Environment and Climate Analytics

Class 4

K. R. Hari Krishna
M2022BSASS020

Climate Data
Climate Data is data collected over an extended period concerning various atmospheric
conditions. It provides insights into the long-term trends and patterns of weather in specific
regions or globally. Climate data is necessary to understand climate change, its impacts on
ecosystems and societies, and adaptation and mitigation strategies.

Climate Data is Geospatial Data, which is data that includes information related to locations
on the Earth’s surface and can map objects, events, and other real-world phenomena to a
specific geographical area identified by latitude and longitude coordinates. Geospatial data
combines location information with characteristics or attributes of other business datasets
over a given period. For example, geospatial data can represent where a car is parked and
also track its location when it starts moving.

In reality Climate Data is recorded from Earth, in shape of a Geoid, which is the shape that
the ocean surface would take under the influence of the gravity of Earth, including
gravitational attraction and Earth's rotation, if other influences such as winds and tides were
absent. This surface is extended through the continents. If the geoid really existed, the
surface of the Earth would be equal to a level in between the high-tide and low-tide marks.

Although a geoid may seem to be a smooth, regular shape, it isn't. The Earth's mass is
unevenly distributed, meaning that certain areas of the planet experience more
gravitational "pull" than others. Because of these variations in gravitational force, the
"height" of different parts of the geoid is always changing, moving up and down in response
to gravity. The geoidal surface is an irregular shape with a wavy appearance, thus there are
rises in some areas and dips in others. While theoretically possible, actually measuring this
surface is impossible because of the huge amount of data needed to process this shape and
the variability of the surface. Instead for Climate Modelling, geodesists use a Spherical or
Ellipsoid shape is instead used in the place of a geoid model as the next best solution. The
ellipsoid is created by rotating an ellipse around its shorter axis. This matches the real
Earth's shape, because the earth is slightly flattened at the poles and bulges at the equator.
When converting from a three-dimensional model to a two-dimensional surface, we use
projection techniques. A map projection is a method for taking the curved surface of the
earth and displaying it on something flat, like a computer screen or a piece of paper. Map
makers have devised methods for taking points on the curved surface of the earth and
"projecting" them onto a flat surface. These methods allow map makers to control the
distortion that results from creating a flat map of the round earth.

Every map projection has some distortion. Equal area projections attempt to show regions
that are the same size on the Earth the same size on the map but may distort the shape.
Conformal projections favour the shape of features on the map but may distort the size.
Two-Dimensional Projection methods may be classified into: Cylindrical, Pseudocylindrical,
Conical, Pseudoconical, Azimuthal and Polyhedral. Some examples are given below.

As we can see there are differences in how Gall-Peters and Mercator projections are
constructed, with each trying to show different areas in a perfect form even if they are the
same type of projection. Using these projections, we can also map our co-ordinates into the
models and maps. But maps can be also made based on the context and usage. Pilots use
directional maps, while casual users and geospatial analysts use more of relative map for
our purposes, leading to the development of various types of maps.

In climate data analysis, several special characteristics are included for understanding and
interpreting the information effectively. These include coordinates, elevation, data values,
and temporal elements. Coordinates provide geospatial locations, allowing one to pinpoint
specific areas on the Earth's surface. Elevation denotes the height and this can be used as
another positional variable for the three-dimensional data. The data value represents the
measurement units, such as temperature, pressure or rainfall, which are the primary
characteristic of climate data, used for analysis. Additionally, the temporal element, is
included using time intervals or timestamps, is useful for tracking changes over periods and
doing trend analysis and forecasting. To ensure the reliability and integrity of the data (data
quality), metadata serves as a data flag, giving information about the methods and tools
used for measurement, locational conditions, and other relevant details. This metadata is
useful as it improves the quality and credibility of the climate data by giving context to the
scenarios of measurement into the context and circumstances surrounding its collection.]

Example of Data from NOAA on Guwahati, India containing coordinates, data values, and temporal elements
along with attributes for showing metadata.
Locational Data

Locational data in climate observation means information tied to specific places on Earth.
There are two main types:

1. Point Location Data: This refers to data collected at specific observation points, such
as weather stations or observatories. These points gather data on various climatic
characteristics like temperature, humidity, wind speed, and more. However, certain
data, like rainfall, is considered area data, meaning it covers a broader geographical
region rather than being specific to a single point, but still are collected in a point
format as area data is harder to collect.

2. Area Data: Area data includes information for a wider geographic area rather than
being confined to a single point. Gathering area data can be harder compared to
point location data, as it involves combining measurements from multiple sources to
provide a view of climatic condition for a parameter over a specified region.
Examples include rainfall distribution, cloud cover, and temperature across a
landscape.

Rainfall is measured as an area data but is recorded per point and methods to record are
also point wise, thus having many rain gauge stations, like those in Santa Cruz and Colaba,
gives us a better understanding of rainfall patterns across different areas. For example,
during the 2005 floods, when Mumbai recorded 930 mm of rainfall in a day, there was much
scepticism toward the figure, as there were only 2 measuring stations, which was
understandable, as even in Chennai, the annual rainfall is around 800 mm which is lesser
than the figure recorded at Mumbai. But there were many questions over the relevancy of
data, as it only included two points of measurement, which lead to the development of
more stations, giving a more reliable picture of rain in Mumbai. When it comes to area data
in climate observation, one common method used is interpolation and extrapolation of
point data.

1. Interpolation: This involves estimating values for locations where data isn't directly
measured by using information from nearby points where data is available. It's like
filling in the gaps between known points. For example, if we have weather stations in
certain areas but not everywhere, interpolation helps us estimate what the weather
might be like in between those stations. Different mathematical techniques, like
linear interpolation, can be used for this purpose.

2. Extrapolation: This extends beyond the known data points to make predictions
about conditions in areas where no data is available. It's like making educated
guesses based on trends observed in the existing data. For instance, if we have data
for a certain region but need to predict conditions in nearby areas where no
measurements exist, extrapolation can help us make those predictions. However,
extrapolation comes with more uncertainty compared to interpolation because it
involves projecting beyond the observed data range.
Georeferencing
Georeferencing means that the internal coordinate system of a digital map or aerial photo
can be related to a ground system of geographic coordinates. A georeferenced digital map
or image has been tied to a known Earth coordinate system, so users can determine where
every point on the map or aerial photo is located on the Earth's surface.

The relevant coordinate transforms are typically stored within the image file (GeoPDF and
GeoTIFF are examples of georeferenced file formats), though there are many possible
mechanisms for implementing georeferencing. Georeferencing in the digital file allows basic
map analysis to be done, such as pointing and clicking on the map to determine the
coordinates of a point, to calculate distances and areas, and to determine other information
useful for analysis.

Simplified Steps to do Georeferencing: scan data → choose points with locational co-ordinates to help
computer know the structure and pixel locations → use GCP to map important data and conduct analysis.

When estimating rainfall in areas where direct measurements are unavailable, two common
methods used are Nearest Neighbour and Inverse Distance Weighting (IDW) 1.

Now let us consider an example. We have cities A and B with all geolocational co-ordinates
mapped. Now A has Rainfall of R 1 which is known, while B has rainfall R 2, which is unknown.
Now to calculate the Rainfall of B, we can assume that the amount of Rainfall might be
changing, given below.

1
Further Readings: Fujita, Thais & Morais, Marcos & Monteiro, José Alberto & Santos, Vanessa & Rudke,
Anderson & Abou Rafee, Sameh & Martins, Leila & Souza, Rodrigo & Freitas, Edmilson & Martins, Jorge.
(2018). Nearest Neighbour and Inverse Distance Weighting for Rainfall Estimation in SWAT Application.
https://www.researchgate.net/publication/
328102200_Nearest_Neighbor_and_Inverse_Distance_Weighting_for_Rainfall_Estimation_in_SWAT_Applicati
on
Note that ΔR can be positive, negative or zero and thus has the flexibility to use assumptions
to prove for climate data analysis. An adaptation of this can be done in extrapolation where
people calculate the rainfall at a certain point using the nearby points. This can be called the
nearest point calculation.

This method simply assigns the value of the nearest measured point to the location being
estimated. It assumes that nearby locations will experience similar rainfall conditions. While
easy to implement, this method may not always provide accurate estimates, especially if
there is significant variability in rainfall over short distances. For an example below with 5
points A,B,C,D,E with X as the value to be found,
This is an example of how the closest points influence the climatic conditions of X and it
makes sense. But when we also want to see how the rest of the points influence X, we need
to explore Inverse Distance Method.

IDW calculates estimates based on the distances and values of nearby measured points.
Points closer to the target location are given more weight than those farther away, with the
weights inversely proportional to the distances. Essentially, this method assumes that closer
points have a greater influence on the estimation. It's more sophisticated than Nearest
Neighbour and often produces more accurate results, especially when there's a smooth
spatial gradient in rainfall. Some variations are:

 Inverse Square: This refers to the mathematical principle that the influence of a
point on the estimation decreases with the square of the distance. In IDW, this
means that points farther away have a significantly smaller impact on the estimation
compared to those nearby.

 Spline: A spline is a smooth curve that passes through a set of points or data values.
It's used to create a continuous surface from discrete data points. Splines are often
employed in interpolation methods like IDW to generate smooth estimates of rainfall
distribution across an area. But here This refers to the mathematical principle that
the influence of a point on the estimation decreases with the cube of distance.

 Thin Plate Spline: This is a specific type of spline interpolation method used in spatial
analysis. It's particularly useful for estimating surfaces with complex spatial patterns,
such as rainfall over mountainous terrain or irregular landscapes. Thin plate spline
interpolation balances smoothness and accuracy, making it suitable for capturing
intricate variations in rainfall distribution. To calculate such examples, unlike the
other three earlier methods it also takes in the elevation as a variable thus making it
a three-dimensional model. This is not often used in public but rather mostly used in
meteorological centres and n research.

These principles can be explained with an example below.

We first need to measure the distance, for that we have two methods specific to the each of
the method used.

Now when we measure the rainfall, for these methods, we use a weighted mean of all the
rainfall recordings recorded nearby. The weighted arithmetic mean is similar to an ordinary
arithmetic mean (the most common type of average), except that instead of each of the
data points contributing equally to the final average, some data points contribute more than
others and this notion is applied that those recordings from afar influence X lesser than
others nearby and thus the weights used are an inverse of a power of distance (d, d 2, d3).
These interpolation of data and others lead to the two common forms of data: Station Data
which is Point Data and Gridded Data which is interpolated data.

Station data refers to observations collected from specific locations, typically from weather
stations or other monitoring sites. These stations are strategically placed across regions to
provide localised measurements of various meteorological parameters such as temperature,
precipitation, humidity, wind speed, and atmospheric pressure and offers detailed and
precise information about weather conditions at individual points on the Earth's surface.
However, coverage can be sparse in some areas, especially in remote or less populated
regions thus needing to use gridded data.

Gridded data, on the other hand, uses methods of interpolation or extrapolation of

observations from weather stations to create continuous spatial datasets covering larger
geographic areas. This is achieved by dividing the Earth's surface into a grid of evenly spaced
cells and estimating climate variables for each cell based on nearby station data using
interpolation techniques like Inverse Distance Weighting (IDW), kriging, or splining. Gridded
datasets provide great coverage and are useful for analysing regional and global climate
patterns, assessing climate change impacts, and supporting various scientific and
environmental applications.

Station Data of Chennai and Gridded Data of India from IMD

To gain such data, IMD in India is a wonderful source. The India Meteorological Department
(IMD) is India's representative in meteorology and weather forecasting. Established in 1875,
it operates under the Ministry of Earth Sciences of the Government of India. With its
headquarters in New Delhi, IMD has regional offices nationwide, it has great coverage and
dissemination of weather information. IMD uses technology and a network of observatories,
radars, and satellites to monitor weather patterns, predict storms, cyclones, and other
atmospheric phenomena, and provide accurate forecasts to the public, government
agencies, and various sectors such as agriculture, aviation, and disaster management and
also in climate monitoring and research as well. Equivalent organisations are Meteo France,
NOVA, and the UK Meteorological Office.