similar climate and exposure as discussed in Lammers et al. (Lammers et al. 1997). The distribution of climate stations can also be used to define zone partitioning, where each zone defines the area associated with a particular climate station.
2.4. Patches
Patches represent the smallest-resolution spatial unit and define areas of similar soil moisture and land-cover characteristics. Vertical soil moisture processing and soil biogeochemistry are modeled at the patch level. Patch variables include fluxes such as infiltration, saturation zone recharge, and soil nutrient cycling. Patches are often derived using an overlay of several different maps, such as a wetness index, vegetation cover, and stream and road network layers. Patch definition, however, must also be designed to maintain the underlying topographic controls on drainage patterns in the watershed. Patches can be defined that contain a stream channel and parameterized to regulate land-channel drainage flux. In landscapes modified by humans, patches may also be defined to contain a stream channel, road segments, or storm sewers that provide specific drainage conditions. Human sources of water and nutrients including irrigation, fertilization, and septic system input are also defined at this level.
2.5. Canopy strata
Canopy strata are a separate object type but they define vertical, aboveground layers rather than horizontal spatial layers. The spatial resolution of the canopy stratum is defined by the patch partitioning. Processes such as photosynthesis and transpiration are modeled at the canopy stratum level. Each stratum corresponds to a different layer such as overstory or understory in the canopy structure. The user defines the number of vertical layers. A height state variable is associated with each layer and defines its processing relative to the other layers. Incoming radiation, precipitation throughfall, and wind are extinguished through the multiple layers according to the height and vegetation characteristics of each layer. RHESSys also permits multiple strata at the same height. This allows mixed vegetation types within the same spatial area to be represented. Finally, a litter layer is defined at the patch level and receives input from the overlying canopy layers. This litter layer acts as the interface between each patch and its associated canopy strata layers. For nonvegetated patches, a dummy litter layer (with no attenuation or processing of hydrologic, energy and carbon/nutrient fluxes) is used.
2.6. Interface
Parameterization and management of RHESSys is complex due to the multiple levels of spatial partitioning and the associated parameter sets. Most of these parameters are derived from topographic, land-cover, and soil map layers. Associated with RHESSys are a number of interface programs, which organize input data into the format required by the simulation model. These include standard GIS-based terrain-partitioning programs and RHESSys-specific programs that derive landscape representation from GIS images and establish connectivity between spatial units. These various programs can be run in stand-alone mode or as part of an integrated RHESSys interface, RAINMENT. Additional details about the RHESSys interface can be found in Band et al. (Band et al. 2000).
Key inputs into RHESSys include (see Figure 1) the following.
A description (worldfile) of the landscape representation and initial-state variables associated with each component of the spatial hierarchy (i.e., basins, hillslopes, zones, etc.) Because of the length and spatial complexity incorporated in this description, GRASS and ARCVIEW GIS-based programs (GRASS2WORLD, ARCVIEW2WORLD) were developed to generate this file.
The flow table describes connectivity between patches within a hillslope when the explicit routing approach is used to model distributed hydrology. The flow table is also generated automatically prior to running the main RHESSys simulation using a software program (CREATE-FLOWPATHS).
The temporal event control (TEC) file describes the timing and nature of temporal events that will occur during the course of the simulation, that is, disturbances such as forest harvesting, fire, or road construction. Temporal events refer to events that initiate a change in landscape-state variables or parameterization or new processes in the simulation sequence. A spinup period is typically run prior to output to remove transient behavior due to initialization.The TEC file is also used to control data assimilation for output.
Time series inputs include a range of climate variables as discussed in section 3, along with descriptions of the station at which this information was collected. Each zone in the spatial hierarchy is associated with a particular climate station. Note that base station coverage can be defined at any level of the hierarchy and need not be spatially contained by hillslopes, zones, or patches. Input stations can also be assigned at the patch level to organize fertilizer and irrigation inputs.
Parameter files are associated with each level of the spatial hierarchy. A library of commonly used parameter files assigned to specific soil and vegetation types is available. The current library of parameter files include standard parameters for conifer, deciduous, grass, chaparral, tundra, as well as a number of more refined files for specific species groups such maple, pine, spruce, oak, aspen, or douglas fir. The current library of parameter files include all major soil textures, such as clay, sandy loam, etc.
2.7. Implementation and future development
This current version of RHESSys is implemented in the C programming language using an object-based approach to facilitate the substitution of different process algorithms. All of the processes as well as input/output routines are contained in separate procedures with appropriate names. This allows for relatively easy modification of specific process algorithms. Data structures are similarly defined and named to maintain clarity about the level of the spatial hierarchy associated with a specific algorithm.
The object-based approach and hierarchical landscape representation are also designed to facilitate future RHESSys development. Structures exist to allow for the development of subdaily time step models for certain processes such as infiltration. Object-based landscape representation also facilitates the implementation of different land covers such as those found in urban areas. Hillslope-level organization of drainage and particularly the implementation of roads as mechanisms by which flow can be redirected serve as a foundation for implementation of other man-made controls on flowpaths such as sewers.
3. Atmospheric and environmental forcing
Atmospheric variables or external forcing variables drive RHESSys hydrology and biogeochemical cycling. These include climate variables such as daily temperature and precipitation as well as material inputs such as atmospheric nitrogen deposition. Climate input and processing in RHESSys is done at the zone level. Each zone is assigned a particular base station that manages external time series inputs to the model. Multiple zone objects can be assigned to the same base station; however, each zone must have a unique base station associated with it. The appendix lists current atmospheric variables that are used internally in the model. For most of these variables, the user has the option of providing a daily time series as input (associated with its base station) or allowing RHESSys to estimate the daily value from user parameters. Daily temperature and precipitation, however, must be input by the user. Input time series may be derived from field observation or from coupling with external models. Internal estimation of most meteorological variables, including incoming radiation, is based on algorithms from the MTN-CLIM model (Running et al. 1987).
