dataset.varDataset

	Creates an instance of varDataset object based on requested files & variable.
    
	Parameters
	----------
	varlabel : str
		Label for object, consistent with namelist use_vars dictionary.
	path : str
		Directory containing files for concatenation.
	varname : str
		Name of variable in files.

	kwargs : dict
		see Other Parameters
        
	Other Parameters
	----------------
	level : float
		If file contains data for multiple levels
	climoyears : tuple
		(start year, end year) to slice data
	latbounds : tuple
		(south lat, north lat) to slice data
	lonbounds : tuple)
		(west lon, east lon) to slice data
	time_window : int
		Used for running_mean, number of time units.
	time_sample : str
		Integer + string, indicating number of time sample which can be 
		'D' = Day, 'W' = Week, 'M' = Month, 'Y' = Year.
		Example: '3D' would be a time sample of one ob every 3 days.
	landmask : bool
		If True, only use values over landmass.    

	Returns
	-------
	Dataset : object
		An instance of Dataset.

dataset.varDataset.subset

	Creates a new instance of varDataset object subset by dates.

	Parameters
	----------
	datebounds : tuple
		Start and end dates of season, in form "m/d".
	season0 : bool
		If True, starts data at beginning of season. If False, starts data at beginning of first year.
		Only matters for seasons that cross Jan 1st.

	Returns
	-------
	newobj : object
		A new instance of varDataset.

dataset.varDataset.plot_map

	Takes space vector or time, and plots a map of the data.

	Parameters
	----------
	z : ndarray / list
		Vector with same length as number of space points in self.anomaly
	time : datetime object
		Optional to plot self.anomaly data from specific time in self.time.
	ax : axes instance
		Can pass in own axes instance to plot in existing axes.

	kwargs : dict
		see Other Parameters

	Other Parameters
	----------------
	prop : dict
		Options for plotting, keywords and defaults include
		* 'cmap' - None
		* 'levels' - negative abs data max to abs data max
		* 'figsize' - (10,6)
		* 'dpi' - 150
		* 'cbarticks' - every other level, determined in get_cmap_levels
		* 'cbarticklabels' - same as cbarticks
		* 'cbar_label' - None. Optional string, or True for self.attrs['units']
		* 'extend' - 'both'
		* 'interpolate' - None. If float value, interpolates to that lat/lon grid resolution.
		* 'drawcountries' - False
		* 'drawstates' - False

	Returns
	-------
	ax : axes instance

dataset.eofDataset

	Creates an instance of eofDataset object based on requested files & variables.

	Parameters
	----------
		varobjs : list
			List of variable objects
		Max_eofs : int
	Maximum number of modes to save to object. Default is 100. (This is different than the truncation in the LIM)
   
	Returns
	-------
	Dataset : object
		An instance of eofDataset.

	Saved attributes
	-------------
	Accessible using a period after the object, followed by the name.
	Can also view what attributes are available in any object by typing objname.__dict__.keys()

		* eof, pc, totalVar, varExplByEOF

dataset.eofDataset.reconstruct

	Method for reconstructing spatial vector from PCs and EOFs

	Parameters
	----------
	pcs : list or tuple or ndarray
		list of floating numbers corresponding to leading PCs
	order : int
		1 = forecast PCs, 2 = error covariance matrix
	num_eofs : int
		truncation for reconstruction. Default is to retain all EOFs available.
	pc_wt : dict
		keys corresponding to PC number. Values corresponding to weight        
	
	Returns
	-------
	recon : dict
		dictionary with keys = variable label, and values = reconstructed spatial vector.

dataset.eofDataset.plot

	Map EOFs and plot PC timeseries in subplots.
	Include percent variance explained.

	Parameters
	----------
	num_eofs : int
		number of EOFs to plot. Default is 4
	return_figs : bool
		Return list of the figures. Default is False
	
	Returns
	-------
	List of figures

model.Model

	Parameters
	----------
	tau0_data : ndarray
		Data for calibrating the LIM.  Expects
		a 2D MxN matrix where M (rows) represent the sampling dimension and
		N(columns) represents the feature dimension (e.g. spatial grid
		points).
	tau1_data : ndarray, optional
		Data with lag of tau=1.  Used to calculate the mapping term, G1,
		going from tau0 to tau1.  Must be the same shape as tau0_data.  If
		not provided, tau0_data is assumed to be sequential and
		nelem_in_tau1 and tau0_data is used to calculate lag covariance.
	tau1n : int, optional
		Number of time samples that span tau=1.  E.g. for daily data when
		a forecast tau is equivalent to 1 week, tau1n should be 7.
		Used if tau1_data is not provided.

