Workshop Report – The DRI Precipitation and Drought Indices Workshop

There were two well-attended break-out sessions to discuss the key issues that arise in the development and applications of  precipitation and drought indices datasets.  The key recommendations from the discussions are summarized in the following. 

1.  Precipitation dataset issues 

Overall, it was felt that the users could usually get access to the precipitation datasets that they need but it is quite often through appropriate personal contacts rather than through well advertised public channels or data portals. The QC and additional processing of the source data as well as the limitations of the final datasets certainly need to be better documented. The development of cross-border datasets should be first attempted at smaller scale projects while the development of better Canadian datasets should be included as part of the next big research network program. A summary of the discussions at the break-out session is given in Annex 3 and the following recommendations are proposed. 

• A short document describing currently available precipitation and drought indices datasets including a metadata summary, information about observational errors or the quality of the data, and details on how the data can be accessed and acquired. 
  

• EC Management should be strongly encouraged by DRI and others to properly resource its data management program.

• DRI researchers should develop plans for the follow-up project which among its objectives would encourage the development of a comprehensive Canadian data set.

• EC should prepare clear documentation for the QCs and corrections made to datasets that are maintained in their archives.

• DRI should address the need for more cross-validation of different precipitation datasets (may be as part of a Canada-wide water and energy budget study) in its follow-up project. 
  

• Links to EC (and non-EC) precipitation and drought indices datasets  should be put on the DRI website and DAI. 
  

• The US-CANADA GEO committee should launch cross-border test bed studies to assess different approaches to cross border data sharing (e.g. Red River Valley, recent flooding). 
  

• At REDE workshop in May, someone should report on the recommendations of this workshop. 

• At end of DRI, DRI scientists should produce an assessment of how precipitation over the Prairies have changed over time.   

2.  Drought Indices dataset issues 

Overall, it was felt that the current available drought indices such as SPI and PDSI are suitable for some applications, but are inadequate for others. From the break-out session, the following recommendations are proposed. 

• More research is required that links the various drought indices to specific impacts (e.g. crop yields, streamflow, etc.) to determine the most suitable index for the impact in question. 
  

• A careful systematic comparative analysis of the various drought indices over a long period of time and different geographical regions is required. 
  

• All indices have uncertainties and these need to be documented. 
  

• Many of the drought indices datasets used for particular studies are derived by the study itself. It is difficult to access specific long-term, consistently derived indices for scientific, operational, and monitoring purposes. 
  

• Easier access to LDAS daily products is required. 
  

• A huge problem involves the ability to obtain drought indices in real time for operational and monitoring requirements (unlike the US where there is free and open access). 
  

• There are huge spatial and temporal differences between drought indices that are calculated using observed data versus gridded data versus modeled data. Research is required to determine the pros/cons and differences among the various input methods. 
  

• Recommended that historical/real-time drought indices be merged with NWP forecast products (e.g. GEM) to provide drought forecasts of up to one month. This could be achieved through collaborations of operational (e.g. Agriculture Canada) with EC and DRI. As a first step, this could be tested as an experimental product to determine its feasibility. 
  

• We need to rethink the current drought indices such as SPI and PDSI that reflect precipitation variability over long periods (i.e., month or longer). Several sectors including agriculture are impacted more by precipitation events (e.g., intense 1-day rainfalls, prolonged dry spells at critical periods for crop growth, etc.). It is therefore recommended that additional drought indices be incorporated that reflect these precipitation events using daily data. 
  

• Statistical downscaling of the aforementioned precipitation events from GCMs is not reliable. NWP models and RCMs are needed for these event-based analyses. 
  

• Recommended that over the next few years, we work toward a monthly water cycle monitoring program that links the atmospheric and land surface communities. This could be part of the Canada-wide water and energy budget study proposed earlier and a combined government and university initiative. 
  

