Problem with sample_lsm() about landscapemetrics

[mhesselbarth]

Actually all possible options (level, metric, type, what, ...) needs to be NULL. Definitely worth to mention. I will update the documentation next week.

[mhesselbarth]

Oh...we forgot to adapt sample_lsm() to the new structure how metrics are selected (basically list_lsm()). However, list_lsm() doesn't has the option what = "all" anymore, but all metrics are selected if nothing is specified at all. Do we want to bring back the "all" option?

Anyways, should work now (ddfa819)

library(landscapemetrics)
data("landscape")
points <- matrix(c(10, 5, 25, 15, 5, 25), ncol = 2, byrow = TRUE)
points_sp <- sp::SpatialPoints(points)
sample_lsm(landscape,
           points = points_sp, 
           size = 15)

[Nowosad]

Thanks @mhesselbarth.
Bringing back the all option is not necessary, but it would be nice to add info that what = NULL means all metrics are calculated...

[Nowosad]

One more question:
sample_lsm() with return_plots = TRUE gives a tibble with two list columns - one for the landscape summary and one with new rasters. Why is that? Shouldn't it be just the same output as the first one, but with one additional column for rasters?

library(landscapemetrics)
points <- matrix(c(10, 5, 25, 15, 5, 25), ncol = 2, byrow = TRUE)
sample_lsm(landscape, points = points, size = 15, what = "lsm_l_np")
#> # A tibble: 3 x 8
#>   layer plot_id level     class    id metric value percentage_inside
#>   <int>   <int> <chr>     <int> <int> <chr>  <dbl>             <dbl>
#> 1     1       1 landscape    NA    NA np         8              80. 
#> 2     1       2 landscape    NA    NA np         7              74.7
#> 3     1       3 landscape    NA    NA np         7              64.
sample_lsm(landscape, points = points, size = 15, what = "lsm_l_np", return_plots = TRUE)
#> # A tibble: 3 x 4
#>   layer plot_id metrics          raster_sample_plots
#>   <int>   <int> <list>           <list>             
#> 1     1       1 <tibble [1 × 8]> <S4: RasterLayer>  
#> 2     1       2 <tibble [1 × 8]> <S4: RasterLayer>  
#> 3     1       3 <tibble [1 × 8]> <S4: RasterLayer>

^{Created on 2019-01-19 by the reprex package (v0.2.1)}

[Nowosad]

(Btw, the name return_plots could be confusing - I assume it would give some kind of figure as an output. How about using return_rasters instead?)

[mhesselbarth]

Sounds good!

[mhesselbarth]

Done (11df79e)

[Nowosad]

One more question:
sample_lsm() with return_plots = TRUE gives a tibble with two list columns - one for the landscape summary and one with new rasters. Why is that? Shouldn't it be just the same output as the first one, but with one additional column for rasters?

library(landscapemetrics)
points <- matrix(c(10, 5, 25, 15, 5, 25), ncol = 2, byrow = TRUE)
sample_lsm(landscape, points = points, size = 15, what = "lsm_l_np")
#> # A tibble: 3 x 8
#>   layer plot_id level     class    id metric value percentage_inside
#>   <int>   <int> <chr>     <int> <int> <chr>  <dbl>             <dbl>
#> 1     1       1 landscape    NA    NA np         8              80. 
#> 2     1       2 landscape    NA    NA np         7              74.7
#> 3     1       3 landscape    NA    NA np         7              64.
sample_lsm(landscape, points = points, size = 15, what = "lsm_l_np", return_plots = TRUE)
#> # A tibble: 3 x 4
#>   layer plot_id metrics          raster_sample_plots
#>   <int>   <int> <list>           <list>             
#> 1     1       1 <tibble [1 × 8]> <S4: RasterLayer>  
#> 2     1       2 <tibble [1 × 8]> <S4: RasterLayer>  
#> 3     1       3 <tibble [1 × 8]> <S4: RasterLayer>

Created on 2019-01-19 by the reprex package (v0.2.1)

@mhesselbarth @marcosci and what do you think about that?

[mhesselbarth]

The reason why we nested the tibble was that I thought it becomes quite tricky if you select e.g. a patch-level metric and a class-level metric. In this case, you don't know the number of rows the resulting tibble has because that depends on the number of patches and classes in each sample plot.

But actually, using the plot_id this is straightforward. So IMHO, we could just return the tibble with an extra row (see 9630236; @marcosci please revert if you don't like that).

[mhesselbarth]

Please reopen if someone is not happy with the solution

> points <- matrix(c(10, 5, 25, 15, 5, 25), ncol = 2, byrow = TRUE)
> sample_lsm(landscape, points = points, size = 15, what = c("lsm_c_ca", "lsm_l_np"), return_raster = TRUE)
# A tibble: 12 x 9
   layer plot_id level     class    id metric  value percentage_inside raster_sample_plots
   <int>   <int> <chr>     <int> <int> <chr>   <dbl>             <dbl> <list>             
 1     1       1 class         1    NA ca     0.0044              80.  <S4: RasterLayer>  
 2     1       1 class         2    NA ca     0.0051              80.  <S4: RasterLayer>  
 3     1       1 class         3    NA ca     0.0085              80.  <S4: RasterLayer>  
 4     1       1 landscape    NA    NA np     8                   80.  <S4: RasterLayer>  
 5     1       2 class         1    NA ca     0.0016              74.7 <S4: RasterLayer>  
 6     1       2 class         2    NA ca     0.0054              74.7 <S4: RasterLayer>  
 7     1       2 class         3    NA ca     0.0098              74.7 <S4: RasterLayer>  
 8     1       2 landscape    NA    NA np     7                   74.7 <S4: RasterLayer>  
 9     1       3 class         1    NA ca     0.0006              64.  <S4: RasterLayer>  
10     1       3 class         2    NA ca     0.0038              64.  <S4: RasterLayer>  
11     1       3 class         3    NA ca     0.01                64.  <S4: RasterLayer>  
12     1       3 landscape    NA    NA np     7                   64.  <S4: RasterLayer>