Graph as HMM

[Rafnuss]

Modification

• [x] Seperation of the grl$p in T and O
• [x] Return edge marginal probability in graph_marginal() too computational expensive and not useful
• [x] Implement forward ... backward sampling for graph_simulation() work as well in forward.

[Rafnuss]

graph structure:

• s: source node (index in the 3d grid lat-lon-sta),

• t: target node (index in the 3d grid lat-lon-sta),

• gs: average ground speed required to make that transition (km/h) as complex number

• ws: windspeed (added with graph_add_wind()

• as: windspeed (added with graph_add_wind() can be computed from gs and ws, no need to store

• ps: static probability of each target node,

• obs: Observation likelihood store as matrix

• p: probability combining static and transition.

• stap

• stap_mode

• flight_duration: list of flight duration to next stap in hours,

• flight:

• sz: size of the 3d grid lat-lon-stap,

• lat: list of the likelihood$map latitude in cell center,

• lon: list of the likelihood$map longitude in cell center,

• equipment: node(s) of the first stap (index in the 3d grid lat-lon-stap),

• retrieval: node(s) of the last stap (index in the 3d grid lat-lon-stap),

• extent: Geographical extent as c(xmin, xmax, ymin, ymax)

• resolution: resolution of likelihood$map,

• temporal_extent: start and end date time retrieved from likelihood.

• mask_water: