Climate Indicators

In previous posts we have discussed systematic ways of making sense of local changes to weather patterns. We did this by primarily quantifying anomalous and extreme events in terms of well understood meteorological variables (e.g. temperature, wind, humidity, precipitation) for different climate periods (in the past and future) and estimated their trends.

In this post, we will discuss approaches we have taken to further transform these quantifications into metrics that are tailored to specific domains. The umbrella term ‘climate indicators’ or ‘indicators’  will be used to refer to the industry specific metrics built atop these meteorological variables. 

Global communities and industries who are impacted by weather and climate have developed proxies that estimate outcomes relevant to them. For example, the indicator Cooling Degree Days(CDD) (Weighted Sum of days with Tavg  > 65F) is used in the US (and other temperate countries) as a proxy for power consumption used for cooling. This indicator is dependent on the shape of the distribution of Tavg (mean temperature) and does not merely count the number of days above a certain threshold.  Another example is the indicator Heat Wave Days  that uses the nth percentile of summer overnight temperature (Tmin) as a proxy for increased hospitalizations and mortality amongst vulnerable populations.

In a shifting climate, understanding how climate indicators are changing at a decadal interval will be critical in defining adaptation plans and strategies.

Sample Climate Indicators

HDD and CDD trends shown for New York.

New York, projected heat wave days from 1951 to 2060.

Existing Offerings 

There are many indicators that are relevant across domains and a few governments and emerging startups have started publishing them as part of their emerging climate services. However there are familiar problems and gaps. The obvious gaps are 

Scope: The national (or regional government) climate services cover their regions. For interests spanning regions, it becomes necessary to rely on multiple offerings. Even then, the values of indicators are not comparable as methodologies and source data vary significantly across offerings.

Customizability: It’s hard to customize indicators (e.g. tweak thresholds of HDD to 15C for UK) in the tools offered, and even more impractical if one had to mail in a request to get a new indicator incorporated (e.g. WetBulbDays) into a climate indicator report.

But why aren’t the indicators comparable across offerings? The most common reasons are:

1. Different baseline data for Observations.
   1. Observations  are not  up-to-date and indicators use merely what is readily available.
   2. Baselines used tend to vary (1976-2005, 1991-2020, 1961-1990) and few tools have the flexibility to choose. 
2. Different Methodology 
   1. Calculations based on statistical vs empirical vs ML models.
   2. Transparency levels vary widely. (e.g.  gap prevalence if non-interpolated data sets were used,  assumptions of the statistical models,  lineage of the indicator  - is it derived from one station, gridded average or averaged across nearby stations)
3. Different data for Future Climate Models
   1. The actual CMIP6 model used , bias correction methods and downscaling methods vary across offerings. Though the methods tend to be disclosed, comparing or combining indicators across offerings are tedious or even infeasible.
4. Different definitions of indicators
   1. Indicators that appear to be standard (and canonical) are ever-so-slightly different. e.g. different threshold values for Heating Degree Days (18C, 15C) or different percentile for determining threshold of Heat-Wave Days (p90 or p95)). 

Our Approach

We have attempted to address these concerns by incorporating configurability, customizability and transparency in all aspects of our technical architecture.

Underlying source data models , their quality (completeness of data) are shown or referenced. Moreover, it’s possible to choose the type of source (Observation, Reanalysis and Climate Models) for periods in the past and present. This provides a type of transparency that will bolster confidence in the future models (as well as in  Observations)

1. Baselines used are clearly shown and are configurable to well defined climate epochs.
2. The indicators are of global scope and can be compared across locations. The same methodology applies to all regions.
3. Location specification is flexible and intuitive.
   1. We can either specify coordinates (latitude, longitude) to map the location onto the nearest grid (for sources that are based on climate models) or the nearest station (for sources based on observations).
   2. We can also search by city/metropolitan to identify a city location.  Spatial aggregation expands the analysis to nearby stations or broadens the area selected around the specified coordinates.
4. The indicators are extensible and customizable. 
   1. The effort required to define a new indicator is minimal as the platform on which they are built is extensible. 
   2. The functionality required to define custom indicators (based on configurable thresholds) will be added to the existing tool.
   3. The functionality to define new indicators based on existing meteorological variables is being investigated. At the time of publishing this blog,  requests for particular indicators can be sent to the dev team.

What’s Next

We have received requests to develop indicators based on decades of hourly data such as IDF curves for rainfall intensity and humid-heat based indicators such WetBulb/HeatIndex. We are also planning to add the ability to customize thresholds to the existing page. As always, all of the results that you see on the page can be downloaded as json files. 

 If you want to see a particular indicator or if you want to provide feedback, please drop us a line.

Update (Jan 2024)

We have added Anomaly Days for Max WetBulb Temperature above significant thresholds.

Update (Apr 2024)

We have added a new indicators page that shows annual variation of key metrics like degree days and heat-wave days. The thresholds of these indicators can be altered dynamically, for instance the base temperature of HDD can be set to 15.5C instead of 18.3C (65F). In addition, it is possible to identify trends in the indicators.