Planning and Research

Record of the discussion following page 1. Firming up the plan as a premise for translation (overseas deployment) ｜ 2026-10-19

Contents: 0. Plugin premise ／ 1. Problem statement ／ 2. Organizing the issues ／ 3. Decided premises ／ 4. What the three site data elements are ／ 5. Ways to supply site index ／ 6. Country comparison ／ 7. Open items

0. Plugin premise: how the site index works

Added 2026-10-19. Based on comparing the "Guide to Forest Zoning Considering Profitability and Disaster Risk" (Forestry Agency, revised March 2026, in Japanese; freely available from the G-Spatial Information Center: Forest Zoning Guide (R8.3 revision)) with the plugin code (processes/raster_writer/siteidx.py, profit.py)

The guide's formula differs from the code's formula

Formula in the MEMO on p.70 of the guide:

Sugi site index = 21.28 + 1.028*NPP − 1.001*SRAD − 1.111*VTEX

Formula the plugin actually calculates:

Sugi site index = 21.28 + (NPP − 15.82)×1.028 − (SRAD − 1347.09)×0.01×1.001 − (VTEX − 58.85)×0.01×1.111

The guide's formula omits subtracting a reference value from each input (15.82, 1347.09, 58.85) and the ×0.01. Calculated as written in the guide, with SRAD around 1400 the result would be something like −1400 m, which makes no sense. This is an issue with how the original guide is written.

How to read the formula

It has the form "tree height at an average site ± adjustments for conditions".

a (21.28 for sugi): the site index when NPP, SRAD and VTEX are at their reference values (b, d, f; probably the means of the fitting data). At an average site, about 21 m at age 40
Each term: adds for conditions better than the reference, subtracts for worse ones
×0.01: likely a conversion to align the magnitude of SRAD and VTEX values (basis unconfirmed)

What the three elements are and why they matter

Element	What it is	Effect on site index	Rationale
NPP (net primary productivity)	How much plants can grow in a year on that land, determined by climate alone: carbon absorbed by photosynthesis minus carbon released by respiration. Calculated on a 1 km grid by running Prof. Mitsuda's growth model with climate data	Positive (higher means better site)	Trees grow better where it is warm and the air is not too dry. Represents broad-scale (regional) differences
SRAD (solar radiation index)	How sunlight hits the surface, determined by slope aspect and steepness. Highest on steep south-facing slopes	Negative (stronger sun means poorer site)	Strong sunlight dries the soil. Sugi is sensitive to drought, so it grows better on shaded north-facing or valley slopes
VTEX (topographic exposure / concavity)	How far the cell looks down on low points in the eight surrounding directions. High on ridges, low in valleys	Negative (ridges mean poorer site)	Ridges are wind-exposed and dry, and water does not collect there. Valleys collect water and stay moist

In short, the idea is to take "the land's base capacity determined by climate (NPP)" and subtract "susceptibility to drying (SRAD, VTEX)". The statement on p.16 of the guide that "topographic indices related to soil moisture, which strongly affects plant growth, … are highly correlated with site index" is the reason SRAD and VTEX are used.

Differences in coefficients between species

Species	NPP (climate)	SRAD (sunlight)	VTEX (ridge)	Character it suggests
Sugi	1.028	1.001	1.111	Sensitive to both climate and drying. Prefers moist valleys and north-facing slopes
Hinoki	0.969	0.093	0.862	Sensitive to climate, but almost unaffected by sunlight
Karamatsu	0.129	0.294	1.125	Climate differences matter little; ridge versus valley is the main factor

How much each element matters (estimate with zone VIII values, sugi)

Values obtained by putting the 1st–99th percentiles of the data (table on page 1) into the formula.

NPP: −4.5 m to +1.6 m (range about 6 m)
SRAD: +6.8 m to −3.1 m (range about 10 m)
VTEX: +0.7 m to −1.2 m (range about 2 m)

For sugi in zone VIII, sunlight (SRAD) has the largest effect.

