The science models#

This page provides an overview of the implementations of each science model. These descriptions are intentionally brief to allow you to see all of the different science model components in a single location. Each section provides links to additional detail on the implementation or to the documentation of the actual Python classes and function that provide the implementation.

Abiotic models#

Abiotic models provide the three-dimensional microclimate for the Virtual Ecosystem. Using a small set of input variables from external sources such as reanalysis or regional climate models, the model calculates atmospheric and soil parameters that drive the dynamics of plants, animals, and microbes at different vertical levels:

  • above canopy (canopy height + reference measurement height, typically 2 m)

  • canopy (dynamic heights provided by plant model)

  • surface (10 cm above ground, including understorey vegetation)

  • topsoil (25 cm below ground)

  • subsoil (minimum of one layer at 1 m depth)

The abiotic_simple model is a simple regression model that estimates microclimatic variables based on empirical data for a monthly model timestep. It can be used for quick simulations and it is used to initialise the process-based abiotic model.

The process-based abiotic model is the default model in ve_run. At the moment, the model solves energy balance equantions assuming steady state; in the future we plan to run this model with a (sub-)daily resolution to capture diurnal cycles and weather extremes.

Both versions of the abiotic model provide the following key variables at relevant vertical levels:

  • Air temperature (°C), relative humidity (-), and vapour pressure deficit (VPD, kPa)

  • Soil temperature (°C)

  • Atmospheric \(\ce{CO_{2}}\) concentration (ppm)

  • Atmospheric Pressure (kPa)

  • Wind speed (m s-1)

Simple Abiotic Model#

The abiotic_simple model is a one-column model that operates on a grid cell basis and does not consider horizontal exchange of energy, atmospheric water, and momentum. The model uses linear regressions from (Hardwick et al., 2015) and (Jucker et al., 2018) to predict atmospheric temperature, relative humidity, wind speed, and vapour pressure deficit at ground level (1.5 m) given the above canopy conditions and leaf area index of intervening canopy. A vertical profile across all atmospheric layers is then interpolated using a logarithmic curve between the above canopy observation and ground level prediction. Soil temperature is interpolated between the surface layer and the soil temperature at 1 m depth which roughly equals the mean annual temperature. The model also provides a constant vertical profile of atmospheric pressure and atmospheric \(\ce{CO_{2}}\) based on external inputs.

Process-based Abiotic Model#

The process-based abiotic model contains a mechanistic representation of the radiation balance, the energy balance, and wind profiles. Submodules are closely coupled to the hydrology and plants models through the exchange of energy and water. The model also provides a constant vertical profile of atmospheric pressure and atmospheric \(\ce{CO_{2}}\) based on external inputs. Most processes are calculated on a per grid cell basis; horizontal exchange of properties will be considered at a later stage.

Note

Some of the features described here are not yet implemented.

Radiation and Energy balance#

The microclimate submodule contains the equations to solve the radiation and energy balance in the Virtual Ecosystem. This includes the radiation balance with reflection and scattering of shortwave radiation from canopy and surface, a vertical profile of net shortwave radiation, and outgoing longwave radiation from canopy and surface. The net radiation is then partitioned in sensible and latent heat fluxes from canopy layers and surface to the atmosphere. Part of the net radiation will be converted into soil heat flux. Based on these turbulent fluxes, air temperature, canopy temperature, relative humidity, and soil temperature will be updated at each level. The vertical mixing between layers is assumed to be driven by heat conductance because turbulence is typically low below the canopy (Maclean and Klinges, 2021).

Wind#

The microclimate submodule will also calculate the above- and within-canopy wind profiles for the Virtual Ecosystem. These profiles determine the exchange of heat and water between soil and atmosphere below the canopy as well as the exchange with the atmosphere above the canopy.

Hydrology Model#

The hydrology model simulates the hydrological processes in the Virtual Ecosystem. We placed hydrology in a separate model to allow easy replacement with a different hydrology model. Also, this separation provides more flexibility in defining the order of models and/or processes in the overall Virtual Ecosystem workflow.

Note

Some of the features described here are not yet implemented.

Vertical hydrology components#

The vertical component of the hydrology model determines the water balance within each grid cell. This includes above ground processes such as rainfall, intercept, and surface runoff out of the grid cell. The below ground component considers infiltration, bypass flow, percolation (= vertical flow), soil moisture and matric potential, horizontal sub-surface flow out of the grid cell, and changes in groundwater storage. The model is loosely based on the LISFLOOD model (Van Der Knijff et al., 2010).

Horizontal hydrology components#

The second part of the hydrology model calculates the horizontal water movement across the full model grid including accumulated surface runoff and sub-surface flow, and river discharge rate, see above ground details. The flow direction is based on a digital elevation model.

