453+16.073 Nhat’s simple scaling factor for derivation of shorter duration, d (h) events intensities, Pd, from NMIA 24-h precipitation depths, P24 (mm) equation(6) Pd=P24d240.178 Nhat’s simple scaling factor for derivation of shorter duration, d (h) events intensities, Pd, from SIA 24-h precipitation depths, P24

(mm) equation(7) Pd=P24d240.152 The ANN formulae used for determining 1, 2, 5 and 10 days durations in Eq. (8) performed credibly. Predictions of the tuned ANN for NMIA and SIA stations are shown in Fig. 6. Lumacaftor manufacturer Output of an ANN for daily precipitation (mm) from a number (n) of re-analysis predictors (x), with weights (W) and constants (C) with time in days (t) in a Sigmoid function. equation(8) Outputt=wk∑i=0n11+e−∑ni=0xi−1⋅wi−ti.wj+c1+c2⋅(Outputt−1+Outputt−2)2 Correlation analysis varied between 0.52 and 0.72 for NMIA and 0.46 and 0.68 for SIA and suggested some skill of the ANN’s 1–10 days predictions. NMIA ANN model predictions ABT 199 were marginally better than SIA’s. Daily precipitation performance was expectably lower with correlations of 0.40 and 0.28 for NMIA and SIA respectively and reinforced that downscaling techniques do better with longer temporal

scale. Daily events are likely to be influenced by orographic factors not captured in the gridded re-analysis predictions. Scatter plot assessment of the ANN AMS predictions versus the observed (see Fig. 6 bottom panels) revealed that the NMIA model performed better than the SIA model for the 10 days durations. The gradient was 1.097 or slight over-prediction versus 0.638 or moderate under-prediction for SIA. Linear model correction of the differences explained most of the biases and the corrected ANN predictions had second a gradient of 0.96–1.0 (near perfect agreement). This approach is consistent with that of Van Roosmalen et al. (2009). The climatology of monthly precipitation was accurately predicted by the ANN for both

stations with a correlation of 0.76 and 0.88 for NMIA and SIA respectively in Fig. 7. Both the observed and predicted climatology are consistent with Taylor et al. (2002), Angeles et al. (2010), and CSGM (2012). Bias averaged 38.0 mm for NMIA and was maximized for October that corresponds to the late wet season. Bias was relatively small and consistent at 3.7 mm for SIA. High correlations and low biases confirm the ANN’s applicability to both AMS analysis and seasonal precipitation analysis (see Fig. 7). AMS predictions from the ANN were derived. NMIA’s predictions were determined to be 40–60% higher than SIA typically and follow a similar trend in the original data of 1957–1991. Gaps in the data set were reduced by both empirical and downscaling methods. NMIA and SIA data sets typically increased from 13% of the maximum number of data set values to 65% for the 5 min to 10 days durations. Both methods can be used to increase AMS for frequency analysis reliably.