How the plugin uses it

Create site index rasters for the three species (siteidx.py)
The user picks one species and sets two thresholds to divide it into scores of 1, 2 and 3 (profit.py)
That score is added to the scores for "yarding efficiency" and "distance from roads" to give "profitability"
Profitability and disaster risk are combined for zoning

The site index is used only to divide sites into scores; the height values themselves do not remain in the calculation. This is why the guide says to "treat it as a relative value": the design works as long as the ranking of sites is correct. Outside Japan, the formula need not be reproduced exactly; any indicator that correctly ranks site quality can be used.

1. Problem statement (user)

If morizon_next copy is to be translated: page 1 lists Spain, Canada and Brazil as examples, but whether to proceed with these as the first countries needs to be considered
Assuming open data is used as page 1 suggests, there is also the question of whether that is possible
The plan needs to be firmed up as a premise

Note: the CRS notation in the code mentioned in page 1 (JGD2011 vs. JGD2000) is an issue of the original MORIZON and is not addressed.

2. Organizing the issues (Claude)

Current state of translation

Files using tr(): 0. Python files with hard-coded Japanese: 74; .ui files: 3; docs/manual.html is also in Japanese. There is no translation infrastructure yet.

Translation and localization are separate tasks

Translation: wrapping strings in tr() and preparing .ts/.qm files. Needed for any country and can proceed without waiting for the country choice
Localization: replacing site index, DEM, CRS and harvesting systems. Directly depends on the country choice and open data feasibility

Questions on whether open data really works

Are the definitions known: how NPP, SRAD and VTEX are made (prerequisite)
Licence: does the plugin fetch source data at run time, or do we host and distribute processed data
Access: stable URLs, whether authentication is required, whether automated access is permitted (precedent: blocked by the Shizuoka Prefecture open data portal)
Scientific validity: the gap between satellite canopy height and "dominant tree height". A trial in one region is needed

Proposed steps

Decide the purpose and audience
Confirm the definitions of the three elements
Compare candidate countries
Open data feasibility for each candidate country
Trial in one region (translation infrastructure in parallel)

3. Decided premises (user)

Item	Answer
Purpose	Simply interested in whether there are countries where this system can be used effectively; no local contacts are assumed. As a characteristic of QGIS, if users can prepare the data and the plugin can guide or create that preparation, it can be offered as a good tool. Interested in this as a designer
Translation scope	Everything, as a general-purpose tool. English first. Depending on where the open data comes from, that country's language may take priority
Next task	Proceed with research (definitions of the three elements, country comparison, open data feasibility)

4. What the three site data elements are (research)

Source

"Mitsuda, Y. and Kitahara, F. (2017)" is a conference presentation, not a paper: Mitsuda, Y. and Kitahara, F. "Estimating the distribution of site index of sugi, hinoki and karamatsu across Japan", 128th Annual Meeting of the Japanese Forest Society (Kagoshima), 27 March 2017. An outcome of KAKENHI 15K07483 (2015–2017, PI: Yasushi Mitsuda). The abstract itself was not yet confirmed at this point (see page 3).

Definitions (from the papers)

For the meaning of each element and why it affects the site index, see 0. Plugin premise.

Element	Definition	Can it be made abroad?
SRAD	Solar radiation index. Calculated from the solar zenith and azimuth angles and the slope and aspect of the surface (Smith et al. 1980). Sunlight intensity related to soil drying. 50 m	Yes From a DEM with the QGIS raster calculator
VTEX	Vertical exposure index. Sum of nadir angles in eight directions (Mitsuda et al. 2007). Susceptibility to wind-driven evapotranspiration. 50 m	Yes Search distance etc. to be confirmed in Mitsuda et al. 2007
NPP	Mean growth over 10 years simulated on a 1 km grid with a carbon-balance-based stand growth model, with uniform initial conditions. Climate 1971–2000. Higher where temperature is high and atmospheric dryness is low	Not the same thing Output of a model tuned to Japanese species

The fitting used the GSI 50 m DEM, the Japan Meteorological Agency 1 km gridded climate data, and the nationwide Japanese National Forest Inventory (NFI).