Plant Model#

The Plant Model models the primary production from plants in the Virtual Ecosystem. We use the P Model (Prentice et al., 2014, Wang et al., 2017), to estimate the optimal balance between water loss and photosynthetic productivity and hence gross primary productivity (GPP). The P Model requires estimates of the following drivers:

  • Air temperature (°C)

  • Vapour pressure deficit (VPD, Pa)

  • Atmospheric pressure (Pa)

  • Atmospheric \(\ce{CO_{2}}\) concentration (parts per million)

  • Fraction of absorbed photosynthetically active radiation (\(F_{APAR}\), unitless)

  • Downward shortwave radiation (DSR, \(\text{W}, m^{-2}\))

GPP is then allocated to plant maintenance, respiration and growth using the T Model (Li et al., 2014).

This growth model is used to simulate the demographics of cohorts of key plant functional types (PFTs) under physiologically structured population models developed in the Plant-FATE framework. The framework uses the perfect-plasticity approximation (PPA, Purves et al. (2008)) to model the canopy structure of the plant community, the light environments of different PFTs and hence the change in the size-structured demography of each PFT through time.

Soil Model#

The principal function of the Soil Model is to model the cycling of nutrients. This cycling is assumed to be primarily driven by microbial activity, which in turn is heavily impacted by both environmental and soil conditions. Plant-microbe interactions are taken to principally be either exchanges of or competition for nutrients, and so are modelled within the same nutrient cycling paradigm. Three specific nutrient cycles are incorporated into this model:

Carbon cycle#

The Carbon cycle uses as its basic structure a recently described soil-pool model termed the Millennial model (Abramoff et al., 2018). This model splits carbon into five separate pools: particulate organic matter, low molecular weight carbon (LMWC), mineral associated organic matter, aggregates and microbial biomass. Though plant root exudates feed directly into the LMWC pool, most biomass input is direct less direct and occurs via litter decomposition. Thus, we utilize a common set of litter pools (Kirschbaum and Paul, 2002), that are divided between above- and below-ground pools, and by biomass source (e.g. deadwood).

Nitrogen cycle#

The Nitrogen cycle is strongly coupled to the carbon cycle, therefore tracking the stoichiometry of the carbon pools is key to modelling it correctly. In addition, specific forms of nitrogen are explicitly modelled. They are as follows: a combined \(\ce{NH_{3}}\) and \(\ce{NH_{4}^{+}}\) pool to represent the products of nitrogen fixation and ammonification, a \(\ce{NO_{3}^{-}}\) pool to represent the products of nitrification, and a \(\ce{NO_{2}^{-}}\) pool to capture the process of denitrification.

Phosphorous cycle#

The Phosphorus cycle is similarly coupled to the carbon cycle. The additional inorganic pools tracked in this case are as follows: primary phosphorus in the form of weatherable minerals, mineral phosphorus which can be utilized by plants and microbes, secondary phosphorus which is mineral associated but can be recovered as mineral phosphorus, and occluded phosphorus which is irrecoverably bound within a mineral structure.

Further details#

A separate page documents the further theoretical background for the Soil Model.

Animal Model#

The Animal Model simulates the animal consumers for the Virtual Ecosystem. We follow the Madingley Model (Harfoot et al., 2014) to provide the foundational structure as well as some of the dynamics. The key processes of the model are:

  • foraging and trophic dynamics

  • migration

  • birth

  • metamorphosis

  • metabolism

  • natural mortality

Functional Groups#

Animals within the Animal Model are sorted into functional groups, not biological species. Functional groups share functional traits and body-mass ranges and so behave similarly within the ecosystem. Defining a functional group within the Animal Model requires the following traits:

  • name

  • taxa: mammal, bird, insect

  • diet: herbivore, carnivore

  • metabolic type: endothermic, ectothermic

  • reproductive type: semelparous, iteroparous, nonreproductive

  • development type: direct, indirect

  • development status: adult, larval

  • offspring functional group

  • excretion type: ureotelic, uricotelic

  • birth mass (kg)

  • adult mass (kg)

A set of these functional groups are used to define an instance of the Animal Model.

Animal Cohorts#

Animals are represented as age-specific cohorts, containing many individuals of the same functional type. The key Animal Model processes are run at the cohort level. We track the internal state of the average individual of that cohort over time to determine the resulting dynamics, such that events like starvation and metamorphosis occur based on that cohort’s internal state. Predator-prey interactions, likewise, occur between animal cohorts as part of foraging system.

Disturbance Model#

Warning

This model is not yet in development.

Introducing disturbances (e.g. logging) into the model will usually require making alterations to the state of multiple models. As such, different disturbance models are collected in a separate Disturbance Model. This model will be capable of altering the state of all the other models, and will do so in a manner that allows the source of the changes to be explicitly identified.