Dear all,
The overall goal is to disaggregate RCM output from daily resolution to 15 min resolution for different stations. For the stations, I have time series of about 10 years with 15 min resolutions for "calibration". Any suggestions on suitable approaches ...? I would be happy if the code (matlab, python) can be distributed. In the optimal case, several methods will be tested and compared w.r.t. their performance to represent 15 min resolution.
Cheers, Patrick
Hi Patrick Laux
You could try the Method of Fragments. I used it for disaggregation from daily to hourly resolution and compared it to dynamical downscaling with a CPM:
Article Comparison and Evaluation of Statistical Rainfall Disaggrega...
The available station data was up to 17 years in that example - so it could possibly work for your case.
Cheers, Ben
Dear Benjamin,
thank you very much for the hint about the MoF. This is exactly what I need! The paper also gives useful insights into suitable validation measures.
Cheers, Patrick
Python code will be:
This code uses MBLRP model to receive data from csv file with 2 columns date and rainfall and then disaggregates it on hourly basis
## Start here
import numpy as np
import pandas as pd
# Read daily rainfall data from the csv file
data = pd.read_csv('daily_rainfall_data.csv')
# Convert date column to datetime format
data['date'] = pd.to_datetime(data['date'])
# Define date column as index
data = data.set_index('date')
# Define the MBLRP model parameters
a = 0.7
b = 0.3
lambda_1 = 0.25
lambda_2 = 0.01
# Disaggregate daily rainfall data into hourly rainfall data
hourly_rainfall_data = []
for i in range(len(data)): ## loop through
daily_rainfall = data.iloc[i]['rainfall']
n = int(np.round(np.random.poisson(lambda_1)))
u = np.random.uniform(size=n)
v = np.random.uniform(size=n)
t = np.random.exponential(scale=1/lambda_2, size=n)
t = np.cumsum(t)
for j in range(n):
if j == 0:
p = a
q = b
else:
p = a + (1 - a) * np.exp(-lambda_2 * (t[j] - t[j-1]))
q = b + (1 - b) * np.exp(-lambda_2 * (t[j] - t[j-1]))
if u[j]
Hi,
here are two models in R and Matlab, both capable to work for sub-hourly scale
Preprint Simulating sub-hourly rainfall data for current and future p...
Best,
Ivan
Hi, Patrick Laux , I am also working in a similar field. What method have you used for the disaggregation of precipitation?... Can you share the code here or send me in my inbox.
Regards
Tapas
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps