You are on page 1of 46

Historical Data Analysis

General Information Needed


1. Station data
2. Station history info: obs times, changes, location, moves
3. Topographic data

Analysis of Precipitation Analysis of Temperature Analysis of Evaporation


Information needed Information needed Information needed
Non Mountainous Mountains Non Mountainous Mountains Non Mountainous Mountains
1 isohyetal map - area vs elev. 1 evaporation maps - evap. vs elev.
-basin boundary 2 station weights -basin boundary curve 2 station weights curve
3 mean monthly evap.

PXPP MAT
1 check consistency - check consistency
2 compute monthly
means
TAPLOT MAPE
- get mean max/min 1 check consistency -compute 12
for mean zone elev. 2 generate daily time monthly ET
series of MAPE demand values

MAP MAT

1 recheck consistency - generate time


2 generate time series of MAP series of MAT.

MAPX
1. ‘Poor man’s” reanalysis
NWS Calibration Workshop, LMRFC March, 2009 slide 1
Analysis of Precipitation
• Non-Mountainous Areas
– Long Term Means Vary Slightly Across the Region
– Station Weights Based Totally on Location
• Mountainous Areas
– Long Term Means Vary Across the Region
– Ratio of Monthly Normals Used when Estimating
Missing Data
– Long Term Areal Mean Based on Isohyetal Analysis
– Station Weights Typically Don’t Sum to 1.0

NWS Calibration Workshop, LMRFC March, 2009 slide 2


Analysis of Precipitation
Criteria for
Mountainous vs Non-Mountainous Area Analysis
• Mountainous Areas: any area where the long-term mean
precipitation varies significantly over the area such that mean areal
values cannot be computed as a weighted average based solely on
the geographical location of the stations.

Station Variation
0 1-10% >10%

MAP Use Judgment PXPP


Analysis

NWS Calibration Workshop, LMRFC March, 2009 slide 3


Analysis of Precipitation
Station Selection
• Be conservative
• All stations within basin
• A few outside the basin for coverage and
estimation of missing data
• At least 5, preferably 10 years of data
• Complete as possible record
• Hourly stations for time disaggregation of daily
stations
• Go further out for mtn. areas to represent higher
elevations.
NWS Calibration Workshop, LMRFC March, 2009 slide 4
Selection of Potential Precipitation Stations
in Non-Mountainous Areas

H H

D
D D
D
H

D D
D
D D
H

D
H
D

H
D

NWS Calibration Workshop, LMRFC March, 2009 slide 5


Selection of Potential Precipitation Stations
in Non-Mountainous Areas
Hourly station needed to
H H
distribute nearby daily station
D values
D D
D
H
D
Daily station used as
D D
D D estimator for nearby
H
daily station

D
H
D

H
D

NWS Calibration Workshop, LMRFC March, 2009 slide 6


Precipitation data

Main river
channel
Standard Rain Gauge

Boundary of drainage area


NWS Calibration Workshop, LMRFC March, 2009 slide 7
Data Quality Control
• Method: Double Mass Analysis (DMA)
• Reasons
– Station moves
Need station history
– Equipment changes
• (e.g., add wind shield)
– Site Changes (vegetation,
buildings, etc)
• Legacy Programs that use DMA Wind Shield

– PXPP/MAP/MAT/MAPE

NWS Calibration Workshop, LMRFC March, 2009 slide 8


Standard Double Mass Analysis
Accumulation of station

Accumulation of the group of stations.


NWS Calibration Workshop, LMRFC March, 2009 slide 9
Analysis of Precipitation
NWS Double Mass Analysis
More logical that a single gage is inconsistent rather than entire group
Deviation of Station Accumulation from
Accumulation of Group Base

estimated data
+

0
documented station change
-

Accumulation of Average Precipitation of Group Base

NWS Calibration Workshop, LMRFC March, 2009 slide 10


Analysis of Precipitation
NWS Double Mass Analysis
Goal:
• one set of parameters that is good for entire period
• real inconsistencies are removed, not natural variations
Deviation of Station Accumulation from
Accumulation of Group Base

Documented station change


0

A B
calibration verification

Accumulation of Average Precipitation of Group Base


Given: Station 1 receives 50% of the weight for MAP. Without correction, it catches
20% more precip in verification period. MAPA< MAPB ,hard to calibrate
NWS Calibration Workshop, LMRFC March, 2009 slide 11
NWS Double Mass Analysis:
Definitions

