The Drought Monitor was introduced as an operational weekly product in 1999 to provide an overview of conditions averaged across a broad array of time scales and impact indicators, leaning toward those that seem most relevant to observed impacts. This approach has led to an unprecedented degree of cooperation and coordination among a variety of disparate Federal, state, and local government agencies, in addition to many interested members of the academic and private research communities. The result has boiled the complex issues of drought and drought-related impact assessment down to a single, simple, visually-intuitive summary of conditions which has replaced the uncoordinated, disparate, and often contradictory assortment of opinions and data that formerly characterized responses to requests for drought information.
While this approach has been successful and well-received overall, there are situations where it can be substantially misleading. Drought and its related impacts operate on a variety of time scales, and the Drought Monitor depiction (which usually portrays some semblance of an "average" condition across all time scales and impact types) cannot accurately confer information when conditions and impacts dependent on one time scale differ dramatically from those related to a much longer (or shorter) time scale. Hypothetically, a region which has received consistently and substantially inadequate precipitation over the course of several years might experience a day, or a few days, or even a few weeks of heavy rain. What is the overall drought status after this occurs? The Drought Monitor would likely depict a substantial improvement in conditions (in deference to major short-term relief) but maintain some indication of continuing drought (in deference to the multi-year dryness which likely changed only slightly in response to the heavy rains). This is all that a single-image depiction could possibly do. In reality, however, the degree to which drought-related impacts would continue to be a concern would depend on what time scale a given class of impacts responds to. Obviously, in this situation, wildfire danger would decline sharply, at least for the immediate future. Also, unregulated streamflows would swell from runoff and topsoil moisture would be substantially recharged if the precipitation lasted long enough, thereby providing at least a temporary respite for non-irrigated agriculture. On the other hand, reservoir stores might increase only slightly, having been depleted by a few years of precipitation failing to keep up with demand, and ground water levels and/or well water depth, if they were low, might be barely (or at best belatedly) affected by the heavy short-term rains, since much of the water was likely dispersed by swollen streams or absorbed by parched topsoil.
To confer information about drought status on different time scales to those users that need such information, two new experimental products are being made public which will serve as timescale-specific supplements to the Drought Monitor at a basic level. Both assess conditions based on a blend of several drought indicators, and are depicted relative to the local historic record.
The Short-Term Blend approximates drought-related impacts that respond to precipitation (and secondarily other factors) on time scales ranging from a few days to a few months, such as wildfire danger, non-irrigated agriculture, topsoil moisture, range and pasture conditions, and unregulated streamflows.
The Long-Term Blend approximates drought-related impacts that respond to precipitation on time scales ranging from several months to a few years, such as reservoir stores, irrigated agriculture, groundwater levels, and well water depth.
It should be noted that the relationship between indicators and impacts varies, sometimes markedly, with location and season. This is particularly true of water supplies, which are additionally dependent on the source (or sources) tapped, management practices, and legal mandates. Exercise caution when attempting to relate these maps to specific impact implications for a particular location and time of year. The blend-to-impact correlation is not always direct, and will vary spatially and temporally.
The following bullets describe the composition of these experimental blends:
These products are generated using the Climate Prediction Center's real-time daily & weekly climate division data, and the National Climatic Data Center's monthly climate division data archive, back to 1932.
The indices used in the blends and their weights are as follows:
SHORT-TERM: 35% Palmer Z-Index; 25% 3-Month Precipitation; 20% 1-Month Precipitation; 13% Climate Prediction Center Soil Moisture Model; and 7% Palmer (Modified) Drought Index.
LONG-TERM: 25% Palmer Hydrologic Drought Index; 20% 12-Month Precipitation; 20% 24-Month Precipitation; 15% 6-Month Precipitation; 10% 60-Month Precipitation; 10% Climate Prediction Center Soil Moisture Model.
All parameters are first rendered as percentiles with respect to 1932-2000 data using a percent rank method. Most parameters are ranked relative to the National Climatic Data Center's historic climate division data for the current month, except for the Z-Index which is rendered relative to all months on record (this introduces evaporative seasonality into the short-term blend).
For each blend, the averages of the percentile inputs are calculated, with each input weighted as described above. This yields a "weighted raw average" of the individual component percentiles for each blend. Then, each raw average is compared to its historic (1932 - 2000) distribution (these have been retrospectively generated from the climate division data archive). The real-time data are compared to ALL retrospective months, not just the current month, since the individual percentile inputs were each generated (for all but the Z-Index) relative to the history of the current month only. This allows for a more confident estimation of the percentile by using more data to define the historical array (12 times as many as if we assessed the blends' raw weighted averages relative to the current month only).
The precipitation percentile inputs are generated in a somewhat unusual way, combining month-to-date numbers from Climate Prediction Center with the National Climatic Data Center's monthly totals for prior months. As daily precipitation totals for the current month are ingested into the x-month totals, an identical proportion of the monthly precipitation that fell during the first month in the x-month period is eliminated (e.g., to determine a 6-month preciptation total, from which a percentile will be calculated and incorporated into the blend, for the period ending September 21, 2002, we add the daily preliminary precipitation amounts for September 1-21 to the 6-month total for March-August 2002, then subtract 21/30 of the March total from the result, since 21/30 of September have been added). This process (a) emulates natural cycles by adding precipitation as it falls but eliminating early-period precipitation evenly over the course of a month; and (b) ensures that the data utilized in real time are as consistent with the historical array as possible. The near-real-time climate division precipitation data are biased in some areas relative to the final NCDC monthly archive, with wet near-real-time biases in the central and northern Rockies particularly extreme. The data are adjusted where appropriate at the end of each month, but the biases remain in the data for all precipitation time scales since the end of the previous calendar month. In addition, the biased near-real-time data are used in the Palmer Drought Index, the Palmer Hydrologic Drought Index, the Z-Index, and CPC's modeled soil moisture data, and can remain in those calculations for several weeks. For more information, a graphic is available depicting the observed biases (a) for the 12 most recent 3-month periods, and (b) for the last 3 years in each season.
These blends may be subject to change in the future. Feel free to forward any questions or comments to the Drought Monitor authors for consideration or response.