Representing agricultural Land Use Change over time

[Urs Hippelein]

Hello everyone, hello SWAT+ team,

I am currently working on my MSc thesis and the goal is to expand an already existing SWAT+ model for a catchment in Northeastern Brazil with agricultural land use change over time (and then calibrate for crop yield).

The model I am working with has already been set up, but it currently only accounts for land use in a single year (2015), since the input land use dataset is static (and SWAT+ doesn't account for several land use maps over time").

I am mainly reaching out here, because I am interested in conceptual flaws or limitations in the approaches that might lead to them not working, as my understanding of certain model parts have come to a limit (with a lot of research in the forum here and a SWAT+ Advanced workshop that I went through).

-------

Available data:

• Spatial component: Annual LCLU maps indicating the location of “agricultural area” (without distinction between crop types).
• Statistical component: Annual municipality-level statistics on crop area (ha) and crop yield for different crops (e.g., sugarcane, cashew).

Previous approach (existing model):
It was addressed with the Split Land Use Tool in QSWAT+, where agricultural areas were defined separately for each municipality, and crops were split proportionally based on statistics.

My constraints are: SWAT+ requires static input --> I have to use decision tables. No spatial information about where which crops are grown exists within the agricultural area.
Example: Sugarcane in municipality 1:
2015–2017: 250 ha

2018: 280 ha
2019: 0 ha

Approaches I am considering:
1. Grid-based approach

• Create polygons (e.g., one 200 ha “plot” and one 30 ha “plot” for sugarcane) and introduce them into the land use map as separate land uses (e.g., sugc_municipality_x). Each polygon is recognized as a distinct land use, creating separate HRUs.
• Decision tables would then apply planting/harvesting operations for the corresponding years (or switch to a placeholder land use in years when the crop is absent).
• Alternative: use Scenario Land Use decision tables to switch the specific HRUs between “crop” and “default agriculture/placeholder” dynamically.
• Doubts: I have thought about this approach the most and it seems promising, it requires a lot of manual preparation tho and is not very replicable for future implementations/years.

2. Plant community approach

• Define a plant community for each municipality-year combination (e.g., municipality1_2015, municipality1_2016, etc.), each with a linked LUM
• Use Land Use Scenario decision tables to switch the landuse (and LUM) annually (from "municipality1_2015" to "municipality1_2016") and hence switch the corresponding plant community.
• To approximate percentages, let all crops grow simultaneously and adjust initial biomass [kg/ha] to match the proportional crop area
• Concerns: This assumes the crops coexist on the same land unit, which may not reflect reality and I have a limited understanding of how the model handles simultaneous growth of multiple crops (but it was mentioned in other conversations here).
  The inital biomas approach should work for permanent crops, but not for temporary ones.
  There might be issues with the model as I got the message "the model expects LUM with the names [crop-name]_lum..." (I think this can be avoided).

3. HRU fraction update

• I got told by my supervisor, that in older SWAT versions, there was the “land use update file.” option to change HRU areas
• SWAT+ documentation mentions a decision table action type called hru_fr_update (HRU area fraction change).
• However, I have not found further documentation or examples of its implementation, neither in an advanced workshop, nor in the forum here,...
  

----------

I would highly appreciate feedback from the SWAT+ team or from anyone with experience or similar issues. I have looked through the Google groups, but came to a limit about information about DT and plant communities.

Since I don't really have the time to try out every single approach in detail, suggestions about the doability/feasibility would really help me (as mentioned: possible conceptual flaws or limitations in the approaches) and which approach you would consider the most promising.
At this point, I feel that my conceptual understanding of certain SWAT+ functions has reached its limit and I am also lacking a person with a deep model understanding to discuss this with.

Thank you!
Urs Hippelein