	Saved attributes
	-------------
	Accessible using a period after the object, followed by the name.
	Can also view what attributes are available in any object by typing objname.__dict__.keys()

		* tau1n, X0, X1, C0, Ctau, G1, Geigs, L, Leigs, Q

model.Model.forecast

	Forecast on provided data.

	Performs LIM forecast over the times specified by the fcst_leads. 
	Forecast can be performed by calculating G for each time 
	period or by L for a 1-tau lag and then calculating each fcst_lead G 
	from that L matrix.

	Parameters
	----------
	t0_data : ndarray
		Data to forecast from.  
		Expects a 1D or 2D array where M (rows) represent the 
		initialization time dimension
		and N(columns) represents the feature dimension.
	lead_time : List (int)
		A list of forecast lead times.  Each value is interpreted as a
		tau value, for which the forecast matrix is determined as G_1^tau.
	
	Returns
	-----
	fcst_out : ndarray-like
		LIM forecasts in a list of arrays, where the first dimension 
		corresponds to each lead time. Second dimension corresponds 
		to initialization times. Third dimension is PCs

driver.Driver.get_variables

	Load data into variable objects and save to pickle files.

	Parameters
	----------

	read_path : str
		If None, compiles data from paths specified in variable namelist file.
	save_path : str
		If None, does not save variables.
	save_netcdf_path : str
		If a str, saves netcdf files of variable with running-mean anomalies and climatology to specified directory.
		Default is None, which means netcdf files will not be written.
	segmentby : str
		If save_netcdf_path is not None, segmentby specifies how to chunk the netcdf files. Options are "all" or "year". Default is "all".

driver.Driver.get_eofs

	Load data into EOF objects and save to pickle files.

	Parameters
	----------
	read : bool
		If True, reads previously saved eof objects, else compiles data from path.
	save_netcdf_path : str
		If a str, saves netcdf files of EOF object with regridded EOF maps, PCs, and variance explained.
		Default is None, which means netcdf files will not be written.

driver.Driver.cross_validation

	Cross validation.

	Parameters
	----------
	num_folds : int
		Number of times data is subset and model is trained. Default is 10,
		Which means each iteration trains on 90% of data, forecasts run on 10% of the data.
	lead_times : list, tuple, ndarray
		Lead times (in days) to integrate model for output. Default is first 28 lead times.
	average : bool
		Whether to average the forecasts if multiple lead times are provided.
	save_netcdf_path : str
		Path to save netcdf files containing forecast and spread.
	segmentby : str
		If save_netcdf_path is not None, segmentby specifies how to chunk the netcdf files. 
		Options are "day", "month", "year". Default is "day".
	
	Returns
	-------
	model_F : dict
		Dictionary of model forecast output, with keys corresponding to valid time
	model_E : dict
		Dictionary of model error output, with keys corresponding to valid time

driver.Driver.pc_to_grid

	Convect PC state space vector to variable grid space.

	Parameters
	----------
	F : ndarray
		Array with one or two dimensions. LAST axis must be the PC vector.
	E : ndarray
		Array with one or two dimensions. Error covariance matrix. If None, ignores.
		LAST TWO axes must be length of PC vector.
	limkey
		Name of LIM / EOF truncation dictionary. Default is first one in namelist.
	regrid : bool
		Whether to regrid to lat/lon gridded map.
	varname : str
		If just one variable output is desired. 
		Default is None, in which case all variables are output in a dictionary.
	
	Returns
	-------
	Fmap : dict
		If F was provided. Dictionary with keys as variable names, and values are ndarrays of 
		reconstructed gridded space.
	Emap : dict
		If E was provided. Dictionary with keys as variable names, and values are ndarrays of 
		reconstructed gridded space.

driver.Driver.get_rhoinf_time

	Get timeseries of rho infinity.

	Parameters
	----------
	varname : str
		Name of variable of interest.
	fcst : ndarray
		Forecast array with shape (time, space) or (space). Default is None, using the self.model_F output.
	spread : ndarray
		Spread array with shape (time, space) or (space). Default is None, using the self.model_E output.
	latbounds : tuple
		latitude bounds to specify spatial subdomain. Default is None, using full domain.
	lonbounds : tuple
		longitude bounds to specify spatial subdomain. Default is None, using full domain.
	region : str
		Using regionmask, can specify country to define the subdomain. Default is None.
	