Annex 1 – Pre-registered Workshop Attendees

Annex 2 – Workshop Program 

Welcome and Opening Remarks

09:00 Kit Szeto                     Welcome, overview of precipitation dataset and drought indices issues and workshop objectives

                                        09:15  Ron Stewart                DRI overview and updates 

Precipitation and Associated Derived Datasets  

09:30  Xiaolan Wang             Homogenization of station data and development of blended precipitation dataset

                                        09:50  Eva Mekis                   Corrected station precipitation archive and derived datasets

                                        10:05  Ewa Milewska             Gridded temperature and precipitation datasets for Canada 

                                        10:20-10:45                            Break (coffee served outside the Auditorium) 

                                        10:45  Marco Carrera             Development of the Canadian Precipitation Analysis (CaPA)  and Canadian Land Data Assimilation System (CaLDAS)

                                        11:05  Rick Lawford              Global precipitation datasets 

                                        11:20  Phillip Harder and Patrice Constanza                 DRI datasets and legacy dataset development plan   

                                        11:35-12:45                            Lunch (provided, MSC cafeteria) 

Drought Indices and Applications   

                                        12:45  Xuebin Zhang             Climate extreme indices

13:00  Richard Heim             Drought and climate extremes indices for the North American Drought Monitor and North America Climate Extremes Monitoring System

13:20  Eric Wood                 Land data assimilation products and their applications in drought monitoring and forecast

13:40  Lei Wen                     Real time drought monitoring and forecasting over the Canadian Prairies using the VIC model

                                        13:55  Amir Shabbar             On the genesis of prolonged drought in Canada

                                        14:10  Kit Szeto                    Trends of Prairie drought  

14:25  Philippe Gachon          Indices of droughts (SPI & PDSI) over Canada as simulated by a statistical downscaling model: current and future periods    

14:40  Sharon Fernandez       Canadian Drought Alert and Monitoring Program (CDAMP)

14:55  Alison Meinert            Capturing the dynamics of the 1999-2005 Canadian Prairie drought using the SPI and PDSI 

15:10 Kit Szeto                     Issues for discussions at break-outs   

                                        15:15-15:30                          Break (coffee and snacks at MSC cafeteria) 

Break-outs and Discussions

                                        15:30- 16:30                         Breakout sessions (MSC cafeteria)

                                                                                     1. Precipitation dataset issues

                                             2.  Drought indices issues

                                        16:30                                    Summary from breakout sessions and overall discussions

                                        17:30                                     Workshop summary and next steps 
 

Annex 3 – Summary of Discussions at the Precipitation Issues Break-out Session 

1. Do currently available precipitation datasets meet your requirements?

COMMENTS AND DISCUSSION: There was a general view that the precipitation data sets available on the web are not adequately adjusted.  Stations are disappearing, quality is questionable, qc not properly done. Given the large number of data sets it would be good to separate the data sets with current information from those were entirely historical.  There is a need to ensure that data being entered into a national archive should be adjusted (with a minimum time lag).  Modellers who want to assimilate precipitation into their models, would like to see observational error estimates provided with the data that they assimilate into their real-time and historical reanalysis.

  Precipitation rates are important and the data need tol be more effectively disseminated.   (Rainfall rates for 5min, 1min, 1hr, 3hr, 6hr, 12hr, 24hr are archived for tipping bucket however these data are not adjusted and the information is not well known).

With automation the manual snow measurements have disappeared.  Snow data are in disarray and special attention needs to be paid to these data.  Issues such as distinguishing between snow and rain are no longer effectively addressed.  Furthermore, observational errors are not documented.

The archiving of radar data is not well structured.  It is believed that the archiving of Canadian radar data began in 2005 at NCDC.  Some regions (e.g., Western and Northern region) have regional archives of radar data on disks although this varies regionally.  Canada has a national radar program but none of the participant knew how this activity addressed radar data archiving issues.

Precipitation data from satellite is not archived in Canada and users go to NOAA or NASA archives when they need these data.

RECOMMENDATION: Environment Canada should prepare a short document describing all the datasets in the archive including a metadata summary, information about observational errors or the quality of the data, and details on how the data can be accessed and acquired.  

2. Are the inherent limitations of currently available precipitation datasets well-documented?

COMMENTS AND DISCUSSION: No, they aren’t.  Two processes are occurring:  regional climate experts are disappearing (i.e., corporate knowledge is disappearing) and archiving facilities are paperless. However, non-computerized information is still flowing into the data centres and will not be available unless they are keyed in.  There is strong evidence that EC is not committing enough resources to get access to data from because the data are not being keyed in.  Furthermore, the metadata for many stations are not being kept up to date.  With more resources for these activities in Environment Canada, these problems could be solved.  Participants agreed that volunteer stations and long valuable historical stations) should not be closed. It was also noted that many provinces have their own networks but these data are relatively unknown and often hard to access.  There was some interest in having federal, provincial and private sector data integrated into a large national data base,  However, we would need to address the different data policies and instrument standards that used in different jurisdictions.  .

