Home About Free Downloads Register Help Videos More Stuff Site Map
 Gallery of Automated Nowcast Comparisons
Jan 1, 2016 to Aug 14, 2017
 

Rainfall Nowcasts MAE% Table begin Jan 2016:

Basin Run Start
Time
Minimum
Used
Storm
Type
HR1 HR2 HR3 Adjust
Type
Source Final Series
BigCA 1 V 0.1" S 31% 52% 65% 1 N0Q G/R adj
BigCA 2 V 0.1" S 48% 52% 76% 1 N0Q G/R adj
OneiNY 1 hh:33 0.1" S 30% 49% 43% 1 N0Q unadjusted
BigCA 3 hh:01 0.1" S 29% 45%  ---  1 N0Q unadjusted
BigCA 3 hh:01 0.1" S 66% 66% 65% none NWS gage avg
BigCA 4 hh:02 0.1" S 24% 64% 62% N0Q unadjusted
BigCA 4 hh:02 0.1" S 52% 49% 58% none NWS gage avg
AldriAL 1 hh:06 0.1" FS 32% 70% 96% none N0Q unadjusted
AldriAL 1 hh:06 0.1" FS 31% 81% 100% 1 N0Q unadjusted
AldriAL 1 hh:06 0.1" FS 74% 69% 62% none NWS gage avg
BigCA 5 hh:10 0.1" S 16% 33% 64% 1 N0Q unadjusted
BigCA 5 hh:10 0.1" S 72% 72% 72% none NWS gage avg
VillaAL 3 hh:07 0.1" FS 22% 31% 51% none N0Q unadjusted
VillaAL 3 hh:07 0.1" FS 83% 48% 125% none N0Q G/R adj
VillaAL 3 hh:07 0.1" FS 59% 53% 48% none NWS gage avg
Oneida 2 hh:17 0.1" S 26% 47% 39% none N0Q unadjusted
Oneida 2 hh:17 0.1" S 25% 47% 44% none N0Q G/R adj
Below are with out graphs or text.      
AldriAL 2 hh:05 0.1" T 49% 76% 96% none N0Q unadjusted
AldriAL 2 hh:05 0.1" T 120% 80% 97% none N0Q G/R adj
AldriAL 2 hh:05 0.1" T 60% 59% 59% none NWS gage avg
AldriAL 3 hh:10 0.1" T 56% 80% 96% none N0Q unadjusted
AldriAL 3 hh:10 0.1" T 52% 93% 99% none N0Q G/R adj
AldriAL 3 hh:10 0.1" T 62% 61% 53% none NWS gage avg
Oneida 3 hh:15 0.1" S 23% 46% 93% none N0Q unadjusted
AldriAL 4 hh:07 0.1" T 60% 54% 83% none N0Q G/R adj
Oneida 4 hh:07 0.1" T 60% 42% 50% none N0Q unadjusted
BigCA 6 hh:05 0.1" S 36% 71% 91% 1 N0Q unadjusted
BigCA 6 hh:05 0.1" S 27% 56% 86% 1 N0Q G/R adj
BigCA 6 hh:05 0.1" S 63% 66% 76% none NWS-4 gage avg
BigCA 7 hh:05 0.1" S 22% 41% 68% 1 N0Q unadjusted
BigCA 7 hh:05 0.1" S 45% 54% 71% 1 N0Q G/R adj
BigCA 7 hh:05 0.1" S 50% 45% 51% none NWS-4 gage avg
AldriAL 4 hh:06 0.1" T 20% 32% 98% none N0Q G/R adj
AldriAL 4 hh:06 0.1" T 81% 72% 59% none NWS gage avg
BigCA 8 hh:07 0.1" S speed/dir  out of  range none N0Q G/R adj
BigCA 8 hh:07 0.1" S 38% 36% 57% none NWS-4 gage avg
BigCA 9 hh:05 0.1" S 48% 41% 62% none N0Q G/R adj
BigCA 9 hh:05 0.1" S 33% 45% 65% none NWS-4 gage avg
OneiNY 5 hh:06 0.1" T 8% 71% 79% none N0Q G/R adj
OneiNY 5 hh:06 0.1" T 91% 91% 83% none NWS-3 gage avg
AldriAL 5 hh:06 0.1" T 60% 83% 135% none N0Q G/R adj
AldriAL 5 hh:06 0.1" T 52% 51% 60% none NWS-3 gage avg
BigCA 10 hh:05 0.1" S 59% 17% 61% none N0Q G/R adj
BigCA 10 hh:05 0.1" S 29% 38% 53% none NWS-4 gage avg
AldriAL 6 hh:06 0.1" T 31% 64% 64% none N0Q G/R adj
AldriAL 6 hh:06 0.1" T 32% 51% 75% none NWS-3 gage avg
Fargo 1 hh:21 0.1" T 66% 64% 91% none N0Q G/R adj
Fargo 1 hh:21 0.1" T 96% 96% 96% none NWS-3 gage avg
OneiNY 6 hh:06 0.1" T 53% 79% 100% none N0Q G/R adj
OneiNY 6 hh:06 0.1" T 76% 82% 91% none NWS-3 gage avg
Fargo 2 hh:21 0.1" T 9% 54% 100% none N0Q G/R adj
Fargo 2 hh:21 0.1" T 48% 44% 59% none NWS-3 gage avg
Oneida 7 hh:06 0.1" T 48% 65% 69% none N0Q G/R adj
Oneida 7 hh:06 0.1" T 26% 21% 59% none NWS-3 gage avg
TuscaAL 1 hh:05 0.1" T 33% 65% 72% none N0Q G/R adj
TuscaAL 1 hh:05 0.1" T 82% 88% 100% none NWS-3 gage avg
OneiNY 8 hh:12 0.1" T 34% 92% 98% none N0Q G/R adj
OneiNY 8 hh:12 0.1" T 98% 98% 98% none NWS-3 gage avg
Fargo 3 hh:13 0.1" T 60% 99% 100% none N0Q G/R adj
Fargo 3 hh:13 0.1" T 96% 97% 99% none NWS-3 gage avg
DaltGA 1 hh:10 0.1" T 69% 100% 100% none N0Q G/R adj
DaltGA 1 hh:10 0.1" T 95% 94% 79% none NWS-3 gage avg
OneiNY 9 hh:08 0.1" T 24% 100% 99% none N0Q G/R adj
OneiNY 9 hh:08 0.1" T 80% 75% 5% none NWS-3 gage avg
Fargo 4 hh:23 0.1" SQ 10% 89% 91% none N0Q G/R adj
Fargo 4 hh:23 0.1" SQ 20% 86% 78% none NWS-3 gage avg
BEGIN + or - MAE   Errors see note below      
These start in July and are small Tstorms unlike the win ter/spring SQ, F  & S above 
OneiNY 10 hh:10 0.1" SQ -41% -100% -100% none N0Q G/R adj
OneiNY 10 hh:10 0.1" SQ -87% -87% -74% none NWS-3 gage avg
OneiNY 10 hh:10 0.1" SQ -36% -90% -94% none N0Q G/R adj
OneiNY 10 hh:10 0.1" SQ -100% -100% -61% none NWS-3 gage avg
VillaAL 4 hh:07 0.1" T -73% -51% -100% 3 NCQ G/R adj
VillaAL 4 hh:07 0.1" T -93% -95% -97% none NWS-3 gage avg
VillaAL 5 hh:07 0.1" T +493% -88% -100% 3 NCQ G/R adj
VillaAL 5 hh:07 0.1" T -88% -91% -95% none NWS-3 gage avg
VillaAL 6 hh:11 0.1" T -78% -100% -99% 3 NCQ G/R adj
VillaAL 6 hh:11 0.1" T -100% -100% -100% none NWS-3 gage avg
OneiNY 11 hh:10 0.1" T -65% -98% -100% none N0Q G/R adj
OneiNY 11 hh:10 0.1" T -100% -100% -100% none NWS-3 gage avg
Fargo 5 hh:10 0.1" T +-44% -59% -59% 2 N0Q G/R adj
Fargo 5 hh:10 0.1" T -97% -98% -99% none NWS-3 gage avg
TuscaAL 2 hh:12 0.1" T +68% -74% -95% 3 N0Q G/R adj
TuscaAL 2 hh:12 0.1" T -92% -92% -76% none NWS-3 gage avg
Aldridge 7 hh:12 0.1" T -60% -100% -100% 3 N0Q G/R adj
Aldridge 7 hh:12 0.1" T -96% -97% -98% none NWS-3 gage avg
BurroAZ 1 hh:06 0.1" T -22% -59% +43% 3 NCQ G/R adj
BurroAZ 1 hh:06 0.1" T -96% -96% -96% none NWS-5 gage avg
VillaAL 7 hh:07 0.1" T -100% -100% -100% 3 NCQ G/R adj
VillaAL 7 hh:07 0.1" T -100% -100% -100% none NWS-3 gage avg
AldriAL 8 hh:12 0.1" T +6% -100% -100% 3 N0Q G/R adj
AldriAL 8 hh:12 0.1" T -95% -96% -98% none NWS-3 gage avg
BurroAZ 2 hh:08 0.1" T +-40% -99% -100% 3 NCQ G/R adj
BurroAZ 2 hh:08 0.1" T -100% -100% -28% none NWS-2 gage avg
BurroAZ 3 hh:08 0.1" T +-92% -51% -97% 3 NCQ G/R adj
BurroAZ 3 hh:08 0.1" T -91% -90% -84% none NWS-6 gage avg
BurroAZ 4 hh:08 0.1" T -55% -86% -81% 3 NCQ G/R adj
BurroAZ 4 hh:08 0.1" T -100% -100% +21% none NWS-3 gage avg
OneiNY 12 hh:12 0.1" T -20% -20% -80% 3 N0Q G/R adj
OneiNY 12 hh:12 0.1" T -96% -94% -84% none NWS-3 gage avg
VillaAL 8 hh:15 0.1" T -85% -97% -97% 3 N0Q G/R adj
VillaAL 8 hh:15 0.1" T -93% -95% -98% none NWS-3 gage avg
TuscaAL 3 hh:14 0.1" T -22% -100% --- 3 N0Q G/R adj
TuscaAL 3 hh:14 0.1" T -82% -69% --- none NWS-3 gage avg
OneiNY 13 hh:10 0.1" T -43% -55% -100% 3 N0Q G/R adj
OneiNY 13 hh:10 0.1" T -84% -76% -65% none NWS-3 gage avg
OneiNY 14 hh:06 0.1" T +87% -62% +101% 3 N0Q G/R adj
OneiNY 14 hh:06 0.1" T -87% -68% +-72% none NWS-3 gage avg
OneiNY 15 hh:09 0.1" SQ +53% -74% -80% 3 N0Q G/R adj
OneiNY 15 hh:09 0.1" SQ -92% -92% -80% none NWS-3 gage avg
TuscaAL 4 hh:06 0.1" T -83% +102% -88 3 N0Q G/R adj
TuscaAL 4 hh:06 0.1" T -87% -87% -82 none NWS-3 gage avg
BurroAZ 4 hh:10 0.1" T +-94% -97% -99% 3 NCQ G/R adj
BurroAZ 4 hh:10 0.1" T -82% -81% -65% none NWS-3 gage avg
VillaAL 9 hh:08 0.1" T +-45% -98% -100% 3 N0Q G/R adj
VillaAL 9 hh:08 0.1" T -78% -90% -95% none NWS-3 gage avg
Aldridge 8 hh:05 0.1" T -13% -97% -96% 3 N0Q G/R adj
Aldridge 8 hh:05 0.1" T -85% -89% -95% none NWS-3 gage avg
TuscaAL 5 hh:06 0.1" T -+57% -90% -80 3 N0Q G/R adj
TuscaAL 5 hh:06 0.1" T -56% -75% -84 none NWS-3 gage avg
OneiNY 16 hh:10 0.1" T -+16% -+19% +39% 3 N0Q G/R adj
OneiNY 16 hh:10 0.1" T -+56% -+52% -50% none NWS-3 gage avg
Aldridge 9 hh:05 0.1" SQ/T +-59% -+84% -85% 3 N0Q G/R adj
Aldridge 9 hh:05 0.1" SQ/T -86% -90% -95% none NWS-3 gage avg
BurroAZ 5 hh:10 0.1" T -+32% -+73% -+48% 3 NCQ G/R adj
BurroAZ 5 hh:10 0.1" T -70% -78% -86% none NWS-3 gage avg
Fargo 6 hh:11 0.1" T -+64% -91% -99% 2 N0Q G/R adj
Fargo 6 hh:11 0.1" T -31% -50% -75% none NWS-3 gage avg
Aldridge 10 hh:05 0.1" T -66% -86% -100% 3 N0Q G/R adj
Aldridge 10 hh:05 0.1" T -100% -100% -100% none NWS-3 gage avg
DaltGA 2 hh:06 0.1" T +300% -36% -61% none N0Q G/R adj
DaltGA 2 hh:06 0.1" T -70% -89% -100% none NWS-3 gage avg
Fargo 7 hh:11 0.1" T -64% -81% -100% 3 N0Q G/R adj
Fargo 7 hh:11 0.1" T +232% +40% -32% none NWS-3 gage avg
Fargo 8 hh:11 0.1" T +32% -96% -100% 3 N0Q G/R adj
Fargo 8 hh:11 0.1" T -34% -52% -76% none NWS-3 gage avg
OneiNY 17 hh:10 0.1" T +9% +15% +123% 3 N0Q G/R adj
OneiNY 17 hh:10 0.1" T +8% -21% -70% none NWS-3 gage avg
Fargo 9 hh:07 0.1" T -14% +3% -3% 3 N0Q G/R adj
Fargo 9 hh:07 0.1" T -60% -71% -85% none NWS-3 gage avg
BigCA 11 hh:05 0.1" OS -41% -92% -100% none N0Q G/R adj
BigCA 11 hh:05 0.1" OS +-41% +-27% -40% none NWS-4 gage avg
                   

 

****************** ONLY RAINFALL > 0.8 in/hr ****************************

Table values above are for final rainfall greater than 0.1 inch over the hour.

Ok, let's up the ante.  I just read that rainfall above 0.8 in/hr have zero chance of being usable 1 hour out.   Values below are only for predicted OR final rainfall greater than 0.8 inches.  Begin Aug 2016.

Basin Run Start
Time
Error Sources Storm
Type
HR1 HR2 HR3 Adjust
Type
Source Final Series Final Rainfall
VillaAL 8 hh:15 growth T -85% -97% -97% 3 NCQ G/R adj 1.32
VillaAL 8 hh:15   T -93% -95% -98% none NWS-3 gage avg 0.68
BurroAZ5 hh:10 G/R T +259% -94% -99% 3 NCQ G/R adj 0.48
BurroAZ5 hh:10 1.1 v .3 T -83% -94% -97% none NWS-3 gage avg 0.48
AldrAL11 01:08   T -27% -99% --- 3 N0Q G/R adj 0.84
AldrAL11 01:08   T -78% -84% --- none NWS-3 gage avg 0.43
AldrAL11 02:08 g/r T -46% -30% -57% 3 N0Q G/R adj 0.80
AldrAL11 02:08   T -83% -88% -94% none NWS-3 gage avg 0.58
VillaAL9 hh:12 growth T -99% -100% -100% 3 NCQ G/R adj 0.80
VillaAL9 hh:12   T --- --- --- none NWS-3 gage avg 0.00
NashTN1 hh:07 growth T -62% -100% -100% 3 NCQ G/R adj 0.84
NashTN1 hh:07   T -97 -97 -97 none NWS-3 gage avg 0.55
HousTX1 hh:12 growth T -42% -100% -100% 3 N0Q G/R adj 2.01
HousTX1 hh:12   T -94 -93 -81 none NWS-3 gage avg 0.77
DaltGA2 hh:07 g  & d T +72% -100% -100% 3 NCQ G/R adj 0.51
DaltGA2 hh:07   T -87 -91 -95 none NWS-3 gage avg 0.14
VillaAL10 hh:08 decay T +5800% 630% -100% 3 NCQ G/R adj 0.03
VillaAL10 hh:08   T -99% -97% -96% none NWS-3 gage avg 0.03
VillaAL10 hh:08   T -100% -100% -100% none HRRR gage avg 0.03
VillaAL11 hh:10   T 0% 0% 0% 3 NCQ G/R adj 0.01
VillaAL11 hh:10   T +300% +700% +1000% none NWS-3 gage avg 0.01
VillaAL11 hh:10   T +9200% 0% 0% none HRRR gage avg 0.01
BurroAZ6 hh:06 growth T -60% -83% -99% 3 NCQ G/R adj 0.79
BurroAZ6 hh:06   T -96% -98% -100% none NWS-3 gage avg 0.79
BurroAZ6 hh:06   T -99% -84% -84% none HRRR gage avg 0.79
HousTX2 hh:09   T +2638% +72% -99% none HRRR gage avg 0.05
VillaAL12 hh:12 azmith  T -100% -91% -93% 3 NCQ G/R adj 1.08
VillaAL12 hh:12   T -88% -94% -97% none NWS-3 G/R adj 1.08
VillaAL12 hh:12   T -98% -100% -100% none HRRR G/R adj 1.08
HousTX2 hh:10     -90% -99% -100% 3 NCQ G/R adj 1.30
HousTX2 hh:10     -81% -100% -100% 3 NCQ G/R adj 0.47
HousTX2 hh:10     -55% -93% -99% 3 NCQ G/R adj 2.50
HousTX2 hh:10     +16% -94% -100% 3 NCQ G/R adj 1.31
HousTX2 hh:10     -93% -95% -97% none NWS-3 G/R adj 1.83
HousTX2 hh:10     -92% -94% -97% none NWS-3 G/R adj 1.56
HousTX2 hh:10     -70% -79% -89% none NWS-3 G/R adj 1.09
HousTX2 hh:10     -60% -71% -85% none NWS-3 G/R adj 0.79
HousTX2 hh:10     -94% -98% -100% none HRRR G/R adj 2.08
HousTX2 hh:10     -16% +55% -84% none HRRR G/R adj 1.02
HousTX2 hh:10     -56% --- --- none HRRR G/R adj 2.26
HousTX2 hh:10     -45% -48% --- none HRRR G/R adj 1.07
TuscaAL6 hh:12     +32100% +7900% +9600% none HRRR G/R adj 0.004
NashTN2 hh:06     -100% -100% +5734% none HRRR G/R adj 0.03
                     

Note: Update Jul 12, 2016 a lot more often than not, errors less than 100% are event under estimates (alarm bells), 100% means 0 forecast, and greater than 100% means most event predictions over by AT LEAST 100%.  It's event mean UNSIGNED error remember.  I should probably add a sign to them as - under forecast, + over forecast.  Signed unsigned errors.  Hey science aint science till it's enginerring.

 

Storm types: S(stratiform), T(thunder), F(front), SQ(squal line), O(orographic)

NWS=National Weather Service average 6 hour QPF provided in equal hourly values.

NWS-2=Distribution type 2, Dallas 4 hour.

NWS-3=Distribution type 3, Vegas 5 hour.

(Type-3 distribution:  Hr1 is 3*6hrAvg, Hr2 is 2*6hrAvg, Hr3 is 1*6hrAvg, 0, 0. 0 and future 6 hour periods use the 5-minute distribution over the 6 hours.)

NWS-4=Distribution type 4, California 6 hour.

(Type-4 distribution:  Hr1 is 2*6hrAvg, Hr2 is 1.6*6hrAvg, Hr3 is 1.2*6hrAvg, 0.8, 0.4. 0 and future 6 hour periods use the 5-minute distribution over the 6 hours.)

NWS-5=Distribution type 5, Stepped Pyrimid 6 hour.

NWS-6=Distribution type 6, Pyrimid 6 hour.

Adjust type 1=hour 2 values average of hours 2 and 3 and not recommended.

HRRR=Multi Radar Multi Sensor HRRR radar reflectivity forecast.  I believe HRRR is based on the highest reflectivity in the clouds and if it behaves like the composite of angles then probably needs a 0.5 factor adjustment for surface rainfall rates.  I'll wait for some results before adjusting.

---------these adjusts below will change as I get wiser-----------

Adjust type 2=prorated adjust factor 0.6 to 0.4 for minutes 120 to 180 if count of 5-min rain > 0.05" greater than 2. This means heavy rainfall 3 hours out will only be 40% of that when it reaches the basin. I've been burned too often.

Adjust type 3=in addition to the decay of type 2, type 3 will decay hour 2 only with a prorated 0.8 to 0.6 factor if 5-min hour 2 rain count > 0.05" decreases by more than 3 from last run hour 3 or if count is > 6. Hour 1 will decay only with a prorated 1 to 0.7 factor if hour 1 count decreases by more than 3 from old hour 2. Have to try something.

 

Adjust Type 3 Reduction Success (right direction) Rates (begin 8/18/2016):

Hour 1:  11 for 12

Hour 2:  7 for 7

Hour 3:  12 for 12

 

Getrealtime Nowcasting Help Page

 

Getting these comparisons up and going was a bit of a challenge for me.  The rainfall MAE is ok once I figured out what I was looking at.  But how the Hec to table the runoff on small basins and be fair to the NWS 6-hr forecast is beyond me.  You will just have to settle for "the vibe of the thing" by looking at the hydrographs below with an eye toward volumes.  The Adjust Type comparisons use adjacent subbasins so are not the same.  These comparisons make use of GetRealtime's ability to write to M-tables as M-odel runs what ever is written to the R-eal database tables and then not over written by the next run.  I expect to improve all this with experience but for now tighten your thinking cap cause you're in for a bumpy ride.

 

Comparisons of "3" hour radar rainfall Nowcasts.  The forecast period will be at least 2.5 hours to 3.5 hours  depending on when the nowcast was made.  All these model runs start just after the topof the hour every hour so are close to 3.0 hours.  Nowcasts are made using the last available radar image and moving the basins upwind every 5 minutes to cover the 3 hour nowcast and always ending at the top of the hour. Normally when rain is in the area, model runs would be made at intervals of 30 or 15 minutes, but not for the study here  and I don't know what affect that has other than to replace a bad hour 3 as soon as possible with nothing. The direction and speed is based on NWS OP40-latest winds aloft which is updated every hour.  The radar adjustment applied is the defualt Gage/Radar ratio  Hour 2 (13th to 25th step) of the nowcast will be the average of each 5-minute value and the value 60 minutes later if 'Adjust Type=1' used.  The last data series is considered the final G/R Adjusted radar rainfall OR it may be final Unadjusted radar OR rain gage average for NWS comparisons depending on what was being saved with 'Run-' to the 'mtables'.  Under all this is the hourly average of the NWS 6 hour forecasts for the next 7 days, God luv'm, that can change the runoff graphs abruptly as the NWS changes their forecasts.

 

These runs are being saved here to see what can be learned in some sort of fashion and I have no idea how bad Nowcasts should be.  The NWS forecasts are usually pretty good for tomorrow or more and these Nowcasts are usually MUCH better than the NWS 6 hour forecast for the next 3 hours but they've proved me wrong before elsewhere... and often.  MAE% is mean absolute error and generally will use only final values greater than 0.1 inches rain.  If you look closely at the beginning of each graph you can spot the final series as the one that is plotted on top of all the others.

 

 

Big Creek Basin, California Sierra foothills:

 

Note:  There are 2 wind farms with very high reflectivities about 80-90 miles to the west so the 3rd hour for wind speeds greater than 40 mph with headings 80 to 107 degrees East will not be good.  Headings 92-97 should pass between the two demons. 

 

#1) Jan 5, 2016

Nowcast period:  3-hour

Adjust type:  1   (n+12)/2 for hour 2 only

Stormtrack: SAC 12000ft winds aloft

Nowcast radar adjust factor:  1.25  (heavy rains tend toward 1.0 at most radars)

Final series: G/R adjusted

Precip type: Stratiform rain 

Most nowcasts are every half hour with some 1 hour and at noon there was no rain on the nowcast so the model set it's schedule to be run next at 15:00... bad idea to use 3 hour schedule when rain in the area.  The stratiform winds aloft were verified manually a few times using GetNexrad but seemed even better than my eyeball.

National Weather Service SAC 12,000 ft Wind as headings, not froms:

2016-01-05 00:36 :Date 44 MPH NNE 20° winds aloft

2016-01-05 06:36 :Date 20 MPH NE 40° winds aloft

2016-01-05 14:36 :Date 20 MPH ENE 70° winds aloft

2016-01-05 18:49 :Date 21 MPH ENE 70° winds aloft

Basin Lag Time: 3.0 hours

I had set the runoff to be written just every 6 hours so I missed most of it but the NWS forecast did fine until hour 10:00.

 

##) Jan 6, 2016

Nowcast period:  3-hour

Adjust type:  1   (n+12)/2 for hour 2 only

Stormtrack: SAC 12000ft winds aloft

Nowcast radar adjust factor:  1.25  (heavy rains tend toward 1.0 at most radars)

Final series: G/R adjusted

Precip type: Stratiform rain 

Most nowcasts are every half hour with some 1 hour.  The stratiform winds aloft were verified manually a few times using GetNexrad.  Snow occurred and Wundergages showed about 0.5" of melt the next day and who knows how much wind induced undercatch due to snow there was so radar adjustments are iffy.  So take with a grain of salt and why final values should be unadjusted raw radar for the mission here.

National Weather Service SAC 12,000 ft Wind as headings, not froms:

2016-01-06 00:49 :Date 43 MPH NE 40° winds aloft

2016-01-06 06:20 :Date 31 MPH NNE 30° winds aloft

2016-01-06 12:50 :Date 17 MPH NE 40° winds aloft

2016-01-06 18:55 :Date 12 MPH NNE 30° winds aloft

Basin Lag Time: 3.0 hours

 

#3) Jan 15, 2016

Nowcast period:  3-hour

Adjust type:  1   (n+12)/2 for hour 2 only

Stormtrack: OP40 :25 miles west of Buck Meadows

Nowcast radar adjust factor:  1.25  (heavy rains tend toward 1.0 at most radars)

Final series: raw N0Q radar

Precip type: Stratiform rain 

Most nowcasts are every half hour with some 1 hour.  No final hourly radar>0.1" so no MAE.  There are 2 wind farms with very high reflectivities about 80-90 miles to the west so the 3rd hour for wind speeds greater than 40 mph with headings 80 to 107 degrees East will not be good and explains the high 3:19 hour 3.  And to top it off, this is the typical direction for rain events.  Where did the good ol' days of nuclear go?

National Weather Service OP40 - 9,800 ft Wind as headings, not froms:

Jan 15,  02:19: 51 MPH East 99° ALT 9967

Jan 15,  03:19: 58 MPH ESE 103° ALT 9917

Jan 15,  04:19: 51 MPH ESE 110° ALT 9983

Jan 15,  05:19: 51 MPH ESE 110° ALT 9983

Jan 15, 08:19: 35 MPH SE 127° ALT 9976

Jan 15, 09:19: 34 MPH SE 132° ALT 9970

Jan 15, 10:19: 36 MPH SE 133° ALT 10032

Basin Lag Time: 3.0 hours

 

#3) Jan 18, 2016

Nowcast period:  3-hour

Adjust type: none

Stormtrack: OP40 :25 miles west of Buck Meadows

Nowcast radar adjust factor:  1.25  (heavy rains tend toward 1.0 at most radars)

Final series: raw N0Q radar

Precip type: Stratiform rain 

Nowcasts are every 1 hour.  The wind speeds and direction missed the wind farms.  I screwed up the hourly radar flagging until about 10AM where only 2 hours were being flagged.  So I think hours 1 and 2 are really hours 2 and 3... but I'm not sure so I will leave it alone.

National Weather Service OP40 - 9,800 ft Wind as headings, not froms:

Jan 17, 23:01: 26 MPH NE 46° ALT 10098'

Jan 18, 00:01: 26 MPH NE 46° ALT 10098'

Jan 18, 01:03: 19 MPH NE 51° ALT 10088'

Jan 18, 02:01: 17 MPH ENE 64° ALT 10125'

Jan 18, 03:01: 14 MPH East 86° ALT 10121'

Jan 18, 04:01: 13 MPH ENE 64° ALT 10098'

Jan 18, 05:01: 13 MPH ENE 64° ALT 10098'

Jan 18, 07:01: 24 MPH ENE 59° ALT 10059'

Jan 18, 08:01: 18 MPH NE 52° ALT 10072'

Jan 18, 09:01: 18 MPH NE 52° ALT 10072'

Jan 18, 10:01: 28 MPH SE 132° ALT 10151'

Jan 18, 11:01: 27 MPH ESE 120° ALT 10138'

Jan 18, 12:01: 29 MPH ESE 114° ALT 10161'

Basin Lag Time: 3.0 hours

NWS Forecasts:

 

#4) Jan 19, 2016

Nowcast period:  3-hour

Adjust type: none

Stormtrack: OP40 :25 miles west of Buck Meadows

Nowcast radar adjust factor:  1.25  (heavy rains tend toward 1.0 at most radars)

Final series: raw N0Q radar

Precip type: Stratiform rain 

Nowcasts are every 1 hour.  The wind speeds and direction missed the wind farms. After the rain stopped I raised the winds aloft altitude from 10,000 ft to 12,000 ft because I didn't like the direction and speed.  The change put it back in line with what I was seeing but now it's right back over the wind farms.

National Weather Service OP40 - 9,800 ft Wind as headings, not froms:

Jan 19, 04:02: 27 MPH NE 40° ALT 10128'

Jan 19, 05:02: 27 MPH NE 40° ALT 10128'

Jan 19, 06:02: 25 MPH NNE 31° ALT 10092'

Jan 19, 07:02: 35 MPH NNE 29° ALT 10092'

Jan 19, 08:02: 44 MPH NE 50° ALT 10082'

Jan 19, 09:02: 63 MPH NE 45° ALT 10065'

Jan 19, 10:02: 58 MPH NE 47° ALT 10046'

Jan 19, 11:02: 69 MPH ENE 61° ALT 10065'

Reusing Winds Aloft: Jan-19 12:02

Jan 19, 13:02: 56 MPH East 88° ALT 12164'

Jan 19, 14:02: 59 MPH ESE 103° ALT 12144'

Basin Lag Time: 3.0 hours

NWS Forecasts:

 

#5) Jan 22, 2016

Nowcast period:  3-hour

Adjust type: none

Stormtrack: OP40 :25 miles west of Buck Meadows

Nowcast radar adjust factor:  1.25  (heavy rains tend toward 1.0 at most radars)

Final series: raw N0Q radar

Precip type: Stratiform rain 

Nowcasts are every 1 hour.  The wind speeds and direction missed the wind farms.

National Weather Service OP40 -Speed 11,000 ft, Direction 11,000 ft  Wind as headings, not froms:

Jan 22, 15:10: 37 MPH NE 34° ALT 12157'

Jan 22, 16:11: 35 MPH NE 41° ALT 12171'

Jan 22, 16:42: 35 MPH ENE 60° ALT 12171'

Jan 22, 17:10: 35 MPH ENE 60° ALT 12171'

Jan 22, 18:10: 36 MPH NE 56° ALT 12180'

Jan 22, 19:10: 39 MPH NE 56° ALT 12184'

Jan 22, 20:10: 37 MPH NE 42° ALT 12161'

Jan 22, 21:10: 35 MPH NNE 32° ALT 12134'

Jan 22, 22:10: 44 MPH NE 44° ALT 12138'

Reusing Winds Aloft: Jan-22 23:10

Jan 23, 00:11: 49 MPH NE 37° ALT 12125'

Basin Lag Time: 3.0 hours

NWS Forecasts:

 

 

Birmingham, Alabama:

 

#1) Jan 9, 2016

Nowcast period:  3-hour

Adjust type:  0 and 1   (n+12)/2 for hour 2 only

Stormtrack: BHM 18000ft winds aloft      BAD IDEA!!!

Nowcast radar adjust factor:  1.0  (heavy rains tend toward 1.0 at most radars)

Final series: Unadjusted radar

Precip type: Thunderstoms small squall lines 

Nowcasts are every 1 hour at hh:46.  The winds aloft at 18000ft were 101 MPH.  Not a good thing.  I loaded the winds aloft file on a GetNexrad loop and the 3 rd hour was halfway to New Orleans.  This completely missed the storms traveling about 50 mph and I have since lowered the altitude to 12000ft.

National Weather Service BHM 18,000 ft Wind as headings, not froms:

2016-01-09 00:47 :Date 55 MPH NE 40° winds aloft

2016-01-09 06:46 :Date 86 MPH NE 50° winds aloft

2016-01-09 12:36 :Date 101 MPH ENE 60° winds aloft

   altitude lowered from 18000ft to 12000ft

2016-01-09 16:50 :Date 60 MPH NE 50° winds aloft

2016-01-09 18:51 :Date 54 MPH ENE 60° winds aloft

Basin Lag Time: 3.7 hours

The NWS had the right volume from the start of the day but a few hours early (probably used 100mph winds) and I had nothing with 100 mph winds.

 

#2) Jan 21, 2016

Nowcast period:  3-hour

Adjust type: none

Stormtrack: OP40 :  lat/long of basin

Nowcast radar adjust factor:  1.0  (heavy rains tend toward 1.0 at most radars)

Final series: raw N0Q radar

Precip type: Cold front.

Nowcasts are every 1 hour.  I was comparing wind speed and direction every hour with my eyeball using GetNexrad and didn't like what I saw.  17:11 shows typical eyeball speed and direction.  I finally decided to hardwire the direction in GetRealtime at 18,000 ft and let the user set the speed altitude which I kept at 9800 ft.  This put the direction and speed right on target but a little late in the game starting 21:08.

National Weather Service OP40 - 9,800 ft Wind as headings, not froms:

Jan 21, 16:11: 29 MPH NNE 26° ALT 9937'

Jan 21, 17:11: 29 MPH NNE 26° ALT 9937'

Jan 21, 17:32: 55 MPH NE 52° GetNexrad  ***Expires 0 hours so not used***

Jan 21, 18:11: 31 MPH NNE 33° ALT 9947'

Jan 21, 19:11: 39 MPH NNE 25° ALT 9930'

Jan 21, 20:11: 49 MPH NE 37° ALT 9930'

begin direction from 18000 ft, speed at same 9800 ft.

Jan 21, 21:08: 58 MPH NE 45° ALT 9884'

Jan 21, 22:08: 56 MPH NE 46° ALT 9865'

Jan 21, 23:08: 56 MPH NE 46° ALT 9865'

Basin Lag Time: 3.0 hours

NWS Forecasts:

 

#3) Jan 26, 2016  FUBAR FUBAR FUBAR!  Although the MAE's were 22%, 31%, 51%.

Nowcast period:  3-hour

Adjust type: none

Stormtrack: OP40 :  lat/long of basin

Nowcast radar adjust factor:  1.0  (heavy rains tend toward 1.0 at most radars)

Final series: raw N0Q radar

Precip type: Cold front stratiform rain.

Nowcasts are every 1 hour.  I checked each Nowcast and speed and direction was right on... until the cold front approached and thats bad news.  You have to manually do the nowcast then.  I didn't till to late.  The major problem was not the Nowcast radar, it was that the radar was so far out of wack it needed a 0.5 G/R adjustment, not the default 1.0.

National Weather Service OP40 - 9,800 ft Wind as headings, not froms:

Jan 25, 10:33: 38 MPH NE 53° ALT 9907'

Jan 26, 09:02: 39 MPH ENE 61° ALT 9993'

Jan 26, 10:07: 41 MPH ENE 57° ALT 9980'

Jan 26, 11:07: 40 MPH ENE 61° ALT 9970'

Jan 26, 12:07: 40 MPH ENE 61° ALT 9970'

Jan 26, 13:07: 33 MPH ENE 61° ALT 9940'

Jan 26, 14:07: 37 MPH ENE 58° ALT 9927'

Reusing Winds Aloft: Jan-26 15:07

Jan 26, 16:07: 17 MPH NE 43° ALT 9891'

Jan 26, 16:46: 53 MPH ENE 71° GetNexrad  and set to expire in 4 hours

Basin Lag Time: 3.0 hours

The NWS whipped the snot out of this runoff compared to Nowcasts.  FUBAR!!!

But the problem wasn't the Nowcasts as shown by their nice MAE's, it was this Unadjusted radar compared to final G/R adjusted radar: 

 

NWS Forecasts:

 

#5) Feb 15, 2016 

Nowcast period:  3-hour

Adjust type: none

The hour 1 adjusted radar nowcast MAE was 89%... but the peak flow looks fine, just early:

 

 

Huntsville, Alabama (you would think same as Birmingham... NOT!!!!:

 

#1) Jan 22, 2016

Nowcast period:  3-hour

Adjust type: none (and type 1 not shown)

Stormtrack: OP40 :  lat/long of basin

Nowcast radar adjust factor:  1.82 UH OH! Bad Idea???  (heavy rains tend toward 1.0 at most radars)

Final series: raw N0Q radar

Precip type: Cold front rain then snow

Nowcasts are every 1 hour. I only checked the radar at the end of the event and the directions were going 120% apart from one side of the basin to the other.  Must have had a low pressure move right over the basin screwing all of this completely up.... and SNOW did not help the rain gage readings.

National Weather Service OP40 - 9,000 ft Wind as headings, not froms:

Jan 22, 00:06: 38 MPH NNE 27° ALT 9782'

Jan 22, 01:06: 44 MPH NNE 18° ALT 9746'

Jan 22, 02:06: 31 MPH NNE 28° ALT 9713'

Jan 22, 03:06: 30 MPH NNE 26° ALT 9680'

Jan 22, 04:05: 33 MPH NNE 27° ALT 9723'

Jan 22, 05:06: 33 MPH NNE 27° ALT 9723'

Jan 22, 06:06: 18 MPH NNE 20° ALT 9638'

Jan 22, 07:06: 18 MPH NNE 20° ALT 9638'

Jan 22, 08:06: 36 MPH North 7° ALT 9598'

Jan 22, 09:06: 37 MPH NNW 339° ALT 9565'

Jan 22, 10:06: 30 MPH NNE 13° ALT 9555'

Basin Lag Time: 3.0 hours

NWS Forecasts:

 

#3) Feb 15, 2016

Nowcast period:  3-hour

Adjust type: none

 

 

Oneida Creek Subbasin, central New York:

 

#1) Jan 10, 2016

Nowcast period:  3-hour

Adjust type:  1   (n+12)/2 for hour 2 only

Stormtrack: SYR 12000ft winds aloft

Nowcast radar adjust factor:  1.14  (heavy rains tend toward 1.0 at most radars)

Final series: Unadjusted radar

Precip type: I wasn't watching but afterwards I see a squall line to the southeast

Nowcasts made once an hour at hh:33.

National Weather Service SYR 12,000 ft Wind as headings, not froms:

2016-01-10 00:33 :Date 59 MPH North 350° winds aloft

2016-01-10 06:33 :Date 63 MPH North 10° winds aloft

 

Basin Lag Time to Gage: ~12 hours

I didn't think conditions warranted running the runoff... dang.  The pink series is using yesterday morning's NWS forecast.  In this business you need the USA radar going 24/7 to keep up with it all.... or have all your basins running everything 24/7.

 

Lucky I only dabble in this business... although I do have a Win10 laptop doing 24/7 for California with web updates.  I recently put it on a Windows HomeGroup to copy the database to my REAL Win7 desktop for daily web graphs but I don't like it.  I could map my daily Excel updates to my laptop database over my HomeGroup but I don't know.  Maybe I should see how much can be done with that foul Win10 laptop without database copying.  Cloud avoidance, although my cable provider is not very reliable when it rains.... or a new Star Wars movie comes out.  Stay tuned... Ah.. it works. I just set my connectins for GetAccess and GetGraphs and Excel to //laptop and it works great.  So no more copy database.  But it aint no cloud and I don't like the taste of it yet.  I'll try it for a while...

 

Oneida Creek Subbasin, central New York:

 

#2) Feb 3, 2016

Nowcast period:  3-hour

Adjust type:  none

Stormtrack: OP40 :  lat/long of basin

Nowcast radar adjust factor:  1.14  (heavy rains tend toward 1.0 at most radars)

Final series: Unadjusted radar and G/R Adjusted

Precip type: I wasn't watching.

Nowcasts made once an hour at hh:33.

National Weather Service SYR 7,900 ft speed and direction, Wind as headings, not froms:

Feb 03, 00:17: 55 MPH ENE 58° ALT 8289'

Feb 03, 01:17: 55 MPH ENE 58° ALT 8289'

Feb 03, 02:17: 46 MPH ENE 63° ALT 8312'

Feb 03, 03:17: 40 MPH ENE 58° ALT 8312'

Feb 03, 04:17: 40 MPH ENE 58° ALT 8312'

Feb 03, 05:17: 40 MPH ENE 58° ALT 8312'

Feb 03, 06:17: 64 MPH ENE 62° ALT 8384'

Feb 03, 08:17: 76 MPH ENE 64° ALT 8391'

Feb 03, 09:17: 54 MPH ENE 65° ALT 8411'

Feb 03, 10:14: 55 MPH ENE 72° ALT 8463'

Basin Lag Time to Gage: ~12 hours

 

 

 

 

Findings So Far:

 

1--Ive seen enough alreadty to say Woo Hoo!!!  And you don't have to do anything... but it can and will bite you if your snoozing during the Big One... the NWS winds aloft aint your mama.

 

2--Based on hour 2 MAE I'm starting to think my simple Adjust Type 1 is a bad idea.  I'm saving both Adjust Types at Birmingham so we will see eventually.  I have since stopped using Adjust Type 1 and will now try Adjust Type 2 which is a proration of adjust factor 1 to 0.4 for minutes 60 to 180 if 5-minute rain > 0.05".

 

3--Beginning Jan 10 after 9:00 AM I changed the nowcast period from at least 3 hours but possibly 4 to 2.5 to 3.5 hours based on the final radar image time.  The nowcast always ends at the top of the hour so if it's minute hh:30 to hh:00 then 3 to 3.5 hours will actually be forecast where hh:00 would be 3 hours.  If it's hh:05 to hh:25 then 2.5 to 3 hours where hh:05 would be 2 hrs and 55 min.  This also means for last images dates hh:00 to hh:25 then 3 hourly values will be flagged 14, else 4 hourly values will be flagged 14 for nowcast for database queries used in the analysis here.  So hourly runs NEAR the top of the hour should provide reasonably consistent populations.  When I move to 30-minute samples I.... 'll cross that bridge then.

 

4--I've also added 'Mtables OFF' checkbox under the batch so you can make runs without disturbing the Mtable sequences by injecting ill timed runs.  You would only want to do this if your doing something like what I'm trying to do here with automated Excel sheets. M-tables are used by GetRealtime to save whatever you want as the model is writing to the 'real' database R-tables also. 

 

5--The lake effect snows comming off of Lake Erie are a lot closer to the ground than radar mid-beam scan elevation of 10,000 ft.  I set my winds aloft altitude to 4000 ft for the winter in New York based on manual GetNexrad tracking.  Apparently radar sees everything whether it's looking at it or not.  It will probably bite me when the wind shifts from the south.

 

6--If your wind direction runs you off your N0Q radar image, then set GetRealtime to use the Iowa Mesonet's map cutting web service to create real-time radar images for any location at any size and you probably can get by with half the 1000x1000 pixels of N0Q.

 

7--The Birmingham storms have led me to hard wire the wind direction at altitude 18,000 ft and the user will set the speed altitude to suite, probably below 12,000 ft... or optionally both altitudes may be set. GetRealtime.exe has been updated.

 

8--Cold/Warm Fronts tend to end badly.  The winds aloft countinue as they have but the trailing edge of the front rapidy moves eastward as the storm ends leaving high hour 2 and 3 rains which then turn out to be zero.  I have added the option to use the Maddox 30R70 rule to shift direction and speed if needed.

 

9--But this was a minor snafu compared to unadjusted radar vs G/R adjusted which can be running at 0.5, not 1.  I have updated GetRealtime.exe to use the average of the last ratio and the default ratio for adjusting the nowcast period to help this... but this could be tricky when the last ratio is 3 and then the next is 0.6 like what happened here so I'll keep an eye on this.

 

10--When the winds are blowing 60 to 90 mph to the SE, hour 3 is just looking at some greenish haze, not real radar.  Nowcasts aside, I've been getting better results at Village Cr with the NCQ of Iowa's composite compared to GWX at this basin's 100 mile range so this is a simple change for making Nowcasts get a fair deal.  NCQ way over estimate rainfall and generally need at least a 0.5 G/R ratio adjust.  If as I think it will turn out, hour 3 nowcasts are a pipe dream for Thunderstorms so I doubt much help will be found with composite radar, but we shall see.

 

11--Nowcasts 3 hours out ALWAYS if big never pan out.  I have updated GetRealtime to take hours 2 and 3 with a grain of salt but will probably still over shoot.

 

12--In Arizona mountains with summer winds at 15 mph, sometimes the storm cells advance from the oppisite direction at about the same speed.  I guess condensation is advancing thru the water supply at twice the speed of air mass movement.  In this case one could manually set speed and direction using GetNexrad with expiration in 24 hours.

 

13.--If the best the NWS QPF can do is 50% low then, adding 2*P1 to the GetRealtime setup NWS forecast line is beginning to look like a good idea.

 

14--Using one of the distribution types on the NWS 6-hour QPF's will definately improve NWS forecast if you should decide not to use Nowcasts.  I recommend a 2 hour Nowcast with a NWS distribution.  The 3rd hour of a Nowcast will likely be worse than the distributed NWS QPF for hour 3.  If using a Nowcast then you may (see 11 above) want to copy distribution 2, 3, or 4 to a distribution greater than 4 so that the average NWS QPF is used for the next 6 hours but still distributed beyond 6 hours (see Help).

 

All seems to work pretty well out to 3 hours unless you are in the Southern US with prairee fire squalls and mega thunderstorms the norm rather than exception.  The problem then becomes from what direction to look at the radar.  So 1-hour nowcast is the easy way out here and go with the NWS QPF with a distribution (always use a distribution).  You can manually override winds aloft with GetNexrad for storm direction if your not asleep.  Or like I'm now testing in Alabama, just a permanate 30R70 adjust for the 3 hours and go back to bed.... Or just use the NWS 6-hour forecast with a distribution, my adjustments for hours 1, 2 and 3 seem to work pretty darn well... sometimes.

 

The literature says: "Lee et al. (2009) found that the Critical Success Index CSI decreased with increasing rain rate and forecast lead time: the CSI for 60 minute rainfall forecasts decreased from 0.60 for 0.1 mm/h to 0.2 for 10 mm/h rain rates. Ebert et al. (2004) reported that the CSI for rain greater than 20 mm/h is essentially zero. This implies that the use of nowcasting techniques to predict the precise location of extreme rain for flash flood warning may not be viable."  Hence the 0.8 in/hr table at top to see how true this is.



Some additional page links about ET and Nexrad Radar help and comparisons:

List of How To Videos on Youtube
Help Page for GetNexrad.exe
Nexrad Rainfall-Runoff Comparison Big Creek, San Jouqin Valley, CA
Nexrad Snowfall Comparisons in western central Sierras, CA


 

 WEBSITE MAP


                     
WU   USGS  USBR USCS  CWRD  NWS

More Free Downloads

 

Comments/Questions

Contact  Me
Copyright 2016 Carson, Reproduction permitted with attribution.
Label