sdtaylor / phenograssreplication Goto Github PK

Would be nice to include all the original sites here. 2 are missing for unknown reasons

From the original paper

1. Fit model to all timeseries/all years
2. Do leave-one-out CV, leaving scaling factor h constant to prior fitted value.
   2a. look at mean/var of each parameter as well as avg R2

Writing

• finish discussion
• finalize map
• abstract
• list our software cites
• #7
• cite supplement tables in text
• acknowledgments
• correct R2 everywhere
• bold text in results table
• include CV result in result table

Supplement

• table of individual site info/ R2 values, probably rmse too
• marker whether a site was included in original paper
• giant figure of fitted timeseries

Submission logistics

• collect references
• do latex tables
• get into overleaf w/ biogeosciences template
• upload repo + GrasslandModels package to zenodo
• repo readme

Needs:
soil water holding capacity
plant wilting point

ideally in mm or some easily convertible unit

Currently using this one from the original analysis:
Global Gridded Surfaces of Selected Soil Characteristics (IGBP-DIS)

Others:

Everything below is from:
https://www.isric.org/explore/soil-geographic-databases
A few other datasets have water properties, but nothing with both capacity + wilting point.

WISE Soil Property Databases
has water capacity only (amount 19 mostly chemical vars). unit is in kpa (?)

SoilGrids
No water properties. Though FAQ say they might do them in the future.

gNATSGO
From NRCS, USA only. Looks like it has water holding capacity

POLARIS
Likely has field capacity, not wilting point. From the readme

theta_s - saturated soil water content, m3/m3
theta_r - residual soil water content, m3/m3

Documentation not the best, essentially an FTP of numerous tifs

Look at site level errors per ecoregion on a MAT/MAP plot to potentially gain some insights.

Maybe look at the seasonality of precip too

Its out in the ocean so was missing soil data.

No need to re-run everything. but re-run the shrubland model without it to make sure it didn't screw things up.

Ecological Modelling

EGU biogeosciences
Journal of Arid Environments
AGU biogeosciences - $1000 + $3800 OA (assuming a regular length paper)

Ecohydrology

MDPI - Land - APC 1400 CHF (~1500 USD) Special issue: Land Change Modelling

Do them again with some new things

• 2 new sites added (ae1b502)
• potentially with a Choler model
• increase maximum iterations since some didn't converge

ie. incorporate the MAP scaling factor from phenograss into a naive model

potential names
NaiveGCC
NaiveCorrected
NaiveMAPCorrected

1) soil water balance

W_{t+1} = W_{t} + P_{t} - (1-V_{t})/(D_{t}/(W_{cap}-{W_p}))^{2}E_{t}-gb_{3}D_{t}V_{t}

2) fCover balance

V_{t+1} = V_{t} + gS_{t}b_{1}D_{t-L}(1-V_{t}/V_{max})-db_{2}(1-V_{t})V_{t}

3) Plant available water

D_{t} = max(0,W_{t}-W_{p})

4,5) min/max constraints on water and veg

W_{t} = max(0,min(W_{cap},W_{t}))

V_{t} = max(0,min(V_{max},V_{t}))

6) plant available water decreasing/increasing

\left\{\begin{matrix}
D_{t-L} > D_{t-L-1}; & d=0\\ 
D_{t-L} \leq  D_{t-L-1}; & d=1 
\end{matrix}\right.

7) daylength constraint

S_{t} = (TOA_{t} - Phmin)/(Phmax - Phmin)

\left\{\begin{matrix}
S_{t-1} > S_{t}; & d=1  \\
S_{t-1} \leq  S_{t}; & d=0 
\end{matrix}\right.

8) temperature response function

g = (\frac{T_{max}-T_{m}}{T_{max}-T_{opt}})(\frac{T_{m}}{T_{opt}})^{(\frac{T_{opt}}{T_{max}-T_{opt}})}