heavy precipitation events: a global survey, 1998-2007€¦ · university of miami, miami, florida,...

7
Heavy precipitation events: a global survey, 1998-2007 Brian Mapes 1 and Robert Joyce 2 1 Rosenstiel School of Marine and Atmospheric Sciences University of Miami, Miami, Florida, USA. Email: [email protected] 2 Climate Prediction Center, NOAA, Washington, DC, USA ABSTRACT The heaviest precipitation events (here, 1-day and 3-day accumulations) at every location in the tropics and subtropics have been identified in a 10-year dataset of precipitation estimates at 1/4 degree 3-hourly resolution (the TRMM 3B42 product, based on rainrate-calibrated IR data). Some basic statistics of amount and seasonality are shown here. We use this database of record-setting events for selecting case studies. Animations are made semi-automatically from a Web-accessible satellite imagery archive; the next step will include adding weather analysis and forecast fields, including ensemble forecasts from the TIGGE archive. Examination of many cases selected in this objective manner seems like a useful middle ground between arbitrary case studies and bland statistics, but may require nontraditional forms of dissemination. 1. Introduction Monsoons can bring damaging rains, as well as life-giving ones. Motivated by the topic of ‘monsoon severe weather,’ we have made a systematic survey of satellite-estimated heavy precipitation events, defined here in terms of 1- day or 3-day accumulations. Comparative studies of such events at various locations may reveal interesting synoptic commonalities and differences, as well as giving a sense of the random or unpredictable aspects. With increasing data availability and computing capabilities, it is fairly easy to assemble detailed case animations. It is not our purpose here to reach general conclusions, but rather only to illustrate some methods and products that might be informative and useful for forecaster training as well as research. 2. Data sets used Our 1998-2007 heavy rain event catalog was constructed from the TRMM 3B42 product (Huffman et al. 2007), version 6. This is a 3-hourly 1/4 degree data set of satellite-estimated precipitation rate. In this product, all available sources are blended to obtain best estimates, which means the nature of the product is not homogeneous in time. Infrared data are calibrated to rain units and used to fill space-time gaps, yielding the 3 hourly coverage. For these

Upload: vananh

Post on 04-Jun-2018

213 views

Category:

Documents


0 download

TRANSCRIPT

Heavy precipitation events: a global survey, 1998-2007

Brian Mapes1 and Robert Joyce2

1Rosenstiel School of Marine and Atmospheric SciencesUniversity of Miami, Miami, Florida, USA. Email: [email protected]

2Climate Prediction Center, NOAA, Washington, DC, USA

ABSTRACTThe heaviest precipitation events (here, 1-day and 3-day accumulations) at

every location in the tropics and subtropics have been identified in a 10-yeardataset of precipitation estimates at 1/4 degree 3-hourly resolution (the TRMM

3B42 product, based on rainrate-calibrated IR data). Some basic statistics ofamount and seasonality are shown here.

We use this database of record-setting events for selecting case studies.Animations are made semi-automatically from a Web-accessible satelliteimagery archive; the next step will include adding weather analysis andforecast fields, including ensemble forecasts from the TIGGE archive.

Examination of many cases selected in this objective manner seems like auseful middle ground between arbitrary case studies and bland statistics, but

may require nontraditional forms of dissemination.

1. Introduction

Monsoons can bring damaging rains, as well as life-giving ones. Motivatedby the topic of ‘monsoon severe weather,’ we have made a systematic surveyof satellite-estimated heavy precipitation events, defined here in terms of 1-day or 3-day accumulations. Comparative studies of such events at variouslocations may reveal interesting synoptic commonalities and differences, aswell as giving a sense of the random or unpredictable aspects. With increasingdata availability and computing capabilities, it is fairly easy to assembledetailed case animations. It is not our purpose here to reach generalconclusions, but rather only to illustrate some methods and products thatmight be informative and useful for forecaster training as well as research.

2. Data sets used

Our 1998-2007 heavy rain event catalog was constructed from the TRMM3B42 product (Huffman et al. 2007), version 6. This is a 3-hourly 1/4 degreedata set of satellite-estimated precipitation rate. In this product, all availablesources are blended to obtain best estimates, which means the nature of theproduct is not homogeneous in time. Infrared data are calibrated to rain unitsand used to fill space-time gaps, yielding the 3 hourly coverage. For these

reasons, 3B42 estimates, and comparisons over time (like our heaviestprecipitation events) are not perfect quantitatively. Still, when 3B42 indicatesheavy precipitation accumulations over many sample times (1-3 days), it is verylikely that some significant weather in fact took place. Other rainfall productscould be examined similarly in the future.

At each point in the dataset (50S-50N latitude belt), accumulatedprecipitation was computed using a moving average in time: a 9-point centeredwindow produced 1-day (actually 27-hour) accumulations, while a 25-pointsmoother produced 3-day (actually 3 days + 3 hours) accumulations. The largestvalue of the smoothed time series within the 10-year record was identified asthe heaviest precipitation event. Its amount (in mm), and its central date andtime, were captured in arrays as a function of latitude and longitude. Actually,we isolated the heaviest several precipitation events in each of the 12 calendarmonths, but this paper discusses only the single heaviest event in the entire 10-year record at each location.

The 10-year monthly climatology was also computed from the samedataset, so that “heavy” precipitation events can be selected using bothabsolute criteria (many mm) and relative criteria (say, a month or season worthof precipitation falling in 1 or 3 days). A digital elevation model of topographyand bathymetry was interpolated to the same 1/4 degree grid, so that we canisolate events over land, subdivide by ground altitude or slope, etc. T

High-resolution, 3-hourly geostationary satellite imagery covering thisentire 10-year period (and more) is freely and continuously available athttp://www.ncdc.noaa.gov/gibbs/. Scripts were developed to downloadmultiple images, and the free software convert was used to crop them to anarea of interest and create animation loops. Future efforts will overlay synopticcharts of pressure, heights, winds, etc. from reanalysis and (for year 2007events) from the THORPEX Interactive Grand Global Ensemble (TIGGE)forecast database, to evaluate forecast model performance.

3. Sample of results

The season in which the 10-year record of 3-day rain amount fell isshown in Figure 1. The season of maximum climatological rainfall (from thesame 10—year dataset) is shown in the upper panel for reference.

Land areas tend to have their rainy season in the solstice seasons. Inmany areas the peak is in local summer (we will call these monsoon rains), butwinter is the wet season for lands with so-called Mediterranean climates (theMiddle East, California, the southwestern tips of Africa and Australia). A fewinteresting exceptions to summer or winter rain over land include the MAM wet(and severe weather) season in the south-central plains of North America, anda comparable MAM peak in eastern China and south of Japan. Patches of

springtime-peaked rainfall climatology (SON, dark patches in white areas) alsoappear in some southern hemisphere lands (e.g., southeast Australia).

Figure 1. Season of peak rainfall in the 10-year TRMM 3B42 data. a) season of the peak of theclimatological seasonal cycle. b) season of the maximum 3-day rainfall in the 10-year period.

Over the oceans, tropical areas also tend to have monsoon (summer) rainpeaks, shading into autumn because high heat capacity tends to make that theseason of maximum sea surface temperature (SST). In higher latitudes of Fig. 1,winter precipitation predominates in many ocean areas (storm tracks).

The season of the record rainfall event at each point is shown in Fig 1b.Often it coincides with the climatological wet season (Fig. 1a), but naturallythe extremum is a noisier statistic. In tropical cyclone regions (westernsubtropical basins), autumn rainfall records are frequently set, despitesummer-peaked mean rainfall. This autumnal skew of record rainfall events isless common over land.