Deviation of Station Acc. from


+

Acc. of Group Base


m
 n-1 Pi 
m


i1
Px   
i1  i1 n - 1 0

Acc precip. Average precip. -


of station of group

Acc. of Average Precip. of Group Base


Px= station analyzed
Pi= all stations other than m
 n-1
Pi 
Px
n = total no. of stations; n-1 stations
 n - 1
i1  i1 

in the group; group base acc. varies
slightly for each station.
M = no. of months Average precip. of group
NWS Calibration Workshop, LMRFC March, 2009 slide 12
What is the IDMA Tool?
• A GUI that aids in the quality control of
hydrologic data
– point observations of rainfall, temperature etc.
• Links legacy NWS pre-processors and a data
base of historical data/metadata
• Uses Double Mass Analysis (DMA) as primary
quality check
• Main output: multiplicative correction factors
– Typical range .90 < cf < 1.5

NWS Calibration Workshop, LMRFC March, 2009 slide 13


IDMA Linkages to Historical Data and Preprocessors

Station History Point time series data


Metadata Historical data
inventories
(Postgres)

Pre-processor Pre-processor
controls controls

Pre-processor Legacy
New correction
Input file Calibration
IDMA factors
Pre-processor
Current correction Current correction
-MAP
Factors Factors
-PXPP
-MAT
-MAPE
Accumulated
point time series
Mean areal
(‘dma’ file)
Time series
NWS Calibration Workshop, LMRFC March, 2009 slide 14
Pre-processor: program for analyzing precipitation, temperature, evaporation data
IDMA Steps
• Group stations geographically
• Identify missing data (white lines in IDMA)
• Identify station moves
• Pick period to correct to (usually the most
recent)

NWS Calibration Workshop, LMRFC March, 2009 slide 15


Analysis of Precipitation
NWS Double Mass Analysis
Simple Case
Deviation of Station Accumulation from
Accumulation of Group Base

documented station change


-
CF > ? CF = ?
Early period Later period

Accumulation of Average Precipitation of Group Base

CF = correction factor NWS Calibration Workshop, LMRFC March, 2009 slide 16


Analysis of Precipitation
NWS Double Mass Analysis
Simple Case
Deviation of Station Accumulation from
Accumulation of Group Base

documented station change


-
CF > 1.0 CF =1.0

Early period Later period

Accumulation of Average Precipitation of Group Base

CF = correction factor NWS Calibration Workshop, LMRFC March, 2009 slide 17


Analysis of Precipitation
NWS Double Mass Analysis
Deviation of Station Accumulation from

Complex case
Accumulation of Group Base

0
3
1
-
2
CF>1.0 CF<1.0 CF=1.0

Accumulation of Average Precipitation of Group Base

NWS Calibration Workshop, LMRFC March, 2009 slide 18


Analysis of Precipitation
NWS Double Mass Analysis - Cases
Deviation of Station Accumulation from
Accumulation of Group Base

estimated data

0
documented station change
-

Accumulation of Average Precipitation of Group Base

NWS Calibration Workshop, LMRFC March, 2009 slide 19


Analysis of Precipitation
NWS Double Mass Analysis - Cases
Check for bad data
Deviation of Station Accumulation from

in raw time series


estimated data
(can’t be corrected explicitly)
Accumulation of Group Base

0
documented station change
Good candidate for
- correction

Accumulation of Average Precipitation of Group Base

NWS Calibration Workshop, LMRFC March, 2009 slide 20


Analysis of Precipitation
NWS Double Mass Analysis - Cases
Deviation of Station Accumulation from

Given: no documented station changes


Accumulation of Group Base

Accumulation of Average Precipitation of Group Base

NWS Calibration Workshop, LMRFC March, 2009 slide 21


Analysis of Precipitation
NWS Double Mass Analysis - Cases
Deviation of Station Accumulation from

Given: documented station change in recent period


Accumulation of Group Base

Accumulation of Average Precipitation of Group Base

NWS Calibration Workshop, LMRFC March, 2009 slide 22


Analysis of OHD Basic QPE for DMIP 2
1987 – 2006
Output from STAT-QME Operation
Possible cause: bad data for the Blue Canyon station: “a lot of rain in Jan 95” was
recorded as zeros in the NCDC data. CNRFC set these values to ‘missing’ in their calibration.