Currently, nitrogen and precipitation are the only material inputs into the model. Atmospheric nitrogen deposition as nitrate and as ammonium are input as separate time series in RHESSys. Wet and dry deposition, however, are not distinguished in the current model. If a time series of nitrogen deposition is not provided, a constant daily value for nitrate based on a zone parameter will be used. In this case, deposition as ammonium is assumed to be zero.
3.1. Temperature
Daily minimum and maximum temperature time series must be included in the base station assigned to each zone object. Note that these time series may be derived from field observations or from an external mesoscale atmospheric model such as the Regional Atmospheric Modeling System (RAMS; Walko et al. 2000) or external climate interpolation schemes such as the Parameter-Elevation Regressions on Independent Slopes Model (PRISM; Daly et al. 1994) or Daymet (Thornton et al. 1997). For each zone, the base station temperatures are scaled by a lapse rate with elevation (included as a zone parameter).
3.2. Precipitation
A daily precipitation time series must also be included in each base station and can be adjusted using an isohyetal multiplier assigned to each zone. Orographic patterns in precipitation can be modeled using this isohyetal multiplier. Rainfall duration defaults to the entire day, unless a rainfall duration time series is included as a base station input. The user also has the option of providing hourly, rather than daily, rainfall input and running the hydrologic portion of the model in an hourly mode.
The user may input a snowfall time series directly. If a snowfall time series is not included, precipitation is partitioned into rain and snow by assuming a linear transition from snow to rain across a temperature range defined by
3.3. Vapor pressure deficit
Saturation vapor pressure and relative humidity can be input directly. If these observations are not available from the base station climate file, saturation vapor pressure and vapor pressure deficit are estimated from standard air temperature–vapor pressure relationships (Jones 1992).
3.4. Incoming radiation
Canopy radiation interception (see section 5.2.) depends upon top of canopy inputs of both diffuse (Kdiffuse′) and direct (Kdirect′) solar radiation as well as longwave radiation. These radiation streams can be input by the user as part of the climate base station or can be estimated internally in the model. Within RHESSys, a solar geometric-based estimate of atmospheric radiation is computed based on site latitude, zone slope, aspect, and east–west horizon values. Incoming radiation is then adjusted to account for atmospheric transmissivity, estimated based on daily temperature variation and precipitation. These equations are not described here since they directly follow the MTN-CLIM approach (Running et al. 1987).
3.5. Atmospheric CO2
In the current version of RHESSys, atmospheric CO2 concentration is held constant at 370 ppm. (Note: CO2 concentration is used in estimates of stomatal conductance.) It is included as a state variable, however, such that it can be readily modified in future versions.
4. Soil hydrologic processes and transport mechanisms
Vertical and lateral soil moisture and associated nutrient fluxes are modeled for each patch object. Patch topology (i.e., connectivity between patches) is organized by basin objects.
To model vertical soil moisture processes, a simple three-layer model is used, which includes a surface detention store, an unsaturated store, and a saturated store. Snowpack and litter stores are also included.
4.1. Detention storage and infiltration
If a daily time step is used, it is assumed that all inputs (precipitation and snowmelt) are distributed evenly throughout the day. In regions characterized by high-intensity, short-duration rainfall, this approach will overestimate infiltration rates. In these cases, the user may choose to input a rainfall duration time series, which will override the assumption of a daily duration.
Ponded water that is not infiltrated within the daily time step becomes detention storage. Detention storage beyond a surface detention storage capacity parameter will become overland flow.
4.2. Hydraulic conductivity and porosity profiles
4.3. Vertical unsaturated zone drainage
Here Kunsat(z) can be estimated using either vanGenuchten and Nielsen (vanGenuchten and Nielsen 1984) or Clapp and Hornberger (Clapp and Hornberger 1978) models of soil characteristics.
To compute a depth of unsaturated zone soil moisture at field capacity [θfc in Equation (10)], the relative saturation at field capacity is integrated over the porosity profile from the surface to the water table depth (zs).
4.4. Soil evaporation
Soil evaporation is limited both by energy and atmospheric drivers and by a maximum exfiltration rate as a function of soil properties at a given soil moisture.
Final soil evaporation is a minimum of rates based on the Penman potential evaporation and soil exfiltration rates.
4.5. Capillary rise
Capillary rise is limited to filling unsaturated zone to field capacity. To correct for subdaily plant responses, one-half of the potential capillary rise, computed using Equation (17) is allocated to the unsaturated zone at the start of the day. The remaining potential is available later in the day to satisfy plant transpiration demands (see section 5.6.).
4.6. Snow accumulation and melt
Snow accumlation is based on incoming precipitation and is assumed to fall evenly over each climate zone. Modeling case studies using previous versions of RHESSys, however, have shown that in snow-dominated environments, the redistribution of snow by wind, can have a significant impact on hydrology (Hartman et al. 1999). In the current model, we do not account for this effect; however, an adaptation of a more sophisticated snow accumulation and melt approach will likely be included in future versions.
4.7. Lateral redistribution
Soil moisture redistribution through saturated throughflow and associated runoff production can be modeled using either a quasi-spatially distributed model, TOPMODEL (Beven and Kirkby 1979) or via an explicit routing model, which is a modification of the Distributed Hydrology Soil Vegetation Model (DHSVM; Wigmosta et al. 1994).
TOPMODEL is applied at the hillslope level. DHSVM is applied at the basin level, since it can include some limited streamflow routing. Both approaches are executed at the end of the day, following vertical soil moisture updates.
4.7.1. TOPMODEL
Alternatively, a constant or user-defined profile of hydraulic conductivity can replace the assumption of an exponential decay.
4.7.2. Explicit routing
Alternatively, the explicit routing model is based on the DHSVM (Wigmosta et al. 1994) routing approach that has been modified to account for nongrid-based patches and nonexponential transmissivity profiles. Similar to TOPMODEL, DHSVM assumes that hydraulic gradients follow surface topography. Flow topology is generated by a GIS-based preprocessing routine, CREATE-FLOWPATHS. Multiple flow directions, from any given patch, are permitted.
In the explicit routing approach, three distinct patch types are considered: streams, roads, and land surface. In the current implementation, stream patches include the riparian area adjacent to the stream. Thus vertical processes such as infiltration are modeled using the same algorithms applied for land surface patches. Lateral flow from the stream patch, however, is assumed be channelized. In the current implementation, all channelized flow is assumed to exit the basin in a single time step. Future implementation, however, will include an in-stream routing model between stream patches. Unique characteristics of lateral flow from road patches are discussed below.
Within the model time step, explicit routing is computed for a user-specified time step to achieve stability. This can be useful to maintain a Couret number less than 1 and reduce errors due to over/underestimation of downslope flood wave propagation. All vertical fluxes and storage adjustments including rainfall infiltration are done before the routing of lateral subsurface throughflow. If, however, hourly precipitation data are available, the vertical hydrologic flux portion of the model can be run at a corresponding hourly time step.
For grids, flow widths are assumed to be 0.5 times the grid size for cardinal directions and 0.354 times the grid size for diagonal directions after Quinn et al. (Quinn et al. 1991). For irregular elements, flow widths are summed along the shared boundary between patches a and b.
Surface flow (i.e., saturation overland flow or Hortonian overland flow) produced is routed following the same patch topology used from routing saturated subsurface throughflow. All surface flow produced by a patch is assumed to exit from the patch within a single time step. If the receiving patch is not saturated, surface flow is allowed to infiltrate based on Equation (5) and is added to unsaturated soil moisture storage. Patch routing is sequenced to occur from the uppermost patches first.
Within the explicit routing model, patches that contain roads are treated as special cases.
4.7.3. Road patches
Roads have been shown to alter the routing of both overland and subsurface throughflow (Wemple et al. 1996; Luce and Cundy 1994). Road culverts produce channelized flow that in some cases can connect directly to the stream and effectively extend the stream drainage network. Flow in road culverts is produced from two sources: 1) runoff from the road surface and 2) interception of subsurface routing by the road cut bank. At present, patches can be parameterized using an areally weighted conductivity. If resolution is such that the road covers a significant portion of patch surface, then the infiltration capacity parameter assigned to that patch should reflect low infiltration capacities associated with roads.
The amount of saturated subsurface flow intercepted by the road is a function of the road cut depth and the current saturation deficit. If the road cut bank depth is less than depth to saturation, none of the saturated throughflow is intercepted by the road. If road cut depth is greater than depth to saturation, all subsurface throughflow produced above the road cut depth is captured by the road culvert.
Use of an alternative receiving patch allows flow intercepted by a road to be routed either to an adjacent patch or to a stream/storm network. The former represents the situation in which culverts serve to concentrate flow but allow this flow to be redistributed and reinfiltrated in down-slope patches. The latter models the case in which culverts form part of an extended drainage network. The location of patches to receive flow intercepted by culverts are specified as part of the routing topology in the CREATE-FLOWPATHS preprocessing routine prior to RHESSys execution. A new routing topology can be read in during execution, however, so that disturbances can be modeled.
4.7.4. Vertical and lateral redistribution of nitrate
This approach to modeling nitrate export follows the flushing hypothesis. Use of an exponential profile distribution of nitrate, results in greater supplies of mobile N as saturation throughflow levels rise in the soil, and begin tapping more nutrient-rich near-surface soils. Field investigation of N-transport mechanisms (e.g., McDonnell 1990; Peters et al. 1995) suggests that this may be a reasonable assumption in areas, such as humid forests, where much of the infiltrating water rapidly moves through the unsaturated layer through preferential flowpaths, without significant matric contact. Further model development will investigate the marginal gains made by adopting a more rigorous vertical transport representation.
When using the TOPMODEL approach for soil moisture redistribution, the spatial variation in nitrate export is ignored. For model applications that focus on nutrient export in spatially heterogeneous terrain, the explicit routing option is therefore recommended. When TOPMODEL is used, a basin-scale-lumped approach is used to estimate mobile N transport. A mean hillslope soil nitrate storage is computed and used to determine a hillslope-scale nitrate export based on a mean basin saturation deficit following Equation (36) above. To account for losses due to lateral N export, all patches are assigned the updated mean basin soil nitrate value.
5. Canopy radiative and moisture fluxes
RHESSys models surface radiation and rainfall attenuation through a series of canopy layers.
In vegetated catchments, these layers correspond to overstory and understory vegetation. In urbanizing catchments, nonphotosynthesizing layers, such as buildings, can be represented to model the evaporation of intercepted water. For vegetated canopy, absorbed radiation also controls carbon/nitrogen cycling through photosynthesis and respiration. Carbon and nitrogen cycling are discussed in section 6.
Canopy layers are processed sequentially according to height. Radiation, wind, and rain or snow throughfall are attenuated as they are absorbed/intercepted by each successive layer. Layers at equal height share the same environment. Canopy layers at the same height must have combined fractional coverage less than or equal to 1.
5.1. Leaf area and plant area index
Canopy layer leaf area index (LAI) reflects current leaf carbon storage cs.leaf, scaled by LAIsp, specific leaf area index that varies with vegetation type.
where cs.live.stem and cs.dead.stem are sapwood and heartwood stem carbon stores, respectively, and swa is a vegetation-specific parameter that defines stem wood equivalent area per unit of carbon. For grasses, PAI = LAI.
5.2. Radiation attenuation
Incoming direct and diffuse radiation at the top of the canopy is input directly or estimated as described in section 3.4. Canopy radiation attenuation and absorption by each canopy layer is modeled separately for diffuse, direct, and photosynthetically active radiation (PAR) radiative fluxes.
5.2.1. Direct radiation
5.2.2. Diffuse radiation
5.2.3. PAR radiation
Absorbed PAR (APARdirect; APARdiffuse) are calculated using Equations (41) and (42) above where reflectance coefficients, α and direct radiation extinction coefficients K are replaced by PAR-specific coefficients and LAI replaces PAI.
