local scale landscape scale forest stand/ site level (management unit) Multi-scale upscaling approaches of soil properties from soil monitoring data sampling plot level
Motivation: The Need for Regionalization Available observations are often not sufficient as a basis for decision making in land use management. A tool is needed that describes the area related description of indicators for land use functions.
Motivation: The Need for Regionalization The value of properties (e.g., C stocks) at unsampled sites has to be estimated with the help of an existing sampling?
Scale problems I. measurements and process/parameter variability normally occur at a small scale sampling plot level
Scale problems II. the most important problems have to be solved at more large scales, e.g. ecosystems and landscapes sampling plot level
local scale landscape scale III. forest stand/ site level (management unit) Scale problems but a direct upscaling (regionalization) may fail, if the patterns and processes of the small scales differ from the larger scales sampling plot level
Definitions of terms Regionalization Area related transfer of measurements which are only available for georeferenced points (Jansen et. al.) Up-scaling Spatial and temporal scales in environmental research cover wide range. Regionalization involves taking spatial, temporal and process information at one scale and using it to derive information at another scale
Regionalization methods The point to area problem has no clear answer. There are different methods for the solution of this problem, no one technique is the best Spatial modelling invariably involves the comparison of competing models Conceivable Classifications gradual / abrupt (continuous / discrete) exact / approximate (inexact) deterministic / stochastic global / local (stratified) analysis of uncertainty (calculation of model precision) possible / not possible
Which model should we use? An "optimum" regionalization technique does not exist for any type of environmental data The model we use depends on the questions we wish to ask and the nature of the data.
local scale landscape scale forest stand/ site level (management unit) Frequent Problems In many data sets there are global trends, which vary slowly, overlain by local fluctuations, which vary rapidly and produce uncertainty (error) in the recorded values Antagonistic and/or increasing processes on different temporal and spatial scales suggest complex patterns sampling plot level
Frequent Problems Sampling intensity is low in relation to the pattern of heterogeneity Strong local fluctuations, which vary rapidly and produce great uncertainty The available data set is too small for a reliable estimation of kriging functions Large nugget variance in the sample variogram Mixture of continuous (e.g. topography) and discrete influences (e.g. geology)
Our approach to these problems: 1. Two-stage up-scaling scaling-procedure with global and stratified approaches 2. Complementary method-mixture mixture of statistical models
Complementary method-mixture of statistical models Multiple linear regression analysis Stepwise regressions (logistic regression for binary response variables) Geostatistics Test for spatial autocorrelation of model residuals ICP Forest data ρ(0) ρ( h ) Auxiliary Variables 1 0.8 0.6 0.4 0.2 Measurements Residuals 444 1194 1034 1324 1124 1074 1350 1034 1282 1124 1350 1 topographic variables 498 1126 1324 1144 1194 1282 (DEM, slope, aspect, slope position..) 1066 1144 1066 1074 1126 0.8 land use characteristics 444 498 0.6 classification of the parent material / substrate 0.4 soil information (e.g., hydromorphic sites: yes/no) 0.2... 0 0 8 16 24 32 40 48 56 64 72 8 16 24 32 40 48 56 64 72 h h
Potential Auxiliary Variables (Regressors) Hydrology hydrological regime duration of water saturation water budget... Soil, substrate hydromorphic attrib. soil type soil acidity texture... Vegetation tree species stand age physiognomy life form vegetation height... Response Variable Geomorphology landscape position relief form morphometric relief attributes complex relief parameters watershed size... Hydrochemistry ph salt content nutrient content... Ecology ecotype habitat type nutrient content... Others geology climate genesis...
Relief Analysis provides information about Hengl, Gruber and Shrestha, 2003 energy budget: solar radiation / insolation water budget: evaporation, soil humidity, watershed material budget: erosive power of the terrain
Relief Analysis: Examples TWI Topographic wetness index is a ratio between the specific catchment area and slope. It primarily reflects the accumulation processes. Solar radiation Landform units (canyons, valleys, slopes, ridges, mountain tops,...) on the basis of Topographic Position Index (TPI)
Global modelling or stratified modelling? Stratified / local methods apply an algorithm to a small portion of the total set of points Global methods determine a single function which is mapped across the whole region a change in one input value affects the entire map smoother surfaces with less abrupt changes Interpolated value Geographical Institute University Zürich, GEO 315
C-Models for forest soils in Baden-Württemberg Global Transfer Model, DEM 25m C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg Global Transfer Model, DEM 25m C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg Global Transfer Model, DEM 25m
C-Models for forest soils in Baden-Württemberg Global Transfer Model, DEM 25m C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg Global Transfer Model, DEM 25m, Pre-Alpine Lowlands C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg Stratified Model, DEM 10m, Pre-Alpine Lowlands
C-Models for forest soils in Baden-Württemberg Stratified Model Pre-Alpine Lowlands (DEM 10m) C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg Stratified Model Pre-Alpine Lowlands (DEM 10m) C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg global vs. stratified upscaling Global Model (all data), DEM 25m Stratified Model (Only Pre-Alpine Lowlands), DEM 25m C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg stratified approaches with different DEM's Stratified Model DEM 10m Stratified Model DEM 25m C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg scenarios: C-stocks, mineral soil Coniferous Forests (0 % Deciduous Trees) Mixed Forests (50 % Deciduous Trees) C-stock at 0-60 cm soil depth [t/ha]
C-Models for forest soils in Baden-Württemberg scenarios: C-stocks, humus layer Coniferous Forests (0 % Deciduous Trees) Mixed Forests (50 % Deciduous Trees) C-stock humus layer [t/ha]
Conclusions Upscaling tools on basis of regression analyses may give a relatively fast assessment of environmental features Stratified transfer models = meaningful tool to describe important influences more in detail (= at lower scale) and to improve the model precision The close relation to small scaled landscape attributes is important for the practice relevance of up-scaling results Option to identify landscape compartments with a high potential for, for example, soil carbon storage; measurements can be focussed on these areas Prerequisite data base! (measurements, basic maps) The methodology described cause-effect relationships, scenario-modelling, decision support tool May be applied (not only) to soil physical and chemical properties, and to metrical and binary response variables
Application to soil physical properties from soil profile descriptions Examples: Rooting depth Amount of soil skeleton (>2 mm fraction)
Application to Forest Site Mapping Forest site map Logistic regression model
Application to monitoring data of the forest health condition Spatial distribution of the mean defoliation taking 60 years for age (scenario)