Precipitation Grids - An Explanation
The problems inherent in creating a precipitation time-series:
The vast majority of precipitation data are gathered by volunteer observers. In Minnesota, we benefit from the efforts of over 1400 volunteer precipitation monitors. These individuals provide a valuable public service and the State Climatology Office is grateful for their efforts.
When utilizing a volunteer-based monitoring network to form a long-term data set, one soon discovers that gaps are inherent in the record. Volunteer observers change residences, become unable to make measurements due to illness, or lose interest. Replacing a departing observer with another volunteer in the vicinity can be difficult or even impossible for the network sponsor (i.e. National Weather Service, Soil and Water Conservation District). Therefore, a break in the time-series, or even a termination of the time-series occurs.
Even when a researcher identifies a precipitation monitoring site near their point of interest, they often encounter the data gaps or termination of data mentioned above. If an uninterrupted data series is required, the researcher must then locate the next-nearest monitoring site to fill the gaps and create a data mosaic. If the secondary monitoring site also has gaps in its record, the gap-filling process continues.
What the gridded precipitation data set will do for the researcher:
The gridded precipitation data frees the researcher from the following tasks:
- locating the nearest precipitation monitoring site to the point of interest
- filling gaps in the data record
- locating or calculating historical summary statistics (normal, 30th and 70th percentiles) for the point of interest
How it works ... the precipitation data set:
The gridded database is derived from a monthly precipitation database maintained by the State Climatology Office. Through an act of Congress in 1890, the predecessor to the National Weather Service (NWS) was formed and given the mandate (among other responsibilities) to monitor the climate of the United States. Because a network of professionally staffed weather monitoring sites was economically impractical, the NWS established a network of approximately 200 volunteer weather observers across Minnesota. These observers were well distributed geographically (see map at right), and provided a reasonable depiction of the state’s climate conditions. The network remains in place today and is the backbone for climate monitoring in Minnesota as well as in other states and territories.
While the NWS network offers a large and invaluable data resource, it was long recognized that the spacing between observers is too great to sufficiently describe precipitation patterns formed by isolated thunderstorm activity. Recognizing this shortcoming in the NWS network, farsighted individuals in the early 1970’s formed Minnesota’s High Spatial Density Precipitation Network (HIDEN). This collaborative effort involves many water-sensitive agencies (most notably Soil and Water Conservation Districts), and the combined result is a "network of networks" leading to a precipitation monitoring army of over 1400 volunteers (see map at left).
How it works ... the data gridding process:
To overcome the problems a data gaps across space and time, the State Climatology Office prepares monthly precipitation grids. Grids were prepared using the NWS data from 1891 to 1972. For the period 1973 to the present, the HIDEN data (which includes NWS data) are used. For each month of each year, monthly precipitation totals are estimated for grid nodes at regularly spaced (10 kilometers) intervals (see map at right). The estimates are derived using an interpolation technique called "Kriging", which makes use of the irregularly spaced data in the vicinity of the node to assign it a value. This way, all precipitation data provided by a volunteer observer, be it one month or one hundred months, are fully utilized in the creation of a data time-series. A precipitation total is calculated for every grid node, for every month. There will never be a missing value. Once the grids are created, the calculation of long-term summary statistics such as normals and percentiles can be performed on each grid node.
Possible problems with the gridding process:
No interpolation scheme is without caveats. Some grid nodes are located in sparsely populated areas, or areas without a well developed monitoring network. Obviously, interpolations are most accurate when an array of nearby data exists. The gridding process also tends to "wash out" geographically isolated areas of high or low precipitation. Although the interpolation technique gives greatest weight to the nearest data point, value assignment to a grid node representing an isolated area of high or low precipitation will be influenced by other neighboring data points that may not reflect the small area of dryness or wetness.
Advantages of gridded data for spatial analysis:
One of the advantages of a spatially-complete precipitation coverage is the ability to easily calculate areal averages. By overlaying a boundary (for example, a watershed) upon the grid structure, a researcher can query the values of all grid nodes found within the the polygon, and compute the area average. Example
Using these precipitation grids, researchers have access to a precipitation database that is continuous across time and space. Applications using the gridded monthly precipitation database locate the grid nodes nearest to the user's point of interest and present the time-series interpolated from those nodes. Although the synthetic data will somewhat lack in precision, the database should provide a sound foundation for determining the general precipitation regime experienced at a particular location and period of time.