Radiation inputs per LAI, which are used to drive leaf conductance and photosynthesis submodels, are computed separately for sunlit (ppdfsunlit) and shaded leaves (ppfdshade).
5.3. Aerodynamic resistance
Aerodynamic resistance is computed separately for the top (overstory) canopy and understory layers following the model developed by Heddeland and Lettenmaier (Heddeland and Lettenmaier 1995). This model assumes a logarithmic wind speed decay profile to the top of the canopy and an exponential decay profile within the canopy. A patch-level stability correction is included based upon Oke (Oke 1987).
5.4. Canopy conductance
Finally, the effect of CO2 augmentation on stomatal conductance has not yet been implemented. Many field experiements have shown that increasing atmospheric CO2 concentration may improve plant water use efficiency by reducing stomatal conductance (Medlyn et al. 2001). These relationships will be included in RHESSys in a subsequent version.
Nonvascular conductance
5.5. Interception
5.6. Evapotranspiration
Total evaporative fluxes from each canopy layer may include the evaporation of water intercepted by the canopy, sublimation of intercepted snow, and transpiration by vascular layers. Both evaporation and transpiration rates are computed using the standard Penman–Monteith equation (Monteith 1965).
5.7. Litter interception and evaporation
In the current version, litter PAI is set to 1.0 if litter carbon storage is greater than 0 and 0 otherwise.
Evaporation from the litter is computed based on radiation available following attenuation through canopy layers. Evaporation is computed using the Penman approach, where gs is set to an arbitrarily large value (10 000) and ga is based on a reduction of wind speed through the canopy to the litter layer as discussed in section 5.3.
Note that in cases where detention storage capacity is greater than 0, water in excess of litter interception capacity may be collected at the surface and evaporated.
6. Carbon and nitrogen cycling
The main structure of carbon cycling in RHESSys is based on BIOME-BGC (Thornton 1998) although many of the specific algorithms have been extended and/or modified.
Carbon and nitrogen cycling associated with live vegetation (e.g., photosynthesis, respiration) is included in canopy strata object routines. Carbon and nitrogen cycling associated with litter and soil layers (e.g., decomposition) occurs within patch objects. Thus, all vegetation strata associated with a particular patch contribute and extract material (carbon, water, and nitrogen) to and from the same well-mixed soil and litter pools.
RHESSys carbon and nitrogen stores are partitioned into leaves, roots, stems, and coarse roots. Stem and coarse-root stores also include both live and dead wood components to account for differences in respiration and C:N ratios. For grasses, stores are restricted to leaves and fine roots and include an additional store used to account for dead biomass that remains standing. Vegetation nitrogen stores follow carbon stores based on stoichiometric relationships discussed in more detail below.
6.1. Vegetation: Carbon cycling
6.1.1. Photosynthesis
Finally, the Farquhar model takes into account the control exerted by the leaf nitrogen content. Photosynthesis is also constrained by the amount of nitrogen in the soil that is available for uptake by the plant. As computed, gpsn is a potential photosynthesis subject to the availability of nitrogen. The amount of nitrogen required, however, depends upon the allocation strategy used by the plant since different plant components (i.e., leaves versus stems) have different C:N ratios. The allocation strategy is discussed in section 6.1.4. below. Total nitrogen required from the soil, potential.plant.N.uptake, is determined based on gpsn and the allocation strategy. If sufficient mineralized nitrogen is available, gpsn is used for plant allocation. If, however, nitrogen is limiting, gpsn is reduced until the nitrogen requirements can be met. Section 6.3.1. discusses how soil nitrogen availability is determined.
6.1.2. Respiration
As in BIOME-BGC total maintenance respiration totalmr integrates respiration for each live carbon store including leaves, roots, and stems. Respiration for each component is computed as a function of nitrogen concentration and the current air temperature using the model developed by Ryan (Ryan 1991).
Growth respiration is also computed and subtracted from the carbon allocated to each vegetation component. Growth respiration is computed as a fixed percentage grperc of new carbon allocation, where grperc is a strata-specific (i.e., vegetation type) parameter. Ryan (Ryan 1991) suggests a growth respiration rate grperc of 25% for trees.
6.1.3. Phenology
In RHESSys, as in BIOME-BGC, net photosynthesis can be allocated on a daily or annual basis. Each day, netpsn is partitioned to the various tissues (leaves, roots, and stems). Partitioning between these various tissues is discussed below. The vegetation-type parameter alloc.prop.day.growth sets the percentage of newly assimilated carbon, which is expressed on a daily basis. The remaining carbon is stored and expressed during an annual leaf-out period. For deciduous trees and grasses, alloc.prop.day.growth must be substantially less than 1 to ensure enough stored carbon for spring leaf out. The timing and length of the annual leaf out period is also set in the vegetation-type parameter file, using parameters day.leaf.on and ndays.expand, respectively. Note that although this period is referred to as leaf out, it also controls the assimilate allocated annually to roots. Carbon stored for annual allocation from the previous year is expressed during the leaf-out period. For each day during the leaf-out period, stored carbon is transferred to leaf and fine root pools such that amounts transferred to each component decreases linearly over each day of the leaf-out period.
In addition to annual leaf-out periods, leaf and fine root turnover periods are defined, again using a fixed timing set by stratum-specific parameters (day.leaf.off and ndays.litfal). The amount of leaf and fine root carbon transferred to the litter pool during this period is a fixed percentage (leaf.turnover and froot.turnover, respectively) of current leaf and fine root carbon pools. Parameters are set in a stratum parameter file. For deciduous trees, however, the entire leaf carbon pool is transferred to the litter pool. For grasses, the leaf carbon pool is first transferred to a standing dead leaf carbon pool, which acts as an intermediate store. Transfer from standing dead leaf carbon to the litter pool occurs at the rate defined by deadleafturnover (percentage of dead leaf turnover per year). For all vegetation types, the amounts of leaf and fine root turnover during the fall turnover period follows a linearly decreasing daily transfer schedule similar to that used for leaf out during the spring.
