TO BE DONE WHEN A STABLE VERSION O USER INTERFACE IS ACCEPTED
This service offers model-based estimates of the historic (1986-2016), recent trends (up to 1 year) and projected scenarios (up to year 2100) in soil moisture levels of forests and drought stress of main tree species over Catalonia (NE Spain). Estimates are obtained using a soil plant water balance (SPWB) model and forest plot data of the Spanish National Forest Inventory (De Cáceres et al. 2015). Weather data comes primarily from weather station networks and is processed using the routines described in De Cáceres et al. (2018). Water balance is done at the stand level only, neglecting lateral hydrological processes. The service is intended to complement DEBOSCAT, a monitoring program of observed drought-related forest decline events (mortality, decoloration and leaf loss) conducted in the same area (Chaparro et al. 2016).
The water balance model described in De Cáceres et al. (2015) follows the design principles from BILJOU (Granier et al. 1999) and SIERRA water balance submodel (Mouillot et al. 2001). The model performs daily updates of soil water content as a function of the stand structure and daily weather (radiation, temperature and precipitation). Soil water balance is the difference between processes determining water input (i.e. precipitation) and outputs (canopy interception, tree transpiration, bare soil evaporation, surface runoff and deep drainage). Details of the formulation of each of these processes are given in De Cáceres et al. (2015). The model is implemented in the medfate R package.
The SPWB model requires values of the following daily weather variables: precipitation, potential evapotranspiration (PET). Average daily temperature and solar radiation are also required, but these variables are not currently stored.
|PET||Penman’s potential evapotranspiration||mm/day||Sum|
Details of weather data sources and processing are given below.
Forest water balance
Every simulation day, the SPWB model updates the leaf area index (LAI) of forests according to a simple phenology model and produces the following variables describing the water balance in forest stands:
|Runoff||Surface (overland) runoff||mm/day||Sum|
|DeepDrainage||Deep drainage to groundwater||mm/day||Sum|
|LAI||Leaf area index (including all woody species)||m2/m2||Average|
|Theta||Average soil moisture relative to field capacity.||[0-1]||Average|
Net precipitation is gross precipitation minus the water intercepted by the canopy and evaporated from there. After going through the canopy, water infiltrates into the soil. Runoff is the amount of water that does not infiltrate and leaves the forest stand as an overland flow. Even if water infiltrates into the soil, part of it flows through macroscopic pores and is exported to layers not accessible to plants through deep drainage. Plant transpiration is the amount of water that flows through stems and is evaporated from leaf surfaces. Soil evaporation is the amount of water that is evaporated from the soil surface.
Relative conductance and daily drought stress
Plant transpiration depends on atmospheric demand, soil water content and species identities. Some species close suffer from drought stress at higher soil water contents than others. Whole-plant relative conductance is a relative measure of how much plant transpiration is reduced due to stress. The daily drought stress (DDS) index is defined as the complement ofthe whole-plant relative conductance. In other words, daily stress is higher when the species has a lower overall plant conductance (which can be due to a closure of stomata or the presence of emboli in vessels). For a given soil moisture level in a stand, plants present may experience different levels of drought stress depending on their species identity, root system,…. For each simulated forest stand, the daily drought stress for every species present is calculated by averaging daily drought stress of plant cohorts of the species, using LAI values as weights. The following species are tracked in terms of their daily stress.
|PinusHalepensis||Drought stress for Pinus halepensis||[0,1]||Average|
|PinusNigra||Drought stress for Pinus nigra||[0,1]||Average|
|PinusSylvestris||Drought stress for Pinus sylvestris||[0,1]||Average|
|PinusUncinata||Drought stress for Pinus uncinata||[0,1]||Average|
|PinusPinea||Drought stress for Pinus pinea||[0,1]||Average|
|PinusPinaster||Drought stress for Pinus pinaster||[0,1]||Average|
|QuercusIlex||Drought stress for Quercus ilex||[0,1]||Average|
|QuercusSuber||Drought stress for Quercus suber||[0,1]||Average|
|QuercusHumilis||Drought stress for Quercus humilis||[0,1]||Average|
|QuercusFaginea||Drought stress for Quercus faginea||[0,1]||Average|
|FagusSylvatica||Drought stress for Fagus sylvatica||[0,1]||Average|
|Overall||Average drought stress across all woody plants||[0,1]||Average|
Cumulative drought stress
The cumulative effect of drought stress on forests is tracked by counting the number of consecutive days with daily drought stress values above 0.5 (i.e. less than 50% in conductance). Assuming that in the first of january stress will always be below 50%, within a year values of cumulative drought stress are therefore bounded between 0 and 365. Cumulative drought stress is represented for the same species (and overall) as in daily drought stress.