	Returns
	-------
	rhoinf : ndarray or float

driver.Driver.get_model

	Get the LIM model object.

	Parameters
	----------
	eof_trunc : dict
		Dictionary containing eofs as keys and truncations as values
	tau1n : int
		Training lag
	datebounds : tuple
		Strings (m/d) defining start and end of training period. Default is (‘1/1’,’12/31’)
	load_file : str
		Filename for pickle file containing model to load.
		Default is None, in which case a new model is trained.
	save_file : str
		Filename of pickle file to save new model to.
		Default is None, in which case the model is not saved.
	save_to_netcdf : str
		Filename to save netcdf containing model attributes. 
		Default is None, which does not write a file.

	Returns
	-------
	model : model object
		Object of trained model

driver.Driver.prep_realtime_data

	Compile realtime data, interpolate to same grid as LIM, and convert into PCs.
	If it has already been run before within the defined self object, 
	self.RT_VARS will be the same, and self.RT_PC can be updated for different LIMs with same vars,
	without having to process variables again.

	Parameters
	----------
	limkey
		Name of LIM / EOF truncation dictionary. Default is first one in namelist.
		Saved to self.RTLIMKEY

	Returns
	-------
	self.RT_VARS : dict
		dictionary with keys = variable names + "time", and values = data,
		processed to same time res and space res & domain as training variables.
	self.RT_PC : dict
		dictionary with keys = eof names, and values = truncated PCs, both according to specified limkey

	
	>>> Example use: Can zero out certain PCs in initial conditions by 
		self.RT_PC[eofname] *= 0

driver.Driver.run_forecast

	Run forecasts initialized with self.RT_PC, with EOFs & model specified by self.RTLIMKEY

	Parameters
	----------
	t_init : datetime object
		Date of initialization for the forecast.
		Default is maximum date available in self.RT_VARS['time']
	lead_times : list, tuple, or ndarray
		Lead_times in increment of data to integrate model forward. 
		Default is first 28 leads.
	save_netcdf_path : str
		Path to save netcdf files containing forecast and spread. 
		Default is None, which will not write netcdf files.

	Returns
	-------
	model_F : dict
		Dictionary of model forecast output, with keys corresponding to init time.
	model_E : dict
		Dictionary of model error covariance output, with keys corresponding to init time.

driver.Driver.plot_teleconnection

	Plots teleconnection timeseries analysis, forecast, and spread.

	Parameters
	----------
	list_of_teleconnections : list
		List of names (str) of teleconnections to plot.
	daysback : int
		Number of days prior to t_init to plot analysis.
	save_to_file : str
		Name of file to save figure to. Default is None, which does not save the figure

	Other Parameters
	----------------
	prop : dict
		Customization properties for plotting

driver.Driver.plot_timelon

	Plots meridionally averaged data with longitude on x-axis and time on y-axis

	Parameters
	----------
	varname : str
		Name of variable (consistent with use_var dictionary keys) to plot.
	lat_bounds : tuple or list
		Latitude boundaries to average data between.
	t_init : datetime object
		Time of forecast initialization. Default is maximum time in self.model_F.
	daysback : int
		Number of days prior to t_init to plot hovmoller
	save_to_file : str
		Name of file to save figure to. Default is None, which does not save the figure

driver.Driver.verif

	Plots verifying anomalies, categories, hit/miss maps, and outputs HSS and RPSS skill scores.

	Parameters
	----------
	varname : str
		Must be a variable name in the list of keys used in self.RT_VARS.
	t_init : datetime object
		Time of forecast initialization. 
		Default is maximum time in self.RT_VARS['time'] minus maximum lead time.
	lead_times : ndarray / list / tuple
		Lead times. Forecasts and obs WILL be averaged for given leads.
	Fmap : ndarray
		Default is None, in which case forecast will be taken from self.model_F[t_init]
	Emap : ndarray
		Default is None, in which case spread will be taken from self.model_F[t_init]
	prob_thresh : int
		Probability threshold (50 to 100) above which verification statistics will be calculated and mapped.
		Default is 50 (no masking).
	regMask : str
	Name of region / country to use for masking maps and skill score calculation. Default is None.
	Options listed in regionmask.defined_regions.natural_earth.countries_110

	Other Parameters
	----------------
	prop : dict
		Customization properties for plotting