Accumulated Simulation Error (mm of depth) : North Fork American River


Simulation Period 10/1998 to 8/2006 Dec 2005

0
Oct-88

Oct-89

Oct-90

Oct-91

Oct-92

Oct-93

Oct-94

Oct-95

Oct-96

Oct-97

Oct-98

Oct-99

Oct-00

Oct-01

Oct-02

Oct-03

Oct-04

Oct-05
Accumulated Sim. Error (mm)

-100
-200
-300 March 1998
-400
-500
Accum error
-600
-700
-800
Date

Inconsistent Precipitation?
NWS Calibration Workshop, LMRFC March, 2009 slide 23
DMIP 2: North Fork American River
OHD Streamflow Simulations

OHD
Distributed
Flow (cms)

Lumped

Observed

March 25, 1998

NWS Calibration Workshop, LMRFC March, 2009 slide 24


North Fork American River
Streamflow Simulations
Dec 19-26, 2005
OHD
Observed
Flow (cms)

NWS Calibration Workshop, LMRFC March, 2009 slide 25


Guidelines for Consistency Adjustments
• Use Seasonal Plots in Regions with Snowfall
– Winter - Months when Snowfall Predominates
– Summer - Months with Mostly Rainfall
– Snowfall Affected more by Station Changes
• Large Spikes in Plot Indicate Bad Data
• Group Stations by Location/Elevation
– Changes in Storm Track or Type will Alter the Relationship
between Stations (All Stations in Portion of the Area will Show a
Similar Shift in their Double Mass Plot -- This is Real and Should
Not be Corrected)
• If Any Doubt, Don’t Make an Adjustment
– Precipitation is Naturally Quite Variable
– Double Mass Plots Should Contain Wobbles
• Identify periods of missing data: these can’t be adjusted explicitly
• Station history files not always complete

NWS Calibration Workshop, LMRFC March, 2009 slide 26


Double Mass Analysis
Grouping of Precipitation Stations
in Non-Mountainous Areas

Group stations geographically


H H
in sets of 5
D
D D
D
H

D D
D
D D
H

D
H
D

H
D

NWS Calibration Workshop, LMRFC March, 2009 slide 27


Tests for Precipitation Homogeneity: Graphical procedures
Isolated Station Analyses Neighborhood Analyses

Double Mass Analysis Plots of test statistcs


1.a Plot of pure data: (Potter, 1981)
(Rhoades & Salinger, 1993)
1.b Cusums of isolated data:
Compare one station to another station Compare one station to ‘reference’
(Rhoades & Salinger, 1993)
(relative homogeneity) or ‘base’ series (absolute homogeneity)
(Kohler, 1949; WMO, 1971)

Single Cusum Plots Parallel Cusums Plots


(Kohler, 1949; Arndt and Redmond, 2004; (Rhoades and Reference series
network Reference series
Craddock, 1979) Salinger, 1993) network constant
changes with time
(Peterson and in time
Easterling, 1994)
Specialized Single Cusums Specialized Parallel
(Cumulative deviations) (Craddock, 1979) Cusums Plots (Rhoades Unweighted mean
and Salinger, 1993) Weighted mean of of ref. stations
1. Ratio
1. Deviations ref. stations. (Alexandersson, 1986)
2. Deviation from mean
2. Ratio 1.Using correlation N-1 stations (NWS)
3. Deviation from user defined NWS
3. Ratio of log coeffs. 20 stations
line segment (Arndt and Redmond, 2004)
(Alexandersson, 1986) 5 stations

Specialized Cusum plots


(deviations)
1. Difference NWS
Where do the NWS procedures fit in relation to 2. Ratio
peer-reviewed, published methods? 3. Ratio of logs

NWS Calibration Workshop, LMRFC March, 2009 slide 28


Historical Data Analysis
General Information Needed
1. Station data
2. Station history info: obs times, changes, location, moves
3. Topographic data

Analysis of Precipitation Analysis of Temperature Analysis of Evaporation


Information needed Information needed Information needed
Non Mountainous Mountains Non Mountainous Mountains Non Mountainous Mountains
1 isohyetal map - area vs elev. 1 evaporation maps - evap. vs elev.
-basin boundary 2 station weights -basin boundary curve 2 station weights curve
3 mean monthly evap.