Future versions of RHESSys will incorporate a variable phenology model (i.e., White et al. 1997) to account for environmental (i.e., temperature, radiation, etc.) controls on the timing of spring leaf out and fall leaf drop. This will also allow the model to account for plant responses to environmental stress through increases in leaf fall.
Different vegetation types show different sensitivities to daily versus annual allocation and the timing of phenology. Conifers, because they lose only a portion of their leaves, are less sensitive. For deciduous trees and grasses, some portion of photosynthesis must be reserved for annual allocation (i.e., alloc.prop.day.growth must be less than 1) to restart photosynthesis during the spring and these vegetation types are typically more sensitive to phenology timing.
For trees, the sapwood components of stem and coarse root pools must also be transferred to heartwood stem and coarse root pools. The rate of sapwood to heartwood turnover is also based on a fixed percentage (livewoodturnover) of the associated sapwood pools. Loss of carbon from heartwood pools can only occur due to whole tree mortality (discussed below).
6.1.4. Allocation
New net photosynthesis must be partitioned between roots, stems, and leaves or in the case of grasses, between roots and leaves. BIOME-BGC (Thornton 1998) maintains a fixed partitioning strategy such that the ratio of carbon allocated to each store is constant and set by species-specific parameters.
Once the fraction allocated to roots has been determined, the remaining assimilate is allocated to leaves and, in the case of trees, stem wood based on a fixed ratio (allocstemc.leafc). For trees, carbon allocated to stem wood is further partitioned into coarse-root and stem components and further into live and dead wood components based on fixed ratios (alloccrootc.stemc and alloclivewood.woodc, respectively).
Allocations to fine and coarse roots and stems follow a similar approach.
6.1.5. Alternative allocation strategy
Once these initial allocation fractions are computed, the impact of soil nitrogen limitations and finally allocation amounts are computed using the same approach described above.
6.1.6. Mortality
An annual plant mortality rate as a fixed percentage of current biomass is set in the stratum parameter file (mortality). Total carbon and nitrogen to be lost due to plant mortality is set each year. The same percentage is taken from each of the available tissue stores (i.e., leaves, roots, stem). Carbon that is lost from leaves and fine roots is transferred to the litter pool. Coarse stem and root material is transferred to a coarse wood debris pool that decays at a species-specific fragmentation rate before it is transferred to litter carbon and nitrogen pools. Fragmentation does not alter wood C:N ratios.
Material transfers due to mortality occur on a daily basis such that the annual plant mortality rate is maintained. The annual rate is reset each year (at the start of leaf out) to respond to changes in available stores. Future versions of RHESSys may incorporate a variable mortality rate as a function of environmental stressors that increase susceptibility to disease, blow down, etc. In the current version, episodic changes in vegetation such as forest harvesting or fire can be implemented through a dynamic redefinition of the stratum-level carbon- and nitrogen-state variables. An independent disturbance model can be used to determine size and frequency of these events.
6.2. Vegetation: Nitrogen cycling
In general, the cycling of nitrogen within the vegetation is tied stoichiometrically to that of carbon. In RHESSys, the C:N ratios of the various plant biomass components are fixed based on species-specific input parameters. A number of studies have shown that plants may vary biomass C:N ratios, particularly of leaves, in response to environmental stress (Aber and Melillo 1991). Thornton (Thornton 1998), on the other hand, summarizes studies that suggest that plants respond to nitrogen limitations through variations in total leaf area and leaf area per unit of nitrogen. At present, we follow the BIOME-BGC approach and hold leaf and other C:N ratios constant. Within the simulation, however, carbon and nitrogen stores and fluxes are maintained separately to facilitate future implementation of algorithms to account for differences in C:N ratios in response to stress.
One exception to the above implementation occurs during leaf fall. Retranslocation of leaf nitrogen as leaves fall results in increased C:N ratios in litter fall and excess nitrogen stored within the plant. This retranslocated stored nitrogen is then available for plant use during subsequent growth (i.e., in addition to nitrogen available through uptake from the soil). Separate parameters are therefore required to set leaf litter and leaf C:N ratios.
6.3. Soil
6.3.1. Decomposition
Daily soil and litter decomposition models are based on an approach developed by Thornton (Thornton 1998) for use in BIOME-BGC, which is similar to the approach used by CENTURY (Parton et al. 1996). Decomposition is based on a set of litter and soil pools, each of which includes both organic material and microbial biomass. Each pool is associated with a specific C:N ratio and a potential decay rate. This decay rate includes both carbon lost due to microbial respiration and carbon transferred to the next soil/litter pool. Respiration is computed as a percentage of decomposition rates and are specific to each soil/litter pool (based on Thornton 1998).
6.3.2. Nitrification
Only mineralized soil nitrogen (i.e., from the decomposition of soil organic material; soil.sminn or fertilizer) is used for nitrification. This ensures the availability of soil carbon for the microbial nitrification processes (Gold et al. 2001).
6.3.3. Denitrification
7. Application: Hydrologic controls on N cycling for a small forested catchment
Specific applications of RHESSys highlight different components of model structure and function. Hydrologic applications of this model focus on predictions of streamflow and spatial patterns of soil moisture under a variety of land-cover and climate forcing conditions. Biogeochemical applications have used the model to examine the mechanisms that control the cycling and transport of both carbon and nitrogen within a watershed. As an example, a brief overview of a current RHESSys application is presented here. Additional detail on this study can be found in Band et al. (Band et al. 2001).
Pond Branch is a 34-ha catchment used as the control forested catchment as part of the Baltimore Long-Term Ecological Research program. As part of this project, continuous streamflow and weekly stream chemistry have been monitored since fall 1998. Watershed topography is characterized by a gently sloping upland area with steep side slopes draining to a broad riparian area. Upland soils are silt loam soil with underlying deep saprolite. Side slopes contain very shallow soils, while the bottomland is characterized by a substantial organic soil layer. Vegetation is dominated by a mature oak–hickory forest.
