barnes

Views:
 
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Slide1: 

By Jennifer Barnes (Supervised by Prof. Mark Saunders)

Slide2: 

Meteorological Agricultural Socioeconomic Hydrological

Slide3: 

Possible to direct aid to specific regions of Africa before drought hits: i) Extra time for Governments and Aid Agencies to prepare – with an indication of severity of drought. ii) Ultimate goal is to reduce the number of deaths from the effects of severe drought: famine, malnutrition, epidemics and population displacements.

Slide4: 

(CRED crunch – Centre for Research on the Epidemiology of Disasters, newsletter)

Slide5: 

(CRED crunch)

Slide6: 

Mean annual rainfall varies considerably over Africa: andlt; 1mm/yr in the Sahara, andgt; 5m/yr in the tropical rainforests (e.g. Guinea coast, Congo and the Great Lakes), andgt; 10m/yr in the mountains (e.g. the Atlas Mountains). Higher diurnal variability in temperature (10-15°C) than interannual (~6 °C). Most of continent has a prolonged dry season (~10-11 months). Strong links between precipitation and global SSTs, high/low pressure and ITCZ. (Nicholson, 2001)

Slide7: 

(http://personal.uncc.edu/lagaro/4100/4100_projects98/AutumnMorgan.html) Mean annual rainfall over Africa (inches). Dark green shows over 80 inches reducing by 20 inches with each shade. Light grey is 10-20 inches and white is under 10 inches.

Slide8: 

Data from the Emergency Disasters Database - CRED (http://www.em-dat.net) Figures exclude deaths from epidemics and extreme temperatures

Slide9: 

Drought Indices have been developed to quantify current drought conditions over past 100 years. 1) Palmer Drought Severity Index (PDSI): uses temperature and precipitation to determine dryness. Developed by Wayne C Palmer in 1965. 2) Crop Moisture Index (CMI): Formula responds rapidly, useful for short term dryness. Developed by W.C. Palmer in 1968. (Heim, 2002)

Slide10: 

(Heim, 2002) Another index is the Standardised Precipitation Index (SPI) (McKee et al., Colorado, 1993), which is probabilistic. This allows flexibility in computing different timescales and can provide early warning of drought. Modern developments: statistical and numerical models; attempt to predict precipitation on a seasonal basis up to a year ahead. Presently we believe there are no models that predict droughts and their severity; hence this PhD is an important extension to the science of drought prediction.

Slide11: 

The Climate Extremes group at UCL have developed a Global Drought Monitor: monitors the strength of drought globally on an ongoing basis. Available free online. This image shows the drought severity based on the precipitation deficit over the prior period of 9 months: Exceptional drought across Madagascar currently affects 671,000 people. (http://drought.mssl.ucl.ac.uk)

Slide12: 


Slide13: 


Slide14: 

Learnt the basics of how to programme in R: a computer statistics system. It has inbuilt commands and the user can write their own functions. It is flexible, simple (once you learn the basics) and has good quality graphical and statistical packages. R is the main programme that I will be using throughout my PhD so these 3 months of practice have been invaluable in building the foundations of my use of R.

Slide15: 

1) Understanding how to get the precipitation data the right way up when dealing with millions of data points in one array (data will be discussed later). 2) Making sure that, when dealing with two datasets, the grids matched in resolution and there was no offset in the latitudes/longitudes. 3) I have dealt with NetCDF, GRIB and text files over the last 3 month; all of which required very individual approaches in programming for: - the collection of data - the opening of the files in R - dealing with the data - getting it into a suitable - manipulating the data format (array) in R to get results. 4) Applying land masks to data to see the outlines of the countries!

Slide16: 

Moved on to studying the correlation between ENSO and global precipitation. This meant correlating the sea-surface temperatures (SSTs) over specific areas of the Pacific with global precip. (NOAA website) Then looked at the correlation of the Nino4 index (average SST anomaly in Nino 4 region) with global precipitation using two different datasets. I will show you the main results in a moment.

These maps are from the Climate Prediction Center. They show the observed links between precip and ENSO. These include a positive correlation between precip and ENSO over central Africa and California to Florida; with negative correlation over S.Africa, S.Australia, Indonesia and Brazil. The maps were used to make sure the data were being handled correctly.: 

These maps are from the Climate Prediction Center. They show the observed links between precip and ENSO. These include a positive correlation between precip and ENSO over central Africa and California to Florida; with negative correlation over S.Africa, S.Australia, Indonesia and Brazil. The maps were used to make sure the data were being handled correctly. (Climate Prediction Center website)

Slide18: 

Negative correlation over parts of Africa, Indonesia, Australia and Brazil Positive correlation over California and Florida

Slide19: 

I have been using 2 main datasets so far: 1) NCEP (National Centres for Environmental Prediction) data are from a reanalysis project – the precipitation data are the result of a numerical forecast so there are potential deficiencies in the data (Janowiak, 1998). 2) GPCC (Global Precipitation Climatology Centre) data are based on monthly mean estimates of precip from gauge observations and satellite derived precipitation estimates. These have been carefully compiled and quality controlled and are used for validation of numerical models, including DEMETER. 3) A third data set is ERA-40, which describes the climate between 1958 and 2001. It is believed to be the best and most time-consistent representation of the climate for the globe as a whole, however there will be regional difficulties in the accuracy of data as with all reanalysis projects including GPCC. The data for ERA-40 come from a wide selection of sources. (http://www.ecmwf.int/research/era/index.html)

Slide20: 

Good positive correlation over the developed world between GPCC, NCEP and ERA-40 precipitation data over 43 year period. Negative or low correlation appears to be mainly over areas of low data and areas of extremely high or low precipitation.

Slide21: 

The first stage of my PhD programme is to assess the long-range forecast skill for drought of the DEMETER model, which provides globally gridded ensemble forecasts of precipitation out to 6 months ahead. DEMETER is the leading dynamical seasonal prediction system in the world. These data have not yet been examined to quantify drought prediction therefore it is important to assess their skill. The skillful outlooks will then be incorporated into UCL’s drought monitor to extend this product to benefit humanitarian relief.

Slide22: 

'Development of a European Multimodel Ensemble system for seasonal to inTERannual prediction'. 7 independent GCMs from across Europe (this reduces model error). Each has 9 different SST initial conditions (this reduces error due to chaos theory). DEMETER has 7x9=63 members. Period 1957-2001, but only 1980-2001 for all members. I will study 3 main models: ECMWF, Meteo France and UK Met Office.

Slide23: 

Assess the skill of the forecasts generated by DEMETER by comparing the hindcasts to historical precipitation records. Ongoing issue of which data set is best suited for assessing the skill of DEMETER over Africa needs to be approached – I will attempt to gain access to observed data sets as well as using the ERA-40 data (used to force the lower boundary SST conditions in DEMETER) and GPCC data set in order to assess the hindcasts from DEMETER . There are many statistical methods that I need to consider in this process (e.g. removing bias drift in hindcasts). Skill of Multimodel ensemble is thought to be andgt; single model ensemble as has better sampling of forecast uncertainty and has verification. This will be tested during the process of assessing the skill of DEMETER and its component models separately. (Hagedorn, R. et al., 2005 and Palmer, T.N. et al., 2004)

Slide24: 

2) Development of statistical probabilistic drought prediction models trained on historical climate data. These will be developed for individual countries with priority being sub-Saharan and eastern Africa. The predictive capability for drought of ENSO, SSTs and anomalies in atmospheric circulation will be quantified.

Slide25: 

3) Compare and combine the statistical forecasts with seasonal precipitation forecasts from the DEMETER project to consolidate skill through the ensemble-averaging of independent forecasts to remove noise. 4) Tune skillful drought predictions to meet the specific needs of humanitarian relief.

Slide26: 

CRED crunch newsletter http://www.uta.edu/depken/data/credcrunch/droughts.pdf (last accessed:22/01/07). Climate Prediction Center http://www.srh.noaa.gov/ssd/html/elnino.htm (last accessed:20/01/07). Global Drought Monitor website http://drought.mssl.ucl.ac.uk (last accessed:23/01/07). Hagedorn, R. et al. 2005 The rationale behind the success of multi-model ensembles in seasonal forecasting – 1. Basic concept. Tellus, (57A):219- 233. Heim, R.R. 2002 A review of twentieth-century drought indices used in the United States. Bulletin of the American Meteorological Society, (83):1149-1165. Janowiak, J.E. 1998 A comparison of the NCP-NCAR reanalysis precipitation and the GPCP rain gauge-satellite combined dataset with observational error considerations. Journal of Climate, (11):2960-2979. Nicholson, S.E. 2000 The nature of rainfall variability over Africa on time scales of decades to millenia. Global and Planetary Change (26):127-158. Nicholson, S.E. 2001 Climatic and environmental change in Africa during the last 2 centuries. Climate Research, (17):123-144. NOAA website http://www.cdc.noaa.gov/forecast1/images/nino4region.gif (last accessed:23/01/07). Palmer, T.N. et al. 2004 Development of a European multi-model ensemble system for seasonal to interannual prediction (DEMETER). Bulletin of the American Meteorological Society, June.

authorStream Live Help