Dear Xinxin Song ,

First thing you should consider is the density of points per unit area (points/area).

So I suggest calculating the number of points for each zone, based on the density (points/area) and from there, choosing where the sample points will be collected (allocated).

You can create "regular grids" to allocate the points, but in the field, this is often not possible due to inaccessible locations. Therefore, it is more important to verify whether the spatial distribution of the points is not clustered, but randomly or preferably evenly spread in space.

You can use some methods to perform the point pattern analysis, which can be cited: the quadrant counts, nearest-neighbor distance, and the G, F and K functions.

There is a tool in ArcGIS that allows to perform the analysis of the distribution of points through the Ripley's K-function. The description of this tool is given through the link below:

https://pro.arcgis.com/en/pro-app/tool-reference/spatial-statistics/multi-distance-spatial-cluster-analysis.htm

I hope this hint has helped you.

In my opinion you can try some simulation in Sagagis.

Thank you for your answer Leandro França . Yes the planned density is 30 samples per ha. I am just not sure how to allocate the 30 samples according to the zones within the one hectare if you know what I mean. For example, the 3 zones in a ha have different sizes, I think the allocated number of the samples should be based on the size of each zone, but it's like very hard to determine exactly.

The tool in Arcgis sounds good unfortunatelyI don't use ArcGis but QGIS. I will see if there is similar options. Thanks!

Thank you Adrian Andrei however my project doesn't use SagaGis, so it's not an option for me

You're welcome Xinxin Song ,

The number of samples for each zone must be proportional to its respective area.

I also suggest creating a regular grid whose size in x,y must be the square root of the total area divided by the number of samples.

If you collect a sample inside each squared polygon of the grid, then you sure that your points will be evenly spread in space.

In relation to the Ripley's K-function, I've suggested the ArcGIS tool, because I'd tested a time ago. However, you can also perform this spatial point pattern analysis in R, accordingly the link bellow:

https://training.fws.gov/courses/references/tutorials/geospatial/CSP7304/documents/PointPatterTutorial.pdf

Leandro França Yes I agree with you that the number of samples for each zone needs to be proportional to the respective area. However, the shape of each zone is very irregular and it's difficult to calculate the area.

Xinxin Song , could you share a shapefile with the zones? I'll use some QGIS tools to show you how to create a regular grid over the study area. We can do that even on very irregular areas.

Hi Leandro França thank you. Please see the attached file. I installed a plugin QSGR tool in QGIS which can create grid. However, I prefer to use a random distribution of sample points rather than grid sampling. From the attached image you can see that the red zone is smaller than the yellow zone. So I don't know how many samples should be assigned to the red and other colours on a proportional basis if use a random distribution.

Hi Xinxin Song ,

I just needed to use the native QGIS tools.

In the first case I created a regular grid with rectangles of dimension 22m by 16.67m with the tool "Create Grid" (Figure 1), then I selected all the created features and, with the vector editing tool "Rotate Feature" and "Move Feature", I adjusted the position of the created grid to the Boundary layer that you made available, saving only the 30 rectangles that were inside the area.

These 30 features correspond to the locations where samples are to be collected. However, if you prefer, you can use your centroids (Figure 2).

For the case of generating random points, I used the tool "Random points inside polygons" (Figure 3), but the result is not so well distributed when compared to the regular grid (Figure 4).

I'm also making the results available as shapefiles for you to check out on your QGIS.