Processing Rainfall Data

Published on by in Academic

Processing Rainfall Data

I am working with raw rainfall data, directly from the files retrieved from a tipping bucket rain gauge.

I have two columns, one is the time log for each tip and at that time the read of 0.5mm of rainfall. Since it is not a continuous time series, I cannot identify whether it is a dry period or a missing value. (Refer to the attached file for an example of the retreived data).

I am using R to convert those readings to a continuous time series; of course, so far I am filling with zeros the time where there are no readings. I am not sure if that is correct. The code just sums up the readings between a time step, say for example 5, 10 and 15 minutes. 

I would much appreciate any further readings or reference on processing rainfall data. Any advice is welcome!

Media

Taxonomy

10 Answers

  1. Se ve rara la serie de datos. ¿Cuál es la marca del pluvio? Creo que la serie corresponde a aquellos datos donde hubieron tips (0.5 mm tal vez es la resolución del pluvio lo cual parece bajo pero depende de la zona en la que está instalado). Con todo puedes agregar a una serie de tiempo de 10 min por ejemplo pero necesitas saber a qué agregación el datalogger obtuvo esos 0.5 mm, de lo contrario no es posible.

    1 Comment

    1. Hola! Gracias por tu respuesta!.

      No se cual es la marca del pluviometro. Los datos los recibo en formato crudo, tal y como lo describo. De hecho, es así, el datalogger guarda el tiempo en el que el balancín gira, entonces se asume que el tiempo que transcurre desde un log y el siguiente, es el tiempo en el que se registró una precipitación de 0.5mm. 

      Hasta ahora, estoy confirmando mis dudas. Está bien agregar en diferentes lapsos de tiempo agregando el numero de lecturas registradas en ese lapso. Y por otro lado, no es posible determinar períodos de datos faltantes. 

      Los datos son iguales para 14 estaciones que tiene ETAPA-EP en las 4 cuencas principales. Lastimosamente, al hacer una correlación entre estaciones cercanas, los resultados tampoco son de mucha utilidad. Obtengo correlaciones inferiores a 0.4 en todos los casos.

      Por lo pronto, creo que voy a seguir con lo que estaba haciendo. Agregar en períodos de 1, 5, 10, 15 y 30 minutos  1 hora.  El problema volverá a aparecer cuando intente calcular las IDF. Pensaba hacerlo para cada estación y esperar similitudes.

  5. If its any use, here is the contact of an old friend/colleague who manufacturer rain gauges. I sent him a copy of image you posted of data and suggest you send him a sample file so he can see if its a format he can process. Here is his email... Mike Lysaght .

     

    Cheers

    Charles

    1 Comment

  6. Our product STORMPAC http://sites.wrcplc.co.uk/stormpac/ is good for disaggregating rainfall data to 5 minute once it is in hourly timesteps. We tend to process into hourly by pulling off the timestamps for the PREVIOUS hour (remember a physical recording of the tipping bucket is rainfall up to that point) then disaggregate to 5 minute to run through sewer models. So it's all about the reason for analysis. As to missing data - well data is data that could be rubbish so I would specify the areas of uncertainty in anything you use the data for.

  7. It's good if you could upload the output as well. It may answer your question...

    1 Comment

    1. I will upload the resulting time series after processing the raw data. I believe that the question really is if it is correct to just aggregate readings in between a timestep, for example 5, 15 o 30 minutes. And if I can somehow identify the missing data, given that the gauging station does not follow a continuos time series.

  8. I recognise the format as one used by a data logger for rain gauges made in Australia. Do you know the manufacturer of the rain gauge?.  You can write an excel macro to convert it in to a standard hourly rainfall total using the date and time. However, I will send a message to an old colleague who might have access to the conversion program.

    Cheers

    charles

    1 Comment

    1. Thanks for the answer. I don't really know where the datalogger is from, I just get the raw data and process it for further needs. 
      I started creating an excel spreadsheet that converts it to regular time series, I used "countif" function since all readings are 0.5mm. It became too slow since I am working with 5 to 10 year readings, so I implemented a similar code to R, it works flawlessly, but my doubdt is wheather I am doing things correct or I am making some kind of dumb mistake with missing data or by aggregating the readings. I hope this makes sense and you understand what I mean. 

  9. I think you need to back up one step.  Why are you trying to take readings?  Just for curiosity sake. If your lakes, rivers, reservoirs, are running low are you using too much or is there not enough rainfall? You are entering a classic situation. Do you want to teach or do? What is your pleasure?  Catch rainwater, use, process, reuse again and again.  Collect your waste, process it, make and use compost and compost teas. When soil is black and rich it will hold the moisture twice as long. When you are growing garden food, flowers, or trees they respire. Long term respiration causes weather patterns to change. This increases the rainfall amount.  Take all trash from a country make and use the compost, grow food in a desert  and slowly you will have more and more rain and soon a forest.

  10. One short quick answer is checking with the nearby site (if any) on daily time step. If missing, use regression to fill then disaggregate to smaller time step (e.g.15 min).  

    1 Comment