Forecast Step
#background trajectory [without correction]
ub[:,k+1] = RK4(Lorenz63,ub[:,k],dt,sigma,beta,rho)
#EnKF trajectory [with correction at observation times]
for i in range(N): # forecast ensemble
uai[:,i] = RK4(Lorenz63,uai[:,i],dt,sigma,beta,rho) \
+ np.random.multivariate_normal(np.zeros(n), Q)