|PinusHalepensis||Drought stress for Pinus halepensis||[0,365]||Maximum|
|PinusNigra||Drought stress for Pinus nigra||[0,365]||Maximum|
|PinusSylvestris||Drought stress for Pinus sylvestris||[0,365]||Maximum|
|PinusUncinata||Drought stress for Pinus uncinata||[0,365]||Maximum|
|PinusPinea||Drought stress for Pinus pinea||[0,365]||Maximum|
|PinusPinaster||Drought stress for Pinus pinaster||[0,365]||Maximum|
|QuercusIlex||Drought stress for Quercus ilex||[0,365]||Maximum|
|QuercusSuber||Drought stress for Quercus suber||[0,365]||Maximum|
|QuercusHumilis||Drought stress for Quercus humilis||[0,365]||Maximum|
|QuercusFaginea||Drought stress for Quercus faginea||[0,365]||Maximum|
|FagusSylvatica||Drought stress for Fagus sylvatica||[0,365]||Maximum|
|Overall||Average drought stress across all woody plants||[0,365]||Maximum|
In this mode, the water balance model is run once every day for each simulation location after gathering weather station data and interpolating weather over the target simulation location. Soil moisture and other state variables are stored from one day to the other, so that the accumulation of drought stress (or the relieving effect of precipitations) can be simulated.
The following diagram illustrates the workflow of tasks necessary for the initialization of the tool and every day actions:
SPWB simulations are conducted on two sets of locations:
Simulation on pixels are only used for mapping purposes, whereas simulations on IFN3 plots are used to examine temporal trends.
Initialization for IFN3 plots
Initialization of pixels
Forest categories are first defined for 200x200m pixels and IFN3 plots according to the MFE50, using the first two dominant species and their occupation level. Each 200x200 m pixel is then assigned an IFN3 plot according to the following rules:
The imputation process of IFN3 plots to 200x200 m pixels may entail gross errors at this resolution, but it is intended to provide sound estimates when averaging the results of 200x200 m pixels at 1km resolution. The imputation of a IFN3 plot provides the species composition for the pixel, a preliminary estimate of forest structure, and soil/root distribution parameters. Tree heights and leaf area index of each tree cohort at each target pixel are then modified according to 20x20 m resolution layers of average tree height and foliar biomass that had been estimated from LiDAR data (0.5 pulses/m2) by CREAF [ref]. Specifically, these two layers are first upscaled to 200x200 m resolution and for each target pixel, the ratio between average height at the pixel corresponding to the location of the assigned IFN3 plot and average height at the target pixel is used to correct tree heights from plot data for the target pixel. Similarly, the ratio in foliar biomass is used to correct the leaf area index of trees for the target pixel.
Weather station data of the current day is downloaded from the Spanish Agencia Estatal de Meteorologia (AEMET) using their OpenData API. This information, complemented with weather data from previous days, is used to interpolate weather variables over the simulation locations (pixels or IFN3 plots). Precipitation, temperature and relative humidity are interpolated for the target locations and solar radiation is estimated from latitude, topography and the other variables (see details in De Cáceres et al. 2018). Interpolation routines are implemented in the R package meteoland.
The spatial scope of displayed results is within Catalonia boundaries and at a maximum 2km distance from IFN3 plots. Two raster resolution levels and three display options are offered:
Values at 1km2 are calculated averaging 200 x 200 m pixels. Smoothing is done on 1 km2 pixels using a spatial kernel of radius 2 km (R package spatstat). Maps are usually displayed at the daily scale, but the user can decide to map values aggregated temporally up to a number of days Temporal aggregation follows the function (sum or average) indicated in variable tables above.
Daily temporal trends are stored for IFN3 plots only (i.e. not for pixels). Average trends can be displayed at four spatial levels: (a) Watersheds; (b) Counties; © Municipalities; and (d) IFN3 plots. The user can select the desired spatial structure depending on each level. In the case of IFN3 plots the trends of the selected IFN3 plot are displayed. For the other levels, the application loads the trends of all the IFN3 plots within county/municipality boundaries and calculates average and 5%-95% confidence interval.
In this mode, the soil water balance model is used to simulate water balance for the historical period between 1986 and 2016. Unlike current-day simulations, climate change simulations are conducted for forest inventory plots only. Moreover, only monthly and yearly summaries are stored.
Simulation locations and initialization
The second and third forest inventories in Spain (IFN2 and IFN3) are comparable in sampling methods. In Catalonia, IFN2 sampling was conducted between 1987 and 1991, whereas IFN3 sampling was conducted between 2000 and 2001. In order to simulate historic soil water balance and drought stress, simulations are run on both IFN2 and IFN3 plot data. Since IFN3 was more complete than IFN2, simulations using IFN2 plot data are restricted to those plots that were resampled in IFN3 (10,469 plots), even if in some cases the exact sampling location may have changed. For those 'common' plots, the simulation results are averaged assigning weights to IFN2 and IFN3 according to the year of simulation, as described in De Cáceres et al. (2015). For those IFN3 plots that do not have IFN2 counterpart, only the simulations with IFN3 data is used. Initialization procedure is the same as for current-day simulations.
The following diagram illustrates the workflow of tasks necessary for the initialization and simulations in the historic mode:
Historic weather series for each forest plot is obtained by interpolating daily weather station data for the 1986-2016 period, using package meteoland. Surface weather station data was obtained from the Agencia Estatal de Meteorologia (AEMET) and Servei Meteorologic de Catalunya (SMC).
In this mode, the soil water balance model is used to project forest drought stress under climate change predictions. Unlike current-day simulations, climate change simulations are conducted for IFN3 plots only. Moreover, only monthly and yearly summaries are stored. Initialization procedure is the same as for current-day simulations.
Climate scenarios were downloaded from ESFG (Earth System Grid Federation, a web site that centralizes climate data from GCM and RCM uploaded in the frame of different international consortia, including the CORDEX regionalization project. Fifth assessment report (AR5) is used and RCP 4.5 and RCP 8.5 scenarios are used. Daily climate projections for the 2006-2100 period are used as forcing. They were produced by two regional climate models (CCLM4-7-13 and RCA4), both driven by the same global climate model (CNRM-CERFACS).
Statistical downscaling is applied to daily climate data, which includes a bias correction. Bias correction is conducted using package meteoland, and requires a local historic meteorological series to use as reference for bias estimation. This is obtained by interpolating daily weather station data over IFN3 plots for the 1986-2005 period (station data from AEMET and SMC).
The following table provides an idea of current computational times of the different tasks:
|MODE 1 (current)||Initialization||10 min||No||No|
|MODE 1 (current)||Plot meteo interpolation||1 min||No||Yes|
|MODE 1 (current)||Plot soil water balance||2 min||Yes||Yes|
|MODE 1 (current)||Plot update trends||2 min||No||Yes|
|MODE 1 (current)||Pixel meteo interpolation||4 min||Yes||Yes|
|MODE 1 (current)||Pixel soil water balance||3 min||Yes||Yes|
|MODE 1 (current)||Smoothing for maps||1 min||No||Yes|
|MODE 2/3||Meteo interpolation 1986-2016||7 days||No||No|
|MODE 2 (historic)||Soil water balance IFN2/IFN3||18 hours||No||No|
|MODE 2 (historic)||Mapping||2 hours||No||No|
|MODE 2 (historic)||Trend mapping||1 hour||No||No|
|MODE 3 (projection)||Bias correction & SPWB||10 days||No||No|
|MODE 3 (projection)||Trend mapping||10 hours||No||No|
Daily routines in total account for 13 min of computation in a server with 8 cores.
This web-based service has been develop by the Vegetation Modeling Group at Forest Sciences Centre of Catalonia (CTFC), in collaboration with the Center for Ecological Research and Forestry Applications (CREAF):
Projects: FORESTCAST (CGL2014-59742-C2-2-R, Ministerio de Economía y Competitividad), DRESS (CGL2017-89149-C2-2-R, Ministerio de Economía y Competitividad).
Fellowships: RYC-2012-11109 to M. De Cáceres (Ministerio de Economía y Ciencia).
The data offered in this website is the result of modelling exercises and the accuracy of results may be impaired by several factors, like inaccuracies of input data, inappropriate model design or parameterization. Users should be aware of all these limitations when using the data offered here. We decline all responsibility for how the data in this website is used by third parties.