PXPP MAT
1 check consistency - check consistency
2 compute monthy
means
TAPLOT MAPE
- get mean max/min 1 check consistency -compute 12
for mean zone elev. 2 generate daily time monthy ET
series of MAPE demand values

MAP MAT

1 recheck consistency - generate time


2 generate time series of MAP series of MAT.

NWS Calibration Workshop, LMRFC March, 2009 slide 29


Mean Areal Precipitation (MAP) Program
MAP weighting options: Grid Thiessen Predetermined

Grid Point Weighting


1. Overlays HRAP grid
5 2. For each grid pt. Finds closest station
in each of 4 quadrants; compute
distance d
3. Compute weight of each station 1/d
2
1 4. Normalize 4 weights
5. Sum all weights for each station
4 6. Normalize station weights to sum to 1.0
2
HRAP grid
3

Thiessen Polygon

2 Precipitation station

NWS Calibration Workshop, LMRFC March, 2009 slide 30


Mean Areal Precipitation (MAP) Program
MAP weighting: Grid Thiessen Predetermined

Thiessen Weighting
.
1. Overlays HRAP grid
2. Examines each grid point
1
3. Assigns grid point to closest
station
4. Station weight =
4 no. assigned points/
2 Total no. of grid points.

3
HRAP grid

1
Precipitation station
NWS Calibration Workshop, LMRFC March, 2009 slide 31
MAP3 Computational Sequence
1. Read in data and corrections
2. Applies corrections to observed data
3. Estimates missing hourly data using only other hourly stations.
n
Px
  Pi  w i 1
Px  i1 Pi w x,i  2
n
dx,i
w
i1
x,i
Px  precipitation at station being estimated
Pi  precipitation at the estimator station
n  number of estimating stations
i  station being used as an estimator
Px  mean monthly precipitation for station x
Px  mean monthly precipitation for station i
w x,i  station weight
dx,i  distance from station x to estimator i
NWS Calibration Workshop, LMRFC March, 2009 slide 32
MAP3 Computational Sequence
continued
4. Time distribute observed daily amounts into hourly values based on
surrounding hourly stations.
1. Procedure uses 1/d2 weighting for surrounding hourly stations.
2. If all hourly stations = 0, then all precipitation is put in last hour of the
daily station. Hour of the observation time. NFAR example
5. Estimate missing daily amounts using both hourly and daily gages; time
distribute these amounts
-If all estimators are missing, then uses 0.0
6. Generates file of station and group accumulated precipitation for IDMA
7. IDMA
1. -Compute correction factors
2. -Preliminary check of correction factors
3. -Insert correction factors into input file
4. -Re-run MAP3 for final check of consistency

8. Applies weights to station for each area


9. Computes hourly MAP time series
10. Sums to selected time interval, e.g., 3hr, 6hr.

NWS Calibration Workshop, LMRFC March, 2009 slide 33


Calibration MAP vs Operational MAP
Two Different Algorithms

Calibration MAP Operational MAP


1. Uses hourly and daily 1. Uses sub-daily and daily
precipitation amounts amounts.
2. Computes hourly MAP, 2. Computes 24 hr. MAP,
then sums to any time step. then distributes into 4 6-hr.
periods based on hourly
stations. Will use uniform
distribution if hourly not
available.
3. OFS Techniques available
for various conditions.
NWS Calibration Workshop, LMRFC March, 2009 slide 34
Importance of Mountainous Area Analysis

NWS Calibration Workshop, LMRFC March, 2009 slide 35


Precipitation Analysis
Objectives of Mountainous Area Procedure

• Compute Unbiased Estimate of Mean Areal Precipitation


• Ratio of Monthly Normals Used to Estimate Missing Data
• Long Term Areal Averages Based on Isohyetal Analysis
• Allow for Operational and Historical Estimates of MAP to
be Unbiased
• Same Method Used for Both Historical and Real Time
Data
• Exact Same Areal Averages Used in Both Cases
• Requires Good Definition of Monthly Station Normals

