Quadtree and Sub-Grid Sampling FAQ - Revision history

Abrar.Alttahir: /* My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why? */

2025-04-30T07:20:30Z

My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why?

← Older revision		Revision as of 17:20, 30 April 2025
Line 32:		Line 32:

	=My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why?=		=My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why?=
	The SGS Sample Target Distance sets the spacing at which elevations are sampled within each 2D cell, not the ~~model~~ ~~output~~ resolution. SGS works by sampling the terrain at a ~~smaller~~ scale ~~inside~~ each computational cell to improve how topography is represented~~, especially across cell faces~~. The hydraulic behaviour across the cell is determined using these smaller elevation samples, without changing the size of the 2D cells themselves.~~<br>~~		The SGS Sample Target Distance sets the spacing at which elevations are sampled within each 2D cell. It does not define the 2D cell resolution. SGS works by sampling the terrain at a finer scale within each computational cell to improve how topography is represented. The hydraulic behaviour across the cell is determined using these smaller elevation samples, without changing the actual size of the 2D cells themselves.

	For example, ~~using~~ an 8 m cell size and a 2 m SGS Sample Target Distance, the DEM is sampled at 5 points across each face. This results in 25 elevation points being used to define the ~~volume~~ versus elevation relationship ~~within~~ ~~the~~ 2D ~~cell~~, and ~~the~~ ~~same~~ points ~~are~~ used ~~along each face~~ to define the ~~conveyance~~ versus elevation relationship. ~~This~~ ~~helps~~ ~~better~~ ~~capture~~ more ~~details~~ from input topography while maintaining the ~~chosen input~~ grid size.~~<br>~~		For example, with an 8 m cell size and a 2 m SGS Sample Target Distance, the DEM is sampled at 5 points across each cell face. This results in 5 elevation points being used to define the flow area versus elevation relationship at each 2D face, and 25 elevation points used to define the volume versus elevation relationship within the 2D cell. This allows more detail from the input topography to be captured while maintaining the original grid size.
	If a smaller output resolution (such as for velocity or velocity related outputs like hazards) is required, this is controlled by the 2D cell size and/or the nesting level of the Quadtree polygon, not by the SGS sampling settings. SGS improves the internal hydraulic calculations within each cell but does not alter the 2D grid or the output resolution.

	<br>
			However, TUFLOW still stores only one velocity value per face and one depth value per cell. This means the resolution of mesh based outputs (e.g. XMDF) and standard grid outputs (e.g. ASC, FLT, TIF) is not reduced to match the SGS Sample Target Distance.

			When SGS is enabled, users can output high resolution water level and depth outputs based on the cell centre water level and the sampled sub-grid elevations (see: [[TUFLOW HR Output]]). However, if higher resolution velocity or velocity derived outputs (such as hazard) are required, this must be achieved by reducing the 2D cell size, not by adjusting the SGS sampling settings.

	=Does using SGS increase model runtimes?=		=Does using SGS increase model runtimes?=

Pavlina Monhartova at 03:00, 29 April 2025

2025-04-29T03:00:50Z

Show changes

Pavlina Monhartova: /* My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why? */

2025-04-29T02:55:41Z

My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why?

← Older revision		Revision as of 12:55, 29 April 2025
Line 92:		Line 92:
	# Read the 2d_zsh merge polygon into the model using the "Create TIN Zpts Write TIN" command in the tgc file.		# Read the 2d_zsh merge polygon into the model using the "Create TIN Zpts Write TIN" command in the tgc file.
	=My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why?=		=My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why?=
	The SGS Sample Target Distance sets the spacing at which elevations are sampled within each 2D cell, not the model output resolution.		The SGS Sample Target Distance sets the spacing at which elevations are sampled within each 2D cell, not the model output resolution. SGS works by sampling the terrain at a smaller scale inside each computational cell to improve how topography is represented, especially across cell faces. The hydraulic behaviour across the cell is determined using these smaller elevation samples, without changing the size of the 2D cells themselves.<br>
		⚫	For example, using an 8 m cell size and a 2 m SGS Sample Target Distance, the DEM is sampled at 5 points across each face. This results in 25 elevation points being used to define the volume versus elevation relationship within the 2D cell, and the same points are used along each face to define the conveyance versus elevation relationship. This helps better capture more details from input topography while maintaining the chosen input grid size.<br>

		⚫	If a smaller output resolution (such as for velocity or velocity related outputs like hazards) is required, this is controlled by the 2D cell size and/or the nesting level of the Quadtree polygon, not by the SGS sampling settings. SGS improves the internal hydraulic calculations within each cell but does not alter the 2D grid or the output resolution.
	SGS (Sub Grid Sampling) works by sampling the terrain at a smaller scale inside each computational cell to improve how topography is represented, especially across cell faces. The hydraulic behaviour across the cell is determined using these smaller elevation samples, without changing the size of the 2D cells themselves.

⚫	For example, using an 8 m cell size and a 2 m SGS Sample Target Distance, the DEM is sampled at 5 points across each face. This results in 25 elevation points being used to define the volume versus elevation relationship within the 2D cell, and the same points are used along each face to define the conveyance versus elevation relationship. This helps better capture more details from input topography while maintaining the chosen input grid size.

⚫	If a smaller output resolution (such as for velocity or ~~depth~~ outputs) is required, this is controlled by the 2D cell size and the nesting level of the Quadtree polygon, not by the SGS sampling settings. SGS improves the internal hydraulic calculations within each cell but does not alter the 2D grid or the output resolution.

	Note SGS provides the most benefit when the input DEM resolution is significantly smaller than the 2D cell size. If the DEM is similar to or coarser than the 2D cell size, SGS offers little improvement.
	<br>		<br>
	{{Tips Navigation		{{Tips Navigation

Abrar.Alttahir at 02:38, 29 April 2025

2025-04-29T02:38:00Z

← Older revision		Revision as of 12:38, 29 April 2025
Line 90:		Line 90:
	# Digitise a 2d_zsh merge polygon around the 2d_zsh lines and points that define the open channel. Tutorial Module 2 provides instructions on how to create this 2d_zsh merge polygon: [[Tutorial_M02#Part_2_-_Other_Topographic_Updates \|Tutorial M02 Part2 - Other Topographic Updates]] <br>		# Digitise a 2d_zsh merge polygon around the 2d_zsh lines and points that define the open channel. Tutorial Module 2 provides instructions on how to create this 2d_zsh merge polygon: [[Tutorial_M02#Part_2_-_Other_Topographic_Updates \|Tutorial M02 Part2 - Other Topographic Updates]] <br>
	# Read the 2d_zsh line and points cross section inputs into the model using the "Read GIS Z Shape" command in the tgc file. <br>		# Read the 2d_zsh line and points cross section inputs into the model using the "Read GIS Z Shape" command in the tgc file. <br>
	# Read the 2d_zsh merge polygon into the model using the "Create TIN Zpts Write TIN" command in the tgc file. ~~<br>~~		# Read the 2d_zsh merge polygon into the model using the "Create TIN Zpts Write TIN" command in the tgc file.
			=My SGS Sample Target Distance is set to 1 metre, but the model output resolution is larger than 1 metre. Why?=
			The SGS Sample Target Distance sets the spacing at which elevations are sampled within each 2D cell, not the model output resolution.

			SGS (Sub Grid Sampling) works by sampling the terrain at a smaller scale inside each computational cell to improve how topography is represented, especially across cell faces. The hydraulic behaviour across the cell is determined using these smaller elevation samples, without changing the size of the 2D cells themselves.

			For example, using an 8 m cell size and a 2 m SGS Sample Target Distance, the DEM is sampled at 5 points across each face. This results in 25 elevation points being used to define the volume versus elevation relationship within the 2D cell, and the same points are used along each face to define the conveyance versus elevation relationship. This helps better capture more details from input topography while maintaining the chosen input grid size.

			If a smaller output resolution (such as for velocity or depth outputs) is required, this is controlled by the 2D cell size and the nesting level of the Quadtree polygon, not by the SGS sampling settings. SGS improves the internal hydraulic calculations within each cell but does not alter the 2D grid or the output resolution.

			Note SGS provides the most benefit when the input DEM resolution is significantly smaller than the 2D cell size. If the DEM is similar to or coarser than the 2D cell size, SGS offers little improvement.
	<br>		<br>
	{{Tips Navigation		{{Tips Navigation

Pavlina Monhartova: /* When should I use a combination of SGS and Quadtree? */

2024-08-08T02:22:46Z

When should I use a combination of SGS and Quadtree?

← Older revision		Revision as of 12:22, 8 August 2024
Line 24:		Line 24:
	=When should I use a combination of SGS and Quadtree?=		=When should I use a combination of SGS and Quadtree?=
	Based on the benchmarking so far there seems to be little reason not to use SGS with Quadtree. The combination of SGS and Quadtree is particularly powerful for large models that require higher resolution in the focus area as this allows much larger cells away from the focus area without affecting the results.<br>		Based on the benchmarking so far there seems to be little reason not to use SGS with Quadtree. The combination of SGS and Quadtree is particularly powerful for large models that require higher resolution in the focus area as this allows much larger cells away from the focus area without affecting the results.<br>
	SGS was not made the default in the 2020-01 release because for some models, especially those with coarse cell sizes over highly variable terrain, substantially different results can be seen due to SGS picking up a lot more detail of the terrain within a 2D cell. Using Quadtree with SGS, and by using coarser cells, much better results and good results convergence can be achieved~~. It's highly likely that SGS will be made the default in a future release once there has been extensive industry benchmarking demonstrating using SGS to be consistently superior to not using it~~.<br>		SGS was not made the default in the 2020-01 release because for some models, especially those with coarse cell sizes over highly variable terrain, substantially different results can be seen due to SGS picking up a lot more detail of the terrain within a 2D cell. Using Quadtree with SGS, and by using coarser cells, much better results and good results convergence can be achieved.<br>
	<br>		<br>

Abigail.lillo at 00:03, 25 January 2024

2024-01-25T00:03:00Z

← Older revision		Revision as of 10:03, 25 January 2024
Line 84:		Line 84:
	=How to decide between using SGS, Quadtree and 1D open channel?=		=How to decide between using SGS, Quadtree and 1D open channel?=
	The number of 2D cells across channel is critical. Based on the <u>[https://downloads.tuflow.com/_archive/Australian_Rainfall_Runoff_Project15_TwoDimensional_Modelling_DraftReport.pdf ARR 2D modelling guidelines]</u> the aim is to have 6 cells across the river, creek or drain (if not using SGS). However, use of SGS does allow use of a much fewer cells across a waterway whilst preserving the total conveyance of the waterway. The only downside of SGS is that if the velocity distribution across the waterway, or for example the super-elevation of the water surface around a bend, is important then sufficient number of cells is needed to accurately reproduce the velocity distribution or water level variation. Quadtree grid can be useful for this purpose.		The number of 2D cells across channel is critical. Based on the <u>[https://downloads.tuflow.com/_archive/Australian_Rainfall_Runoff_Project15_TwoDimensional_Modelling_DraftReport.pdf ARR 2D modelling guidelines]</u> the aim is to have 6 cells across the river, creek or drain (if not using SGS). However, use of SGS does allow use of a much fewer cells across a waterway whilst preserving the total conveyance of the waterway. The only downside of SGS is that if the velocity distribution across the waterway, or for example the super-elevation of the water surface around a bend, is important then sufficient number of cells is needed to accurately reproduce the velocity distribution or water level variation. Quadtree grid can be useful for this purpose.

			=How can I convert 1D open channel into Quadtree?=
			Follow these steps to create a TIN from ESTRY cross section inputs that can be read into a Quadtree model: <br>
			# Use the following script that has been developed to convert ESTRY cross sections inputs into 2d_zsh line and points: https://gitlab.com/tuflow-user-group/tuflow/data-pre-processing/estry_to_zsh <br>
			# Digitise a 2d_zsh merge polygon around the 2d_zsh lines and points that define the open channel. Tutorial Module 2 provides instructions on how to create this 2d_zsh merge polygon: [[Tutorial_M02#Part_2_-_Other_Topographic_Updates \|Tutorial M02 Part2 - Other Topographic Updates]] <br>
			# Read the 2d_zsh line and points cross section inputs into the model using the "Read GIS Z Shape" command in the tgc file. <br>
			# Read the 2d_zsh merge polygon into the model using the "Create TIN Zpts Write TIN" command in the tgc file. <br>


	<br>		<br>

Pavlina Monhartova at 07:03, 28 November 2023

2023-11-28T07:03:37Z

← Older revision		Revision as of 17:03, 28 November 2023
Line 11:		Line 11:

	=When should I use SGS?=		=When should I use SGS?=
	Almost every time to take advantage of the higher accuracy. There is a threshold beyond which SGS isn’t required. The benefit of SGS is focused on cells with variety of elevations within each cell. For flat cells, where all SGS sample points have almost the same elevation, there won’t be much of a difference in comparison without SGS.<br>		Almost every time to take advantage of the higher accuracy. There is a threshold beyond which SGS isn’t required. The benefit of SGS is focused on cells with variety of elevations within each cell. For flat cells, where all SGS sample points have almost the same elevation, there won’t be much of a difference in comparison without SGS, as there is no detailed sub-cell terrain to sample.<br>
	<br>		<br>

