Weather vs. Climate Data
With the power and influence of the mutual orientation of scientists, military, and businesses on accurate forecasting of weather, climate scientists have a long history of frustration with the overwhelming focus of the global observing system on forecasting. Many of them feel that operational agencies mostly ignored climatology’s needs until quite recently, managing instruments and handling data in ways detrimental to the long-term climate record. (Edwards 2010:287). As Edwards discussed in the previous chapter, systems and standards change over time, creating temporal gaps and data inconsistencies; this makes the data unusable for climate study. What needs to be done is data ‘clean-up.’
“Reanalysis of global weather data is producing-for the first time-consistent, gridded data on the planetary circulation, over periods of 50 years or more, at resolutions much higher than those achieved with traditional climatological data sets. Reanalysis may never replace traditional climate data, since serious concerns remain about how assimilation models “bias” data when integrated over very long periods. Nonetheless, the weather and climate data infrastructures are now inextricably linked by the “models of data” each of these infrastructures requires in order to project the atmosphere’s future and to know its past” (Edwards 2010:288).
Differences between weather and climate data: (See Table 11.1, 2010:293-294)
- Data Sources
- Data Quality
- Computation centralization
- Data Preservation
For forecasting the weather, the goal is rapid and accurate prediction. For prediction, data needs to be quickly accessed for analysis, with well-distributed coverage. The Global Observing System (satellites, buoys, planes and ships, ground stations) place a low priority on preserving raw data (which is necessary for use in climate analysis). For climatology, the emphasis is on determining patterns and trends. Data needs to be in long-term and stable data series. Data such as precipitation and paleoclimatic proxies are not relevant to forecasting. As a result, only some weather-station data were useable for climate study.
Over time, there is constant change that affects climatology more than meteorology. For example, changing instruments, locations, and practices all affect the data itself – its reliability (degree to which tool produces stable and consistent results) and validity (how well a tool measures what it is supposed to measure).Changes in how data is gathered, as well as changes in the landscape itself can also impact the quality of data. This is part of the issue that Edwards describes in terms of synchronic vs. temporal calibration.
Changing technologies (especially satellite, computers) reduced data and computational friction: “if you want global data, you have to make them. You do that by inverting the infrastructure, recovering metadata, and using models and algorithms to blend and smooth out diverse, heterogeneous data that are unevenly distributed in space and time” (Edwards 2010:321).
The limited success of the do-over
(This particular effort is a good example of “citizen science.”)
Rebuilding climate data from original weather data needs adjusting for the issues that skew data, as in the example from above. This is what Edwards calls re-analysis. Edwards writes: “reanalysis revealed the depth and intractability of problems in both the observing system and the models. The observing system’s highly heterogeneous instrument types, models, and manufacturers create numerous ongoing disparities in the input data” (2010:331). More specifically:
(From Edwards 2010:333)
This graph is showing the impact of reanalysis in generating data. ERA 40 (from the European Center for Medium-Range Weather Forecasts) is second-generation re-analysis using higher-resolution systems for global atmospheric reanalysis. NCEP reanalysis (National Centers for Environmental Prediction, NOAA), was “designed and executed as a global, high resolution, coupled atmosphere-ocean-land surface-sea ice system to provide the best estimate of the state of these coupled domains over this period” https://reanalyses.org/atmosphere/overview-current-reanalyses . The Climactic Research Unit dataset is considered one of the most authoritative collection of climate data. In the diagram above, I’ve added a red line to indicate the 1979 incorporation of satellite data observations. What Edwards is saying here is that while promising, reanalysis thus far cannot replace traditional climate data. It is also hugely expensive and time-consuming. However, if the focus is determining the extent of anthropogenic climate change, then re-analysis is the only way to go.
So reanalysis has been integrated into the infrastructure of climate knowledge. Each round will bring new revisions to the history of climate. Well into the future, we will keep right on reanalyzing the past: more global data images, more versions of the atmosphere, all shimmering within a relatively narrow band yet never settling on a single definitive line. (Edwards 2010:336)