El proyecto PIJ 18-05 ha concluido el 31 de marzo del 2021. El proyecto tal como se esperaba ha tenido grandes resultados en materia de predicción climática, sentar las bases a futuro y experimentar formas de modelación de las fases de la precipitación.
Los productos del proyecto son:
-Tres artículos científicos.
-Una tesis de grado.
-Una librería de R.
Artículos:
Título: Corrections of raindrop size distribution measured by Parsivel OTT2 disdrometer under windy conditions in Antizana massif, Ecuador.
Resumen: Monitoring precipitation has become a challenging task in sites with strong variations of temperature (Ta) and wind speed (WS). To overcome tipping gauges issues, disdrometers constitute a practical alternative that provides information about drop size distributions (DSD), precipitation type, and amount. Thus, a Parsivel OTT2 disdrometer was installed at 4730 m a.s.l. in the glacier foreland of the Antizana volcano in Ecuador, close to 0°C isotherm. to study precipitation-type. To correct DSDs we filtered spurious particles, and then shift fall velocities such the mean value match with the diameter-fall velocity relationship of rain, snow, graupel, and hail. A clustering approach was employed to partition solid (Ta≤-1°C) and liquid records (Ta≥3°C) into low, medium, and high WS categories. Filters affected both fast and slow particles with D≤1 mm. Also, changes in the snowflakes/raindrops proportion were linked to WS increase due to the increase of frequency of collision and breakup of larger particles. Constant density (1g cm-3) gives reliable results for liquid precipitation, however over estimates solid precipitation. Therefore, density datasets were proposed to correct precipitation yielding 0.76-1.1 times the values measured in disdrometer, and 1-1.43 times the values of nearby rain gauge. These results improve the knowledge of precipitation microphysics of mountainous tropical zone.
Título: Assessing the contribution of glacier melt to discharge in the tropics: the case of study of the Antisana Glacier 12 in Ecuador
Resumen: Tropical glaciers are excellent indicators of climate due to their fast response to temperature and precipitation variations, and in addition play an important role for supplying freshwater. In this study, a hydro-glaciological model was adapted to analyze the influence of meteorological forcing on the melting and discharge variations at Glacier 12 of Antisana volcano (4735-5710 masl, 1.68 Km2, 0⁰29’S; 78°9’W). Energy fluxes and melting was calculated using a distributed surface energy balance (SEBD) using 20 altitude bands from glacier snout to the summit at 30-minute resolution for 684 days between 2011 and 2013. The discharge was computed using linear reservoirs for snow, firn, ice, and moraine. Meteorological variables were recorded at 4750 masl in the ablation area and then distributed through altitudinal and geometrical corrections. The specific mass balance (-0.46 meters of water equivalent -m w.e.-) and the ablation gradient (-2.4 m w.e. (100m)-1) agree with the estimated values for the zone. Sequential validations allowed the simulated discharge to reproduce hourly and daily variability of the discharge at the outlet of the catchment, and was verified the runoff measured did not reflect the potential discharge (0.14 m3 s-1) due to possible losses through the complex geology of the site. Shortwave radiation was the main flux for melting energy at ablation zone, and thus controls the discharge variations. In consequence, the cloud cover and the albedo via snowfall thickness, which module the shortwave components, became the real drivers of melting and discharge during the months with moderate wind. Whereas, the wind speed was the most influencing variable during July-September season. These new insights are crucial in Inner Tropics since cloudiness and precipitation occurs throughout the year yielding a constant shortwave attenuation and continuous variation of snow thickness.
Título: Deriving precipitation phase with artificial intelligence methods in the Andes: unveiling meteorological drivers and comparison with classical methods
Resumen: The precipitation phase has a fundamental role in the hydrologic cycle and energy fluxes between land and the atmosphere, that in turn affect the climatic system. The continuous trends towards a decrease ratio of snow to rain due to climatic change, indicates that the stream flow timing and duration will be affected. Therefore, a better knowledge about the precipitation phase composition, main meteorological drivers and prediction may provide help to take more informed decisions for a better water resource management, especially in cities dependent of water coming from glaciers and constant urbanization and land use change. The classical logistic models rely only on air-temperature and relative humidity to predict PP. However, the processes related to PP are far complex. In the present study data, a data mining approach, Random Forest, RF, to unveil the meteorological variables driving PP. Also, classical logistic, Random Forest, and Artificial Neural Networks models were compared for PP prediction. Interestingly, specific humidity and dew point temperature were the most important drivers for MDA index, followed by outgoing radiation long wave, wind speed and the temperature of the hydrometeor. Further research is necessary in the implementation of simple logistic models using these variables as inputs. With respect to the evaluation of the ability of the classical logistic, ANN, and RF, models, using several performance statistical metrics of amount and occurrence, the results show that the artificial intelligence techniques clearly have a better performance, with the RF models showing the ability to capture the variability of solid, mixed, and liquid phase better. It’s an exercise for the coupling of black box models derived information to gain more knowledge about complex processes and the potential to generate conceptual models at the end.