Because this is a relatively small watershed, patches were based on 5-m pixels. Hillslope delineation is shown in Figure 2. Soil parameters appropriate for sandy loam and vegetation parameters for an oak–hickory forest were used. When RHESSys is run in dynamic mode (i.e., soil and vegetation are not prescribed but are computed by the model), the model must be run for a period of spinup to generate carbon and nitrogen pools. In this case, the model was spun up for 200 years, at which point both canopy cover and soil carbon pools had stabilized (i.e., reached equilibrium values). Once stable carbon and nitrogen pools were achieved, United States Geological Survey (USGS) stream gauge records were used to calibrate the model by varying Ksat0, hydraulic conductivity at the surface and m, the decay of hydraulic conductivity with depth. Calibration used a Monte Carlo–based approach to run the model over a range of values for m and K. The set of values producing the optimal correspondence between observed and modeled streamflow as measured by the highest Nash–Sutcliffe efficiency (Nash and Sutcliffe 1970) were selected for future runs.
Correspondence between modeled and observed streamflow as well as modeled and observed stream nitrate concentration illustrate both the strengths and weakness of RHESSys application in this particular watershed (see Figures 3 and 4). As discussed in an earlier paper (Band et al. 2001), RHESSys tends to overpredict nitrate concentration in this watershed, although it captures the temporal pattern of nitrate export. RHESSys does not model in-stream processes, thus overprediction of the nitrate export may suggest that in this watershed in-stream losses are significant. Field data collected as part of this project support this conclusion. One of the surprising patterns in both field and modeled nitrate export is the increase in N concentration during the summer when biologic activity and hence biologic uptake of nitrogen is typically high. RHESSys output was used to help explain this pattern. Examining spatial outputs show that this increase in nitrification occurs primarily in the riparian region as it dries out during the summer period. The location of these processes within the riparian zones, which maintain connectivity to the stream throughout the summer, is important. Examining RHESSys spatial patterns of soil moisture illustrates temporal and spatial variation in surface saturated areas in which riparian zone processes occur (Figure 5). Modeled spatial patterns of soil moisture suggest that in addition to the classically defined riparian zone, topographic hollows, which drain into the riparian area, are also areas of significant saturation during parts of the year. These spatial patterns suggest that these topographic hollows may play a role similar to the riparian area in terms of both overall control on hydrology and nitrogen cycling, at least during the wetter periods. These results further suggest areas where additional field study may be needed to collect data in these areas, which potentially act as controls on water quality and quantity in this watershed.
Acknowledgments
The Baltimore Ecosystem Study project is supported by the NSF Grant GRS 0095796 to L. Band and C. Tague, and National Science Foundation Long-Term Ecological Research program Grant DEB 9714835. We thank the USDA Forest Service Northeastern Research Station and BES for site management. In addition we thank the University of Maryland, Baltimore County, for their contribution of office and laboratory space at the Research Technology Center. We also thank the Baltimore Department of Parks and Recreation and Department of Public Works, the Baltimore County Department of Parks, and the Maryland Department of Natural Resources for their support.
References
Aber, J. D. and J. M. Melillo. 1991. Terrestrial Ecosystems. Saunders College Publishing,. 429. pp.
Ambroise, B., K. J. Beven, and J. Freer. 1996. Towards a generalisation of the topmodel concepts: Topographic indices of hydrological similarity. Water Resour. Res 32:2135–2145.
Band, L. 1993. Effect of land surface representation on forest water and carbon budgets. J. Hydrol 150:749–772.
Band, L., D. Peterson, S. Running, J. Coughlan, R. Lammers, J. Dungan, and R. Nemani. 1991. Forest ecosystem processes at the watershed scale: Basis for distributed simulation. Ecol. Modell 56:171–196.
Band, L., J. Patterson, R. Nemani, and S. Running. 1993. Forest ecosystem processes at the watershed scale: Incorporating hillslope hydrology. Agric. For. Meteor 63:93–126.
Band, L., C. Tague, S. Brun, D. Tenenbaum, and R. Fernandes. 2000. Modelling watersheds as spatial object hierarchies: Structure and dynamics. Trans. Geogr. Info. Syst 4:181–196.
Band, L. and P. Groffman. 2001. Forest ecosystem processes at the watershed scale: Hydrological and ecological controls of nitrogen export. Hydrol. Processes 15:2013–2028.
Baron, J. S., M. D. Hartman, T. G. F. Kittel, L. E. Band, D. S. Ojima, and R. B. Lammers. 1998. Effects of land cover, water redistribution, and temperature on ecosystem processes in the South Platte basin. Ecol. Appl 8:1037–1051.
Beven, K. and M. Kirkby. 1979. A physically-based variable contributing area model of basin hydrology. Hydrol. Sci. Bull 24:43–69.
Chen, J., P. M. Rich, S. Gower, J. Norman, and S. Plummer. 1997. Leaf area index of boreal forests: Theory, techniques, and measurements. J. Geophys. Res 102:29429–29444.
Chen, J., J. Liu, J. Cihlar, and M. Goulden. 1999. Daily canopy photosynthesis model through temporal and spatial scaling for remote sensing applications. Ecol. Modell 124:99–119.
Clapp, R. and G. Hornberger. 1978. Empirical equations for some soil hydraulic properties. Water. Resour. Res 14:601–604.
Coughlan, J. and S. Running. 1997. Regional ecosystem simulation: A general model for simulating snow accumulation and melt in mountainous terrain. Landscape Ecol 12:119–136.
Creed, I. and L. Band. 1998. Exploring functional similarity in the export of nitrate-n from forested catchments: A mechanistic modeling approach. Water Resour. Res 34:3079–3093.