NWS Calibration Workshop, LMRFC March, 2009 slide 36


Mountainous Area Analysis
Steps
• Select stations, perform quality control
• Determine mean monthly precipitation for each
station for the period of record (Program PXPP)
• Determine annual or seasonal station weighting
• Determine mean annual precipitation for area or
sub area
• Determine station weights (adjust the relative
weights)
=> predetermined weights
• Compute MAP time series
NWS Calibration Workshop, LMRFC March, 2009 slide 37
Program PXPP
• Function: compute monthly means for stations having
different periods of record
• Uses monthly time step
• If any hour or day is missing, sets entire month to
missing
• Computes correlation tables to assist with station
weights.
station

Base station

time
NWS Calibration Workshop, LMRFC March, 2009 slide 38
Analysis of Precipitation in Mountainous Areas

Derivation of Isohyetal Maps

• Use existing map


• Derive using method of Peck (1962)
• Use NRCS PRISM data
– Note:
• May not have used all data NWS uses
• Data may not be consistent
• May need water balance analysis.

NWS Calibration Workshop, LMRFC March, 2009 slide 39


Verification of Isohyetal Maps
• Compare station means, seasonal and annual,
from PXPP to values from isohyetal maps
– Plot Ratio of PXPP mean to isohyetal map value
– Tabulate values, compute differences and average
ratio over the entire region
– Determine isohyetal map adjustment(s) for
historical data period of record
• Perform water balance computations
– Compute actual ET, from MAP and runoff, for
headwaters and local areas with minimal
complications
– Determine if actual ET values are reasonable (Can
adjust MAPs that are clearly in error at this point)

NWS Calibration Workshop, LMRFC March, 2009 slide 40


Determining Relative Weights
MAP weighting options: Grid Thiessen Predetermined

• Information to Consider
– Precipitation - Elevation Relationships, Seasonal and
Annual
– Correlation Relationships (from PXPP)
– Knowledge of Prevailing Storm Types and Tracks
(Anomaly Maps can Assist in Understanding)
• Typical Results
– Seasonal Weights in Intermountain West
– Winter Weights Based More on Elevation
– Summer Weights Based More on Distance
– Annual Weights in East and along West Coast

NWS Calibration Workshop, LMRFC March, 2009 slide 41


Mtn Area Analysis
Examples
• Juniata River, Pennsylvania
– Uses available isohyetal map
• Oostanaula River, Georgia
– Derivation of isohyetal map

NWS Calibration Workshop, LMRFC March, 2009 slide 42


(a)
Effects of Inconsistent Radar QPE
DMIP 1
Cumulative Simulation Error
Cum. Sim. - Cum. Obs. Runoff 800

600

400
Period of known underestimation
(mm)

200 and algorithm changes

-200

-400
Jan-93 Jun-94 Oct-95 Mar-97 Jul-98 Dec-99
Time (months)

ARS ARZ HRC LMP


OHD UTS UWO End Calib

NWS Calibration Workshop, LMRFC March, 2009 slide 43


Bias Correction of Archived Precipitation:
Example of ‘Poor Man’s’ Reanalysis
Cumulative Bias, Monocacy River at Jug Bridge (2100 km 2)
• Bias detected in MARFC MPE 200

archives prior to 2004 0

• Bias corrected precipitation -200


0 20 40 60 80 100 120

Sim - Obs Runoff (mm)


needed to support unbiased
-400
simulation statistics for a bias corrected
original xmrg
reasonable historical period -600

(can extend to ~9 years) -800

-1000

-1200
Time (months)

• Analysis of Monocacy River observed and


simulated flows shows reduction in
cumulative bias and improved consistency
when bias corrected precipitation is used
• A consistent bias can be removed through
calibration or through DHM-TF approach
Monocacy at Jug
Bridge (2116 km2)
NWS Calibration Workshop, LMRFC March, 2009 slide 44
Bias Correction of Precipitation
Monthly RFC MPE Monthly PRISM
Precipitation 03/97 (mm) Precipitation 3/97 (mm)

Adjusted RFC Hourly


RFC Hourly MPE Monthly Bias (ratio) MPE Precipitation
Precipitation 03/01/97 12z (mm)
03/01/97 12z (mm) NWS Calibration Workshop, LMRFC March, 2009 slide 45
Yu Zhang
Bias Correction of Precipitation
Example of typical improvements, particularly for small-medium events.

Monocacy River Original

Re-analysis
Original

Re-analysis

NWS Calibration Workshop, LMRFC March, 2009 slide 46

You might also like