achubaty / mpb Goto Github PK

once we have a working copy of the crop-reproject function/module from @MelinaHoule we can use it instead of separate (yet otherwise identical) data transformation steps for studyArea.

• mpbClimateData (née mpbClimate)
• mpbPine
• mpbMassAttacksData

Currently, we use the Beaudoin et al. (2014) kNN percent pine data, but we really need to get an estimate of stem density (i.e. stems/ha or stems/pixel) in order to model the MPB dynamics.

kNN-StructureBiomass.tar provides pine volume ($m^3/ha$ at 250 m x 250 m). We can also use age (AGE), as well as total volume (VOL_T) multiplied by proportion pine per species, together with stand density managment (SDM) diagrams to get estimates of stems per hectare. However, this isn't necessarily clear-cut (no pun intended).

I've added @BarryCooke's notes on this topic — which follow from our discussions at GLFC last month — to the repo in bf9b19b. Updated in 81cf999.

There's a gap in the AB data from the website (1999 and 2000 are missing), so the raster only uses 1998, 2001-2016.

TO DO: use the 1997-2011 AB data to fill in the gap.

• reflect changes to climate and pine modules
• add corresponding part of mpbMassAttacksData module

• using R for symlinking is creating the link as a subdirectory. fix this.
• make sure the symlinking code works on Windows

Per discussions with Barry, a smaller study area spanning roughly from Lesser Slave Lake to the west, past Cold Lake and across the SK border to the east will be used. This area include the Martin Hills in AB.

I used Google Earth to draw a polygon (https://github.com/achubaty/MPB/blob/master/SpaDES/studyArea/studyArea.kml), which I converted to a shapefile.

TO DO: refine this further

After following up with Remi, we have a means to invoke BioSim from the commandline by passing the -e option using BioSim11.

@BarryCooke will recreate/import the necessary project (XML) files to produce the MPB climatic suitability projections using each of the previously used indices (R, S, L, G) plus the MPB iMod index (I?).

Use the BioSim/ directory for these files.

See also #4

from 4cf688d

• brick currently has no values / NA
• stack is a brick(?)

• update module data descriptions from report draft

• add studyArea polygon to the dropdown list and make it default

• use the selected polygon as the area to run for demo

• add SpaDES/MPB logos (in sidebar?)

• create input widgets for parameters:

  • mpbClimateData$index
  • mpbRedTopGrowth$dataSet
  • mpbRedTopSpread$dispersalKernel

complete the transition of relevant parts of Eliot's old code in cleanup.maps.R into the MPB_maps.Rmd file.

• update paths
• update object names

BC poly data is huge, and something is buggy/wrong with R that is preventing release of memory when working on these files. Thus RAM usage quickly exceeds 12-16 GB (likely more).

Suggestions:

• store the SpatialPolygonsDataFrames in an environment instead of a list (will require some custom looping code because lapply returns a list, so is not suitable for this task)

Since this module is simply doing data import and cropping to the study area, it makes sense to adjust the name.

• use separate growth curves for lodgepole and jack pine

@BarryCooke

mpb-maps/climate-projections.Rmd needs to be updated to explain the generated maps.

• How were these generated, and how can they be regenerated as needed? Include scripts/code.

• What do the maps represent, exactly? What do the raster values mean and how should they be interpreted?

  The response is a float number which can be used as a climatic multiplier on MPB baseline reproductive rate.

  Not sure if this is totally accurate, since values aren't bound between [0, 1].

• What are the caveats to using these maps (i.e., assumptions, limitations, etc.)?

• What is the availability of the maps and their underlying data? Are they publicly available? From where can they be downloaded? With whom can they be shared?

MPB mass attack data from AB and BC are available as points and polygons. Our first crack at generating rasters of red atttacked trees was to use both the points and polygons together for each province. However, looking more closely at the data, while there are differences in the locations for points and polygons, there appears to be some overlap.

I'm not sure how best to deal with points that are inside polygons. What does it mean when a polygon has NUM_TREES attacked but also has points with their own NUM_TREES values? Should the points values be added to the pixel value derived from the polygon average? Or does the points value represent the correct value for that pixel? @BarryCooke, any thoughts about how to effectively merge the point data with the polygons to generate the rasters?

My workaround for now was to produce the rasters using only points for AB and only use polygons for BC (c755614), but this likely needs to be revisited.

Note to self: use points.in.polygon() to determine which points are inside polygons.

Some of the scripts from McIntire-lab repo were (re)moved so there are pieces of this .Rmd that no longer run.

• boreal.R was drastically modified into and Rmarkdown file, and lots of additonal stuff was added and the file became part of mapLoadingManip.Rmd and lazyMaps... Need to re-extract only the download, extract, and load components. Do we need to reproject etc? I don't think so.

mostly completed now, but a few pieces can't yet be implemented:

• fasterize can do point data
• fasterize doesn't have the option to calculate pixels based on 'percent cover' of a polygons in each cell

something is still weird (buggy) with the insect_spread function

• edge problem: neer stops running if spreads off map
• weird overlap and shape when doing repeated iterations of insect_spread

suppressPackageStartupMessages({
  library(amc)
  library(data.table)
  library(sp)
  library(quickPlot)
  library(reproducible)
  library(rgdal)
  library(raster)
  library(SpaDES.tools)
})

clearPlot()
dev.useRSGD(FALSE)
dev()

modulePath <- "~/Documents/GitHub/MPB/SpaDES/modules"

## spread
source(file.path(modulePath, "mpbRedTopSpread", "R", "insect_spread.R"))

# create a dummy raster
XMAX <- YMAX <- 100
r <- raster(extent(0, XMAX, 0, YMAX), res = 1)
pine <- gaussMap(r)
pine[] <- pine[] / maxValue(pine)

r2 <- insect_spread(pine, loci = sample(ncell(r), 3),
                    asymmetry = 5, asymmetryAngle = 75,
                    lambda = 0.12,
                    saturationDensity = 332,
                    total = round(1125 * (250 / 100)^2),
                    debug = TRUE)

# plot the result
clearPlot()
Plot(r2, zero.color = "blue", new = TRUE, col = "Reds")

for (i in 1:1) {
  ids <- which(r2[] > (332 * 2/3))
  r2 <- insect_spread(pine, loci = ids,
                      asymmetry = 5, asymmetryAngle = 90,
                      lambda = 0.12,
                      saturationDensity = 332,
                      total = round(1125 * (250 / 100)^2),
                      debug = FALSE)
  Plot(r2)
}

.gc()

1. Split out each module into a separate repo (https://help.github.com/articles/splitting-a-subfolder-out-into-a-new-repository/)
2. relink these modules here as git submodules (https://git-scm.com/book/en/v2/Git-Tools-Submodules)

Since this module isn't generating the forecast, merely reading them in, it should be renamed to reflect that.

@BarryCooke

Instead of writing multiple otherwise identical modules that differ only by their data source and model fit, I will merge the mpbGrowth_B79fit and mpbGrowth_B79forced modules into a single mpbRedTopGrowth module that takes a parameter identifying which dataset / model fit to use.

• use 2001-2011 data for calibration/POM
• finalise scripts for POM

need to add bib refs for r
values-data.Rmd

• Ried 1962
• Cole and Amman 1969; Amman and Cole 1983

packrat was previously removed due to corrupted library (issues with the old version of the package?).

we should be using it.