Creed, I., C. Tague, R. Swanson, and R. L. Rothwell. 2000. The potential impacts of harvesting on the hydrologic dynamics of boreal watersheds. Proc. American Geophysical Union 2000 Spring Meeting, Washington, D. C., Amer. Geophys. Union.
Croley, T. 1989. Veritable evaporation modeling on the Laurentian Great Lakes. Water Resour. Res 25:781–792.
Daly, C., R. Neilson, and D. Phillips. 1994. A statistical–topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor 33:140–157.
Dickenson, R., M. Shaikh, R. Bryant, and L. Graumlich. 1998. Interactive canopies for a climate model. J. Climate 11:2823–2836.
Dingman, S. 1994. Physical Hydrology. Prentice Hall, 646 pp.
Dunne, T. and L. Leopold. 1979. Water in Environmental Planning. W. H. Freeman and Company, 818 pp.
Eagleson, P. 1978. Climate, soil and vegetation. 3. A simplified model of soil moisture movement in the liquid phase. Water Resour. Res 14.722–730.
Evans, J. and A. Jakeman. 1998. Development of a simple, catchment-scale, rainfall–evapotranspiration–runoff model. Environ. Modell. Software 13:385–393.
Farquhar, G. and S. vonCaemmerer. 1982. Modeling photosynthetic response to environmental conditions. Encyclopedia of Plant Physiology O. Lange et al., Eds., New Series, Vol. 12B, Springer-Verlag, 549–587.
Gold, A., P. Groffman, K. Addy, D. Kellog, M. Stolt, and A. Rosenblatt. 2001. Landscape attributes as controls on groundwater nitrate removal capacity of riparian zones. J. Amer. Water Resour. Assoc 37:1457–1464.
Green, W. and G. Ampt. 1911. Studies on soil physics. 1: The flow of air and water through soils. J. Agric. Sci 4:1–24.
Hartman, M., J. Baron, R. Lammers, B. Cline, and L. T. Band. 1999. Simulations of snow distribution and hydrology in a mountain basin. Water Resour. Res 35:1587–1603.
Heddeland, I. and D. Lettenmaier. 1995. Hydrological Modeling of Boreal Forest Ecosystems. Water Resource Series, Vol. 14, Dept. of Civil Engineering, University of Washington, 38 pp.
Jarvis, P. 1976. The interpretations of the variations in leaf water potential and stomatal conductance found in canopies in the field. Philos. Trans. Roy. Soc. London B273:593–610.
Jones, H. 1992. Plants and Microclimate: A Quantitative Approach to Plant Physiology. Cambridge University Press, 452 pp.
Kelliher, F., T. Black, and D. Price. 1986. Estimating the effects of understory removal from a douglas fir forest using a two-layer canopy evapotranspiration model. Water Resour. Res 22:1891–1899.
Krezek, C. 2001. Climatic regulation of peak flows and their contributing source areas under natural and disturbed conditions in the carnation creek watershed, Vancouver Island, BC, Canada. M.S. thesis, University Western Ontario, Plant Science and Environmental Studies, 255 pp.
Lammers, R. and L. Band. 1990. Automating object descriptions of drainage basins. Comput. Geosci 16:787–810.
Lammers, R. and C. Tague. 1997. Scaling behaviour of watershed processes. Scaling-Up: From Cell to Landscape, P. van Gardingen, G. Foody, and P. Curran, Eds., Cambridge University Press, 295–317.
Landsberg, J. and R. Waring. 1997. A generalized model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning. For. Ecol. Manage 95:209–228.
Laramie, R. and J. Schaake Jr. 1972. Simulation of the continuous snowmelt process. Tech. Rep. 143, Ralph M. Parsons Laboratory, Massachusetts Institute of Technology, Cambridge, MA.
Laundrum, L., C. L. Tague, and J. Baron. 2002. Modeling hydrologic and biogeochemical processes in Loch Vale watershed using the regional hydroecological simulation system (rhessys). Proc. 36th Annual Geological Society of America, South-Central Section Meeting, Alpine, TX, Geological Society of America, South-Central Section.
Luce, C. and T. Cundy. 1994. Parameter identification for a runoff model for forest roads. Water Resour. Res 30:1057–1069.
Mackay, D. and L. Band. 1997. Forest ecosystem processes at the watershed scale: Dynamic coupling of distributed hydrology and canopy growth. Hydrol. Proc 11:1197–1217.
Manley, R. 1977. The soil moisture component of mathematical catchment simulation models. J. Hydrol 35:341–356.
McDonnell, J. 1990. A rationale for old water discharge through macropores in a steep humid catchment. Water Resour. Res 26:2821–2832.
Medlyn, B. E. Coauthors 2001. Stomatal conductance of forest species after long-term exposure to elevated CO2 concentration: A synthesis. New Phytologist 149:247–264.
Mitchell, S. W. and F. Csillag. 2000. Does pattern matter? Handling bias, uncertainty, and stability of predicted vegetation growth in Grasslands National Park, Saskatchewan. Proc. Fourth Int. Conf. on Integrating GIS and Environmental Modeling (GIS/EM4): Problems, Prospects and Research Needs, Banff, Canada.
Monteith, J. 1965. Evaporation and the environment. Proceedings of the 19th Symposium of the Society for Experimental Biology, Cambridge University Press, 205–233.
Nash, I. and I. Sutcliffe. 1970. Streamflow forecasting through conceptual models. J. Hydrol 10:282–290.
Norman, J. 1981. Simulation of microclimates. Biometeorology in Integrated Pest Management, J. Hatfield and I. J. Thomson, Eds., Academic Press, 65–99.
Oke, T. 1987. Boundary Layer Climates. 2d ed. University Press, 450 pp.
Parton, W., A. Mosier, D. Ojima, D. Valentine, D. Schimel, K. Weier, and A. Kulmala. 1996. Generalized model for N2 and N2O production from nitrification and denitrification. Global Biogeochem. Cycles 10:401–412.
