The analyses resulted in satisfactory estimation of E(LT) as is evidenced by the value of COE equal to 75.38% and the mean error equal to −1.15% ( Fig. 5A). Likewise, the values of E(MT) were also predicted using Eqs. (5) and (6) but results were less satisfactory. The E(MT) for rivers exhibiting affinity up to AR-2 dependence structure tended to be

over-predicted while those rivers exhibiting affinity beyond AR-2 dependence structure ICG-001 cost tended to be under predicted. Therefore, the E(LT) was computed using the first order Markov chain model (Eq. (8)) for rivers exhibiting affinity to AR-2 process and by a random or the Markov chain-0 model for rivers in resonance with

AR-1 process. For all other rivers exhibiting dependence structure beyond the second order, the E(LT) was computed based on the second order Markov chain model. It is to be noted that E(LT) can be computed based on a random or the Markov chain-0 model of drought lengths from the expression E(LT) = −[logT(1 − q)/log(q)]. find more The aforesaid expression essentially is Eq. (8) in which qq equals q and also qp equals q. The computations for the drought intensity E(I) remained unchanged as it was unaffected either by the first or the second order probabilities. Using the aforesaid modification, the predicted E(MT) corresponded satisfactorily with the observed counterparts ( Fig. 5B, COE ≈ 86%; mean error ≈ −1%). Succinctly, the computations of E(LT) for estimating E(MT) are based generally on one order less than the best fitting order of the Markov chain model for drought length. That is,

if the drought length is predicted using the Markov chain-2 model, then the corresponding magnitude should be predicted using the drought lengths obtained from the Markov chain-1 model. Likewise, if the lengths are best predicted by the Markov chain-1 model, the magnitude should be based on the drought lengths Alanine-glyoxylate transaminase computed from the random model or the Markov chain-0 model. The hydrologic drought durations and magnitudes at truncation level corresponding to the median flow may not be tangible, although such estimates of drought have relevance to design applications of water resources systems such as reservoirs for water storage to ameliorate droughts. However, hydrologic droughts become tangible at low levels of truncation such as Q90, Q95 etc. on daily or weekly flow series. The first order Markov chain model (Markov chain-1, Eq. (8)) was found satisfactory to predict E(LT) at the uniform truncation levels of Q90 and Q95, which is also evident from the plot ( Fig. 6A with COE ≈ 72% and mean error equal to 0.2%). The drought magnitude can be computed using the relationship E(MT) = α × I × E(LT), where α is a scaling factor for standard deviations.