The characteristic size of heavy rain events is quite small, even for these3-day accumulations. This small patch size can be seen in Fig. 1b, and also Fig.2b, which shows the record rainfall amounts at each location. These fine scalessuggest that mesoscale enhancements to synoptic storms are important aspectsof very heavy rain events everywhere.

a)

b)

DJF

MAMJJA

SON

Figure 2. Precipitation amounts. a) Climatological annual rainfall divided by 12 (mm). b)Maximum 3-day rainfall accumulation occurring in 1998-2007 (mm). c) Ratio b/a.

a)

b)

c) Ratio of the above (record / typical-month)

For comparison, the same gray scale is used for the annual climatology(expressed in mm/month) in Fig. 2a. In wet locations, record 3-day rains areseen to be roughly comparable to ‘a month of rain’ (strictly, 1/12 the annualmean).In climatologically drier locations, dividing the record precipitation bythe average (Fig. 1c) shows that up to a year’s rainfall can happen in a single 3-day event in the most arid regions.

To better glimpse the scale of record-breaking rains, Fig. 3 shows apixelated image (each square or pixel is 1/4 degree) of the year in which theheaviest rain event occurred, for land areas of South and East Asia. Few broadpatterns are evident (except perhaps that early years tended to be wettest inIndochina while late years were wettest over India). The point is just toappreciate the spatial scales involved. Figure 4 shows the same zoomed region,and displays the amount of the heaviest 3-day event at each location, alongwith topography contours. Amounts are patchy, with patch scale notnecessarily tied to the scale of topographic features.

Case finding

As a research tactic, one can use images like Fig. 3 to identify the year,month, day and even central hour of either an especially heavy event in aregion of interest (like choosing a dark patch on Fig. 4), or the record settingevent at a specific location of interest (your city). This date finding exercise isbest done on screen, using a color .pdf file which can be heavily zoomed andstepped between pages showing month, day, year, and amount. This .pdf file isabout 2MB, and is available upon request.

Figure 3. Year of record 3-day rainfall totals over south and east Asia. Square pixel renderingshows the native resolution of the data (1/4 degree) and thus the scale of rain events.

Figure 4. Record 3-day rainfall totals over south and east Asia. Amounts over water weremultiplied by 0.3 for clarity of presentation. Ground elevation contours are 0, 250, 500m (red)and 1, 2, 3, 4, 5km (cyan).

As an example of case finding, Fig. 5 shows the identification of heaviest1-day rain on southern Hispanola island, in the Caribbean. (Color is necessaryto check it in detail.) The record is in May, not in tropical cyclone season.

Fig. 4: Timing of the largest 27h accumulation in the Haiti-Dominican Republic area.

Amounts over waterreduced 70% for clarity

Fig. 5: Satellite image corresponding to Fig. 5 (adding 12h to the central time of the event).http://www.ncdc.noaa.gov/gibbs/image/GOE-12/WV/2004-05-23-18

Delving into satellite imagery archives, this late May storm is seen inwater vapor imagery in Fig. 5 (at lower right corner). Disastrous rainfall at thistime is also confirmed by a news search: one disaster-oriented Web sitehttp://home.att.net/~thehessians/2004monthby.html lists the following entry

MAY 2004. 25th - DOMINICAN REPUBLIC & HAITIFLOODS - 3300 DEAD - 1500 missing

We can use scripts to obtain a sequence of images like Fig. 5, trim them(images are 1200x1200), and assemble animations. Automating this process willallow us to build up a library of case studies. News archives can be consulted toconfirm the events and their impacts, and (for 2007 at least) model-basedforecast fields can be evaluated by delving into the TIGGE archive, availablethrough http://tigge.ucar.edu. Preliminary case studies will be shown at theconference. We hope that such a case study library, based in objectivesampling of long records like this 10-year TRMM product, will combineobjectivity and richness of detail in ways that may be more informative anduseful than either statistics alone, or too few cases selected too randomly.

References

Huffman, G.J., R.F. Adler, D.T. Bolvin, G. Gu, E.J. Nelkin, K.P. Bowman, Y. Hong, E.F.Stocker, D.B. Wolff, 2007: The TRMM Multi-satellite Precipitation Analysis: Quasi-global,multi-year, combined-sensor precipitation estimates at fine scale. J. Hydrometeor, 8(1), 38-55. PDF available at ftp://meso.gsfc.nasa.gov/agnes/huffman/papers/TMPA_jhm_07.pdf.gz