Peters, D., J. Buttle, C. Taylor, and B. LaZerte. 1995. Runoff production in a forested, shallow soil Canadian shield basin. Water Resour. Res 31:1291–1304.
Phillip, J. 1957. The theory of infiltration: 4. Sorptivity and algebraic infiltration equation. Soil Sci 84:257–264.
Quinn, P., K. Beven, P. Chevallier, and O. Planchon. 1991. The prediction of hillslope flow paths for distributed hydrological modeling using digital terrain models. Hydrol. Proc 5:59–79.
Refsgaard, J. and B. Storm. 1995. MIKE SHE. Computer Models of Watershed Hydrology, V. Singh, Ed., Water Resources Publications, 809–846.
Running, S. and J. Coughlan. 1988. A general model of forest ecosystem processes for regional applications: Hydrologic balance, canopy gas exchange and primary production processes. Ecol. Modell 42:125–154.
Running, S. and E. Hunt. 1993. Generalization of a forest ecosystem process model for other biomes, BIOME-BGC and an application for global-scale models. Scaling Physiological Processes: Leaf to Globe, J. R. Ehleringer and C. Field, Eds., Academic Press, 141–158.
Running, S., R. Nemani, and R. Hungerfored. 1987. Extrapolation of synoptic meteorological data in mountainous terrain and its use for simulating forest evapotranspiration and photosynthesis. Can. J. For. Res 17:472–483.
Running, S., D. L. Peterson, L. E. Band, D. F. Potts, L. L. Pierce, and M. A. Spanner. 1989. Mapping regional forest evapotranspiration and photosynthesis by coupling satellite data with ecosystem simulation. Ecology 70:1090–1101.
Ryan, M. G. 1991. Effects of climate change on plant respiration. Ecol. Appl 1:157–167.
Tague, C. and L. Band. 2001. Simulating the impact of road construction and forest harvesting on hydrologic response using rhessys. Earth Surface Processes Landforms 26:135–151.
Tague, C., P. Groffman, and K. Belt. 2000. Spatially distributed modeling of the hydro-geologic and ecologic controls on nitrogen cycling and nitrate export for a small forested watersheds. Proc. American Geophysical Union 2000 Spring Meeting, Washington, D. C., Amer. Geophys. Union.
Thornton, P. 1998. Regional ecosystem simulation: Combining surface and satellite based observations to study linkages between terrestrial energy and mass budgets. Ph.D. thesis, School of Forestry, University of Montana, Missoula, 200 pp.
Thornton, P., S. Running, and M. White. 1997. Generating surfaces of daily meteorological variables over large regions of complex terrain. J. Hydrol 190:214–251.
vanGenuchten, M. and D. Nielsen. 1984. On describing and predicting the hydraulic properties of unsaturated soils. Ann. Geophys 3:615–628.
Vertessy, R., T. Hatton, P. O'Shaughnessy, and M. Jayasuriya. 1993. Predicting water yield from a mountain ash forest catchment using a terrain analysis based catchment model. J. Hydrol 150:665–700.
Walko, R. Coauthors 2000. Coupled atmosphere–biophysics–hydrology models for environmental modeling. J. Appl. Meteor 39:931–944.
Waring, R. and S. Running. 1998. Forest Ecosystems: Analysis at Multiple Scales. Academic Press, 370 pp.
Watson, F., R. Vertessy, and R. Grayson. 1999. Large scale modelling of forest hydrological processes and their long term effect on water yield. Hydrol. Processes 13:689–700.
Webster, K., I. Creed, and C. Tague. 2001. Modeling n export from the Noland Divide watershed: The importance of climate on nitrate-n production and transport. Gordon Research Conference on Forested Catchments: Hydrological, Geochemical, and Biological Processes, Andover, NH.
Wemple, B. C., J. Jones, and G. Grant. 1996. Channel network extension by logging roads in two basins, Western Cascades, Oregon. Water Resour. Bull 32:1195–1207.
White, M., P. Thornton, and S. Running. 1997. A continental phenology model for monitoring vegetation responses to interannual climatic variability. Global Biogeochem. Cycles 11:217–234.
Wigmosta, M., L. Vail, and D. Lettenmaier. 1994. Distributed hydrology–vegetation model for complex terrain. Water Resour. Res 30:1665–1679.
Williams, T. and L. Flanagan. 1997. Effect of changes in water content on photosynthesis, transpiration and discrimination agains 13CO2 and C18O16O in pleurozium and sphagnum. Oecologica 108:34–46.
Zheng, D., E. Hunt, and S. Running. 1993. A daily soil temperature model based on air temperature and precipitation for continental applications. Climate Res 2:183–191.
APPENDIX: RHESSys Variables
This appendix lists the current atmospheric variables that are used internally in the RHESSys model.
A1. Environmental time series inputs
A2. The RHESSys variable list
A2. (Continued)
A2. (Continued)
A2. (Continued)
A2. (Continued)
RHESSys model structure: Inputs, output, and preprocessing
Citation: Earth Interactions 8, 19; 10.1175/1087-3562(2004)8<1:RRHSSO>2.0.CO;2
Hillslope delineation and 30-m DEM for Pond Branch
Citation: Earth Interactions 8, 19; 10.1175/1087-3562(2004)8<1:RRHSSO>2.0.CO;2
RHESSys and measured stream nitrate concentrations (mgN L−1) for Pond Branch during Jan 1999–Dec 2000
Citation: Earth Interactions 8, 19; 10.1175/1087-3562(2004)8<1:RRHSSO>2.0.CO;2
RHESSys and measured streamflow (normalized by drainage area) for Pond Branch
Citation: Earth Interactions 8, 19; 10.1175/1087-3562(2004)8<1:RRHSSO>2.0.CO;2
RHESSys saturation deficit (mm of water) for Pond Branch in Aug 2000
Citation: Earth Interactions 8, 19; 10.1175/1087-3562(2004)8<1:RRHSSO>2.0.CO;2
