Dongchen Zhang Phone Number: 617-320-9561 Email: zhangdc@bu.edu
Breanna van Loenen bvanloen@bu.edu 3474176834
Tessa Keeney Email: tkeeney@bu.edu Cell: 360-854-8174
Katherine Losada Email: klosada@bu.edu Phone number: (603)707-6791
-
Data --
a. Data_Download_Functions: includes scripts that define functions for pulling MODIS data
b. data_dowload_code: includes scripts that pull data from various sources (NOAA, EFI, NEON); also includes a folder with .Rdata files of ensemble parameters
-
Forecast --
a. calibration_code: includes scripts that have calibration code for dynamic linear model (DLM) and random walk
b. forecast_code: includes scripts that define ensemble forecast, run data assimilation (via Kalman filter), and submit forecast to NEON website
Authors: Dongchen, Tessa, Breanna
#Data Model In the ef.out model, our data model relates observations (y and covariates) to the latent states of 1) NEE observations with prescribed priors on the observation error and 2) covariates (temperature, precipitation, and relative humidity) and intercept with observation errors estimated by fixed effects.
#Process Model The process model relates the latent state of NEE to the prior of NEE, intercept, and covariates with prescribed process error. We have chosen to model Gaussian process error, although future iterations of our model will attempt to apply Laplace distribution to account for the non-normal behavior of the NEE variable.
#Priors The priors consist of the observation error and the process error, as well as the initial conditions of the model. We chose gamma distributions for our priors since they are conjugates of the normally distributed mu, but will likely alter this in the model in future iterations to reflect the Laplace distribution of NEE. Fixed effects within the priors include the means and precisions associated with the covariates (temperature, precipitation, and relative humidity).