Search

An improved representation of the spatio-temporal patterns of climatological variables is crucial for ecological, agricultural, and hydrological applications and can improve the decision-making process. Traditionally, precipitation (P) and actual evaporation (ETa) are estimated using ground-based measurements from meteorological stations. However, the estimation of spatial patterns derived solely from point-based measurements is subject to large uncertainties, particularly in data-scarce regions as the Nile Basin, which has an area of about 3 million km2. This study evaluates six state-of-the-art P products (CHIRPSv2, CMORPHv1, CRU TS4.02, MSWEPv2.2, PERSIANN-CDR and GPCCv2018) and five ETa products (SSEBop, MOD16-ET, WaPOR, GLEAM and GLDAS) over the Nile Basin to identify the best-performing products. The P products were evaluated at monthly and annual temporal scales (from 1983 onwards) through a point-to-pixel approach using the modified Kling-Gupta Efficiency and its components (linear correlation, bias, and variability ratio) as continuous performance indices. The ETa products were evaluated through the water balance approach (due to the lack of ground-based ETa measurements) for 2009-2018 at the multiannual scale. Because streamflow data were not available for this period, an empirical model based on the Random Forest machine learning technique was used to estimate streamflow at 21 catchments at the monthly scale. For this purpose, we used streamflow data from 1983 to 2005 as the dependent variable, while CHIRPSv2 precipitation and ERA5 potential evaporation and temperature data were used as predictors. For the catchments where the model performed well over the validation period, streamflow estimates were generated and used for the water balance analysis. Our results show that CHIRPSv2 was the best performing P product at monthly and annual scale when compared with ground-based measurements, while WaPOR was the best-performing ETa product in the water balance evaluation. This study demonstrates how remote sensing data can be evaluated over extremely data-scarce scenarios to estimate the magnitude of key meteorological variables, yet also highlights the importance of improving data availability so that the characterisation of these variables can be further evaluated and improved.

The Intensity-Duration-Frequency (IDF) curves are crucial for urban drainage design and to mitigate the impact of extreme precipitation events and floods. However, many regions lack a high-density network of rain gauges to adequately characterise the spatial distribution of precipitation events. In this work we compute IDF curves for the South-Central Chilean region (26-56°S) using the latest version of the Integrated Multi-satellitE Retrievals for GPM (IMERGv06B) for 2001-2018, with a spatial resolution of 0.10° and half-hourly temporal frequency.First, we evaluated the performance of IMERGv06B against 344 rain gauge stations at daily, monthly and annual temporal scales using a point-to-pixel approach. The modified Kling-Gupta efficiency (KGE’) and its components (linear correlation, bias, and variability ratio) were selected as continuous indices of performance. Secondly, we fit maximum precipitation intensities from 14 long-term rain gauge stations to three probability density functions (Gumbel, Log-Pearson Type III, and GEV II) to evaluate: i) the impact of using 15-year rainfall time series in the computation of IDF curves instead of using the typical long-term periods (~ 30 years); and ii) to select the best distribution function for the study area. The Gumbel distribution was selected to obtain the maximum annual intensities for each grid-cell within the study area for 12 durations (0.5, 1, 2, 4, 6, 8, 10, 12, 18, 24, 48, and 72 h) and 6 return periods (T=2, 5, 10, 25, 50, and 100 years).The application of the Wilcoxon Mann-Whitney test indicates that differences between IDF curves obtained from 15 years of records at the 14 long-term rain gauges and those derived from 25 years of record (or more) are not statistically significant, and therefore, 15 years of record are enough (although not optimal) to compute the IDF curves. Also, our results show that IMERGv06B is able to represent the spatial distribution of precipitation at daily, monthly and annual temporal scales over the study area. Moreover, the obtained precipitation intensities showed high spatial variability, mainly over the Near North (26.0-32.2°S) and the Far South (43.7-56.0°S). Additionally, the intensities from Central Chile (32.2-36.4°S) to the Near South (36.4-43.7°S) were systematically higher compared to the intensities described in older official governmental reports, suggesting an increase in precipitation intensities during recent decades.

Climate models have not achieved a consensus about the future trend of long-term average of precipitation. As well as, the future trend of extreme values (including both extreme, droughts and heavy events) has higher uncertainties, because are unusual events. The Colombian territory is permanently in risk due to precipitation climatic extremes: during El Niño years, the rain amounts are severely reduced, consequently the rivers flow and the water resource availability; nevertheless, during La Niña years, floods and landslides events are common, because the rain is excessive.The precipitation extremes are affected due to long-term trends and the inter-annual variability represented by El Niño/La Niña cycle, then conduct this study is relevant. The selected study area is the Colombian territory. A Satellite Rainfall Estimate (SRE) was used to ensure a whole spatial coverage. The SRE has a daily temporary resolution, then it is suitable for building the selected Extreme Precipitation Indices (EPI). Statistical tests were carried out to verify the long-term change of EPI. The hydrological years were discriminated according to the ENSO, in order to perform a statistical test to probe the hypothesis that EPI, during these particular years (El Niño/La Niña), belong to probability distributions different from that distribution of EPI in “normal” years.Mean annual precipitation in the Andean region drops in El Niño years, and it increases in La Niña years. In the Colombian Pacific basin, the number of wet days is reduced by the long-term trend, but the variable is not affected by the ENSO phenomena. However, in the Andean region and the eastern plains, El Niño has a high effect on reducing the number of wet days. Finally, extreme events are affected by both the long-term trend and the ENSO phenomena too; however, the change spatial distribution reveals a high impact on the Andean region.