Line 17:		Line 17:
	The main advantages of using Quadtree are shorter run times, lower GPU RAM footprint and smaller size of output files. Shorter run times are achieved with considerable cell count reduction compared with using a HPC grid, some models with cell count reduction of only around 30% may even run slower with Quadtree than the original HPC model.<br>		The main advantages of using Quadtree are shorter run times, lower GPU RAM footprint and smaller size of output files. Shorter run times are achieved with considerable cell count reduction compared with using a HPC grid, some models with cell count reduction of only around 30% may even run slower with Quadtree than the original HPC model.<br>
	Reasons for using Quadtree depend on modelling objectives and include:		Reasons for using Quadtree depend on modelling objectives and include:
	*Large models that require higher resolution in the focus area (e.g. the town is the focus but the model due to boundaries, etc goes much further afield)~~. The combination of SGS and Quadtree is particularly powerful as this allows much larger cells away from the focus area without affecting the results~~.		*Large models that require higher resolution in the focus area (e.g. the town is the focus but the model due to boundaries, etc goes much further afield).
	*Where high resolution velocity-based outputs (e.g. velocity, hazard, bed shear stress) are required. SGS is great for enabling the use of larger cells with good results convergence, but it still only produces one (average) velocity per cell. Quadtree will produce higher velocity output resolution (noting run times might become slower due to smaller cell sizes).		*Where high resolution velocity-based outputs (e.g. velocity, hazard, bed shear stress) are required. SGS is great for enabling the use of larger cells with good results convergence, but it still only produces one (average) velocity per cell. Quadtree will produce higher velocity output resolution (noting run times might become slower due to smaller cell sizes).
	*Often popular for urban models where the streets, which take the bulk of the flow, are modelled at a higher resolution than properties, parks, etc. Where buildings and other obstructions are in an urban flowpath a higher resolution along the flowpath may be utilised.		*Often popular for urban models where the streets, which take the bulk of the flow, are modelled at a higher resolution than properties, parks, etc. Where buildings and other obstructions are in an urban flowpath a higher resolution along the flowpath may be utilised.
Line 23:		Line 23:

	=When should I use a combination of SGS and Quadtree?=		=When should I use a combination of SGS and Quadtree?=
			Based on the benchmarking so far there seems to be little reason not to use SGS with Quadtree. The combination of SGS and Quadtree is particularly powerful for large models that require higher resolution in the focus area as this allows much larger cells away from the focus area without affecting the results.<br>
	Based on the benchmarking so far there seems to be little reason not to use SGS with Quadtree. The one exception is if the underlying resolution of the DEM is of similar (or coarser) resolution to the 2D cells, then there is little reason to use SGS as there is no detailed sub-cell terrain to sample. SGS was not made the default in the 2020-01 release because for some models, especially those with coarse cell sizes over highly variable terrain, ~~you~~ ~~can~~ ~~see~~ ~~substantially~~ ~~different~~ ~~results~~ due to SGS picking up a lot more detail of the terrain within a 2D cell. Using Quadtree with SGS, and by using coarser cells, much better results and good results convergence can be achieved. It's highly likely that SGS will be made the default in a future release once there has been extensive industry benchmarking demonstrating using SGS to be consistently superior to not using it.<br>		SGS was not made the default in the 2020-01 release because for some models, especially those with coarse cell sizes over highly variable terrain, substantially different results can be seen due to SGS picking up a lot more detail of the terrain within a 2D cell. Using Quadtree with SGS, and by using coarser cells, much better results and good results convergence can be achieved. It's highly likely that SGS will be made the default in a future release once there has been extensive industry benchmarking demonstrating using SGS to be consistently superior to not using it.<br>
	<br>		<br>

Pavlina Monhartova: /* When should I not use a combination of SGS and Quadtree? */

2023-11-28T06:54:17Z

When should I not use a combination of SGS and Quadtree?

← Older revision		Revision as of 16:54, 28 November 2023
Line 22:		Line 22:
	<br>		<br>

	=When should I ~~not~~ use a combination of SGS and Quadtree?=		=When should I use a combination of SGS and Quadtree?=
	Based on the benchmarking ~~thus~~ far there seems to be little reason not to use SGS with Quadtree. The one exception is if the underlying resolution of the DEM is of similar (or coarser) resolution to the 2D cells then there is little reason to use SGS as there is no detailed sub-cell terrain to sample. SGS was not made the default in the 2020-01 release because for some models, especially those with coarse cell sizes over highly variable terrain, you can see substantially different results due to SGS picking up a lot more detail of the terrain within a 2D cell. Using Quadtree with SGS, and by using coarser cells ~~within your model than you would have using a HPC grid~~, ~~you will achieve a~~ much better ~~result~~ and good results convergence ~~(as~~ ~~discussed~~ ~~further above)~~. It's highly likely that SGS will be made the default in a future release once there has been extensive industry benchmarking demonstrating using SGS to be consistently superior to not using ~~(which is the case thus far)~~.<br>		Based on the benchmarking so far there seems to be little reason not to use SGS with Quadtree. The one exception is if the underlying resolution of the DEM is of similar (or coarser) resolution to the 2D cells, then there is little reason to use SGS as there is no detailed sub-cell terrain to sample. SGS was not made the default in the 2020-01 release because for some models, especially those with coarse cell sizes over highly variable terrain, you can see substantially different results due to SGS picking up a lot more detail of the terrain within a 2D cell. Using Quadtree with SGS, and by using coarser cells, much better results and good results convergence can be achieved. It's highly likely that SGS will be made the default in a future release once there has been extensive industry benchmarking demonstrating using SGS to be consistently superior to not using it.<br>
	<br>		<br>

Pavlina Monhartova: /* When should I not use SGS? */

2023-11-28T06:48:54Z

When should I not use SGS?

← Older revision		Revision as of 16:48, 28 November 2023
Line 10:		Line 10:
	<br>		<br>

	=When should I ~~not~~ use SGS?=		=When should I use SGS?=
	There is a threshold beyond which SGS isn’t required. The benefit of SGS is ~~focussed~~ on cells with variety of elevations within each cell. For flat cells, where all SGS sample points have almost the same elevation, there won’t be much of a difference in comparison without SGS.<br>		Almost every time to take advantage of the higher accuracy. There is a threshold beyond which SGS isn’t required. The benefit of SGS is focused on cells with variety of elevations within each cell. For flat cells, where all SGS sample points have almost the same elevation, there won’t be much of a difference in comparison without SGS.<br>
	<br>		<br>

Pavlina Monhartova: /* When should I not use Quadtree? */

2023-11-28T06:47:34Z

When should I not use Quadtree?

← Older revision		Revision as of 16:47, 28 November 2023
Line 14:		Line 14:
	<br>		<br>

	=When should I ~~not~~ use Quadtree?=		=When should I use Quadtree?=
	The main advantages of using Quadtree are shorter run times ~~(with considerable cell count reduction compared with using a HPC grid)~~, lower GPU RAM footprint and smaller size of output files. If ~~you're~~ ~~not~~ ~~achieving~~ ~~this~~ by ~~using~~ ~~Quadtree~~ ~~then~~ ~~there~~ is ~~little~~ ~~benefit~~ in ~~using~~ ~~it.~~ ~~Some~~ models with cell count reduction of only around 30% may even run slower with Quadtree than the original HPC model.<br>		The main advantages of using Quadtree are shorter run times, lower GPU RAM footprint and smaller size of output files. Shorter run times are achieved with considerable cell count reduction compared with using a HPC grid, some models with cell count reduction of only around 30% may even run slower with Quadtree than the original HPC model.<br>
			Reasons for using Quadtree depend on modelling objectives and include:
			*Large models that require higher resolution in the focus area (e.g. the town is the focus but the model due to boundaries, etc goes much further afield). The combination of SGS and Quadtree is particularly powerful as this allows much larger cells away from the focus area without affecting the results.
			*Where high resolution velocity-based outputs (e.g. velocity, hazard, bed shear stress) are required. SGS is great for enabling the use of larger cells with good results convergence, but it still only produces one (average) velocity per cell. Quadtree will produce higher velocity output resolution (noting run times might become slower due to smaller cell sizes).
			*Often popular for urban models where the streets, which take the bulk of the flow, are modelled at a higher resolution than properties, parks, etc. Where buildings and other obstructions are in an urban flowpath a higher resolution along the flowpath may be utilised.
	<br>		<br>