RE: Modelling

[brenwin1]

data collection

Just finished collecting all the variables and cleaning up to fit the map structure.

Currently the following variables are available:

id: raster cell number (1 to 400) year month fire_count: no. of fire ignitions in that cell daily_rain: daily rainfall (in mm) et_short_crop: short crop evapotranspiration (like drought index) max_temp: max temperature radiation: solar radiation - total incoming downward shortwave radiation on a horizontal surface rh: relative humidity si10: 10m wind speed lai_hv: leaf area index for high vegetation lai_lv: leaf area index for low vegetation forest: 1(forest exist in the cell) or 0(otherwise)

monthly from 2016 (Oct) to 2021 most weather variables available at a daily rate *note: NA values comes from the cells that are not within Victoria (e.g. in the sea)

---

plotting response (fire_count vs. predictors)

Plotting response against predictors end up like the plot of fire_count against max_temp above. Many 0 response value since no fires occured in about half the cells throughout the time frame. The many 0's pulls the OLS line(in blue) down.

For example many cells don't have fires throughout but still experience high temperature causing problems.

*currently fitting a random forest model and creating lag variables

Any input would be helpful!

[dicook]

Try plotting on a log scale, or just the model alone. We know it is a weak model, so trend in the line will be small relative to variation. This is where it is ok to plot the model, without the data.

This is also where looking at temperature in relation to other variables like precipitation, or vegetation.

On 28 Oct 2021, at 2:44 pm, brenwin1 @.***> wrote:

data collection

Just finished collecting all the variables and cleaning up to fit the map.

Currently the following variables are available:

id: raster cell number (1 to 400) year month fire_count: no. of fire ignitions in that cell daily_rain: daily rainfall (in mm) et_short_crop: short crop evapotranspiration (like drought index) max_temp: max temperature radiation: solar radiation - total incoming downward shortwave radiation on a horizontal surface rh: relative humidity si10: 10m wind speed lai_hv: leaf area index for high vegetation lai_lv: leaf area index for low vegetation forest: 1(forest exist in the cell) or 0(otherwise)

monthly from 2016 (Oct) to 2021 many weather variables available at a daily rate *note: NA values comes from the cells that are not within Victoria (e.g. in the sea)

plotting response (fire_count vs. predictors)

Plotting response against predictors end up like the plot of fire_count against max_temp above. Many 0 response value since no fires occured in about half the cells throughout the time frame. The many 0's pulls the OLS line(in blue) down.

For example many cells don't have fires throughout but still experience high temperature causing problems.

*currently fitting a random forest model and creating lag variables

Any input would be helpful!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

cheers, Di

---

Dianne Cook @.***