2 Application of a Machine Learning Technique for Developing Short-Term Flood …
29
Fig. 2.11 Comparison of nearly independent peak flow maxima for the Yanuncay catchment
residuals are represented by the horizontal and vertical differences between each
point and the bisector line. The dependence of the standard deviation on the flow
magnitude was disrupted (constant standard deviation) with a λ-value of 0.25. Results
confirm the observed in Figs. 2.8 and 2.9, where higher scatters and biases were found
for longer forecast horizons.
The gross simplification of precipitation-runoff forecasting models (they do not
include crucial discharge driving-forces) explain the difficulty to accurately forecast
extreme high flows. In addition, for paramo ecosystems, it is well-known that soils
govern flow processes, and therefore, lack of direct measurements limits the forecasting of extreme flows. Apart from missing data, extreme spatial and temporal variability of precipitation in mountainous regions, especially in the Andes of Ecuador,
is hardly collected by a few rain gauge stations within the catchment. Precipitation
events, might occur in regions not covered by rain gauges, and thus, the necessity to
use spatial rainfall estimations (remote sensing imagery) arises.
2.5.2 Evaluation of Drought Forecasts
Contrary to the strong differences found between forecasts and observations for
peak flows (Figs. 2.8 and 2.9), overall results for extreme low flows, for all forecast
horizons and for both catchments, show good fits varying between underestimations
and overestimations (see Figs. 2.12 and 2.13). For the Tomebamba catchment, we
29
Fig. 2.11 Comparison of nearly independent peak flow maxima for the Yanuncay catchment
residuals are represented by the horizontal and vertical differences between each
point and the bisector line. The dependence of the standard deviation on the flow
magnitude was disrupted (constant standard deviation) with a λ-value of 0.25. Results
confirm the observed in Figs. 2.8 and 2.9, where higher scatters and biases were found
for longer forecast horizons.
The gross simplification of precipitation-runoff forecasting models (they do not
include crucial discharge driving-forces) explain the difficulty to accurately forecast
extreme high flows. In addition, for paramo ecosystems, it is well-known that soils
govern flow processes, and therefore, lack of direct measurements limits the forecasting of extreme flows. Apart from missing data, extreme spatial and temporal variability of precipitation in mountainous regions, especially in the Andes of Ecuador,
is hardly collected by a few rain gauge stations within the catchment. Precipitation
events, might occur in regions not covered by rain gauges, and thus, the necessity to
use spatial rainfall estimations (remote sensing imagery) arises.
2.5.2 Evaluation of Drought Forecasts
Contrary to the strong differences found between forecasts and observations for
peak flows (Figs. 2.8 and 2.9), overall results for extreme low flows, for all forecast
horizons and for both catchments, show good fits varying between underestimations
and overestimations (see Figs. 2.12 and 2.13). For the Tomebamba catchment, we