There is only one NPP raster, used in common for all three species. Which species' model was used must be confirmed in the abstract.

Conclusion: outside Japan, it will be rebuilt as "a climate factor replacing NPP (MODIS NPP 500 m, WorldClim temperature and dryness, etc.) + SRAD and VTEX + coefficients fitted locally". The form of the formula (a constant plus three linear terms) can be kept.

5. Ways to supply site index (design draft)

Three levels, assuming users prepare the data and the plugin guides or creates it.

Level	What the user prepares	What the plugin does	Suited regions
A. Ready-made site index	A local site index raster	Point to sources, align CRS and extent	New Zealand, British Columbia (Canada), etc.
B. Calculate from three elements	DEM, climate data (download guided)	Automatically create SRAD and VTEX from the DEM. Coefficients are defaults or entered manually	Regions without a ready-made site index
C. Fit coefficients	Plot CSV with height, age, species and coordinates	Sample the three elements at plot locations and fit coefficients	Users with survey data. Vegetation survey summaries come in here

The current plugin's "custom site index" input field is the basis for A. B and C are new
For licensing, the safest approach is not to bundle or relay data, but to guide users to obtain it themselves. Even non-commercial, no-derivatives data (Scion) remains usable by users themselves
Other replacements: CRS (UTM, etc.), DEM (global default Copernicus DEM 30 m), buildings and roads (OpenStreetMap), yarding efficiency (existing editor), reference age (configurable per species)

6. Country comparison (first pass)

Criteria: fit with the tool's premise (dividing steep plantation forests into production and protection), data, ready-made site index, language. The three countries in page 1 were examples without a stated rationale, so New Zealand was added.

	New Zealand	Canada (BC)	Spain	Brazil
Language	English	English, French	Spanish	Portuguese
Fit with premise	Steep radiata pine plantations (share of cable yarding unconfirmed)	Steep terrain exists but mainly natural forests	Pine and eucalyptus plantations in northern mountains	Short-rotation eucalyptus, mostly flat
DEM	National LiDAR 1 m, CC BY 4.0, over 80% of the country	HRDEM 1–2 m (OGL, partial coverage), MRDEM 30 m (full coverage)	PNOA LiDAR MDT 2 m/5 m (CC BY 4.0 to be confirmed)	No national high-resolution DEM confirmed
Ready-made site index	Scion site index and 300 Index (25 m). CC BY-NC-ND 4.0	PSPL (22 species). Licence unconfirmed	Not found	Not found
Plot data	Unconfirmed	Partly application-based; the province protects coordinates	Georeferenced since IFN3 (Access format), licence unconfirmed	State CSVs have diameter only; licence shown as "none"

Global common data: Copernicus DEM GLO-30 (free; attribution notice and disclaimer required), MODIS NPP (stated as having no restrictions on use or redistribution).

Provisional view (not a decision): 1. New Zealand　2. British Columbia, Canada　3. Spain (gateway to the Spanish-speaking world)　4. Brazil (candidate for later)

7. Open items

#	Item	Purpose	Status
1	Full text of Mitsuda et al. (2007) J. For. Res. 12(3) (paywalled)	Exact calculation of VTEX and SRAD	Open
2	Abstract of the 128th Annual Meeting of the Japanese Forest Society	Which species' model was used for NPP	Open
3	Licence of PNOA MDT02/MDT05	Confirm on the CNIG page	Open
4	Licence and format of BC PSPL	Whether it can be used for level A	Open
5	Licence of Spanish IFN plot data	Whether it can be used as training data for level C	Open
6	Licence and current coverage of the LINZ national DEM	The statements are not consolidated in one place	Open
7	Reference age of New Zealand's site index	Recalled as 20 years; needs confirmation	Open
8	Share of cable yarding and availability of plot data in New Zealand	Support for fit with the premise	Open
9	How to obtain the Scion site index (whether Koordinates requires login)	Whether users can be guided	Open