Scaling Vegetation Dynamics

 

 

SVD

The SVD framework is a deep-learning based modeling framework for upscaling vegetation transitions. iLand is particularly well suited to provide training data for the deep neural networks and ecosystem attribute data. The model already includes the means to classify vegetaiton into distinct "states" that are used in SVD and direct outputs to generate attribute information.
The spatial grain for SVD related data is the resource unit level, i.e. 1ha.

Definition of states

The section model.settings.svdStates in the project file is dedicated to the definition of states in iLand.

Key	Type	Description
	enabled	boolean	if true iLand tracks the vegetation state for each resource unit over time (required for SVD outputs).
	structure	string	type of classification scheme for structure (i.e. canopy height). See below for details.
	functioning	string	type of classification scheme for ecosystem functioning (i.e. canopy height). See below for details.

The classification of structure

The structure component is related to the canopy height on a resource unit; we here use the "top height" which we define as the 90th percentile of the the maximum tree height on the 10m "height grid".
A canopy is irregular when the height on 50% of the area (10m pixels) is below the top height

The following types are available (key structure):

• 4m: classes with a width of 4m (0- <4m, 4- <8m, 8 - <12m, ...) and a maximum height of 80m. For irregular canopies the class width is 12m (<12, 12- <24, ... 48- <60m, >=60m). The total number of classes is 27 (21 standard, 6 irregular). This is the default value.
• 2m: classes with a width of 2m and a maximum top height of 60m (0 - <2m, 2 - <4m, .... 58 - <60m, >60m ). For irregular canopies the class width 8m (<8m, 8 - <16m, 48- <56m, >=56m). The total number of classes is 39 (31 standard, 8 irregular)

The classification of functioning

Ecosystem functioning is based on the leaf area index of a resource unit. The leaf area index includes both adult trees (height >4m) and saplings (<4m).
The following schemes are available (key functioning):

• 3: three classes: LAI 0 - <2, 2 - <4, >=4 (default)
• 5: five classes LAI 0- <1, 1- <2, 2- <3, 3- <4, >=4

SVD related outputs

There are several outputs related to SVD, e.g. SVDUniqueState SVDIndicator.
See Outputs for the details!