Tesis de grado:
Título: Assessment of the impact of the assimilation of observations on short-term weather forecasts in the páramo del antisana.
Resumen: The Antisana glacier is located in the northeastern Ecuadorian Andes, near the Amazonian slope. This area has been the subject of many studies since it provides drinking water for the city of Quito, the capital of Ecuador. The topographic complexity of this zone produces the interaction of several atmospheric phenomena at various spatio-temporal scales, which is still poorly understood. Additionally, the scarcity of stations in the area to evaluate the atmospheric models, the lack of knowledge of the optimal subgrid parameterizations, and the sensitivity of the models to initial conditions, increase the difficulty of weather forecasting. To improve the initial conditions of regional weather models such as the Weather Research and Forecasting Model (WRF), data assimilation is used, either from synoptic, radiosonde, or satellite data. In this study, WRF is evaluated with three assimilation scenarios, i) radiosondes (TEMP assimilation), ii) weather stations (SYNOP assimilation), and iii) the combination of TEMP and SYNOP (S+T assimilation). These assimilation cases were compared with forecasts without assimilation and information from meteorological stations. The forecasts were evaluated from 1 to 3 days at two stations in the glacier area: the PRAA Morrena station and the EORE Glacioclim station, for two time periods from 18 to 21 November 2019 and from 24 to 27 January 2020.
The results were evaluated through three different methods: time series, Tatlor diagrams and evaluation metrics (Nash-Sutcliffe coefficient (NSE), root mean square error (RMSE), and mean absolute error (MAE)). The evaluations indicate that, in November, the forecasts are not accurate and do not show conclusive information. The forecast of the variables is mostly irregular and has low representativeness of the meteorological conditions in the area, this can be observed through metrics the NSE coefficient that has negative values, and the discordance in the time series. Through a simple synoptic analysis, it was determined that these faults are produced because synoptic characteristics in November are more local or mesoscale, with winds from the southeast towards the mountain range. This behavior causes greater difficulties in relation to larger scales circulation states, as is the case in January as shown in the analysis.
In January forecasts, it is observed that the assimilation of soundings and stations (S+T assimilation) has a positive impact on the temperature and relative humidity forecast, improving the representativeness of the forecasts by reducing the mean and quadratic errors and increasing the value of the NSE coefficient in relation to the SIN A. forecast. In the specific case of precipitation, time series show that TEMP assimilation manages to significantly reduce the underestimation of the base forecast in terms of magnitude, but still shows drawbacks in the forecast of short-term precipitation events, since in some cases they present time lags of more than 12 hours. In general, it was observed that in January the forecasts are better than in November, especially those of temperature and liquid precipitation, this is associated with the fact that winds have more synoptic characteristics coming from the northeast presenting the structure of a low level jet, and with less local influences than those shown in November.
Librería de R
Título: DisdroAndes.
Resumen: El uso del disdrómetro ubicado en el Glaciar 12 del Antisana, específicamente en la estación Morrena. Este equipo aloja con el paso del tiempo información que es esencial en futuros análisis de la fase de precipitación entre otras variables. Por lo tanto, para poder extraer esta información es necesario establecer procesos que faciliten el acceso del investigador. La librería DisdroAndes, establece y automatiza procesos como:
-Carga de datos en formato plano.
-Unión de periodos y eliminación de duplicados.
-Tratamiento de datos, mediante el uso de filtros sustentados en referencias bibliográficas.
-Asociación e identificación de la fase de precipitación, mediante el Código Synop.
-Creación de curvas S de acuerdo a la variable elegida.
-Expansión, eliminación de partículas fuera de rango y corrección de matriz DSD en 8 hidrometeoro.
-Cálculo de las propiedades Bulk para 8 hidrometeoros.
Una vez esté disponible en el CRAN de R se puede instalar la librería con la siguiente línea de código “install.packages("DisdroAndes")”.