RECOMMENDATIONS:

• EC Management should be strongly encouraged by DRI and others to properly resource its data management program.
• DRI researchers should develop plans for large project which among its objectives would encourage the development of a comprehensive Canadian data set.

3.  Do you have any problem in finding and accessing the precipitation datasets that are needed for your work? Assuming the data exist digitally, can you find & access the data?

COMMENTS AND DISCUSSION: Users generally find the EC precipitation data on the web but they do not know whether the data are adjusted or they have errors, nor do they have access to adequate metadata about the data sets.  In order to deal with this issue researchers want research quality data sets that are accurate and complete. Details of the data need to be well documented (so researchers don’t mix adjusted and unadjusted data). Since there does not appear to be a single EC person responsible for the quality control of these data, researchers are forced to rely on personal acquaintances with appropriate EC staff to get their data quality controlled.   

4.  Are there any specific precipitation data issues for drought research in Canada?  (winter precipitation sparse network, orographic precipitation, etc.)

COMMENTS AND DISCUSSION: Changes in small (trace) precipitation amounts over time will be important in detecting the effects of climate change.  (However, changes in the definition of trace amounts and techniques for measuring minimum precipitation over the years could introduce erroneous long-term trends).  It is important to seek the same limits and precision across the provincial and US-Canada borders.  Precipitation rates (including drop size distribution) is important for drought research especially in the study of virga (It is important to understand virga because these processes affect the amount of precipitation reaching the ground).  They are also important for feedbacks such as the effects of preceding storms with light precipitation that wet the ground on subsequent flood or drought-ending precipitation events.

  Precipitation information is also required for models.  NWP models need high temporal resolution, near-real time data inputs and require reliable land surface schemes. Drought research has emphasized the need for: 1) longevity of the datasets (many stations have short station records – more stations need to be maintained over long periods), and 2) the spatial resolution of precipitation datasets. Even in 2009, some records only exist in paper format and continue to be recorded in this format.  Data rescue efforts need to be undertaken on an urgent basis for perishable paper records of earlier precipitation records.    

RECOMMENDATION: EC should prepare clear documentation for corrections made to datasets that are maintained in their archives. 

5. Will regular cross-validations between the different precipitation datasets over Canadian regions be useful?  Who should do it?

COMMENTS AND DISCUSSION: Yes, cross-validations would be useful.  Environment Canada should lead because they have all the data (including radar data, in situ data, satellite data, and models).

RECOMMENDATION: DRI should address the need for more cross-validation at its next workshop. 

6. What are the main issues in the development of cross-border datasets?  How could DRI help?

COMMENTS AND DISCUSSION: Issues that need to be addressed in establishing a signle Spatial density of data & length of record, way data are taken (instrumentation & observational methods & variables measured esp. snow measurements, rain gauges, gauge shielding, correction methods different; radar S band US, C band CN).

RECOMMENDATION:  The US-CANADA GEO committee should launch cross-border test bed studies to assess different approaches to cross border data sharing (e.g. Red River Valley, recent flooding). 

7.  Do we have adequate (both modeled and observed) precipitation data for regional climate change and projection studies?  What are the specific issues?  Would RCMs help?

COMMENTS AND DISCUSSION: In general, there was agreement that there may adequate data for regional climate change assessments in some areas such as southwestern Ontario where data density is high and a number of long-term stations exist.  However, over most of Canada station density is inadequate to address regional climate issues.  RCM’s have a role to play in downscaling and could be of use in interpolating between stations but generally they are not adequate in data sparse areas because the error propagation from the boundaries may be larger than the benefits of the higher resolution physics that they incorporate.  The group recognized that these comments were based on a general understanding of the capabilities of RCMs but would need advice from both RCM and GCM modelers before being able to answer this question authoritatively. 

8.  How could DRI and partners collaborate better to address these issues?

COMMENTS AND DISCUSSION: Have workshop to highlight human impact?

RECOMMENDATIONS:

1. At REDE workshop next month, someone should report on the recommendations of this workshop. 
2. At end of DRI, DRI scientists should produce an assessment of how precipitation over the Prairies have changed over time.