Skip to content

Weather Forecasting: Elements, Methods and Hazards

CSIR NET Part C meteorology: forecast ranges, NWP, forecasting temperature, rain, fog and thunderstorms, IMD hazard criteria and forecast verification.

  • 19 min read
  • Updated
On this page · 15 sections
  1. Prerequisites
  2. The Hook: Richardson’s Forecast Factory
  3. Forecast Ranges and Methods
  4. Forecasting Weather Elements
  5. Forecasting Thunderstorms: Stability Indices
  6. Temperature Hazards: IMD Heat and Cold Wave Criteria
  7. Cyclone and Impact-Based Warnings
  8. Forecast Verification
  9. Memory Tricks
  10. Don’t Confuse
  11. Common Mistakes
  12. Practice
  13. Quick Revision Table
  14. Key Takeaways
  15. What’s Next

Prerequisites

The Hook: Richardson’s Forecast Factory

How CSIR NET asks it. Part C questions on this line include computing stability indices (K index, Total Totals, lifted index) from a sounding and judging thunderstorm potential; deciding whether a heat wave, cold wave or heavy-rain category applies from given data; identifying the fog type from conditions; computing verification scores (POD, FAR, CSI, bias, Brier score); and placing methods (nowcasting, NWP, ensembles) on the right range.


Forecast Ranges and Methods

RangeLead timeMain tools
Nowcasting0 to a few hoursradar and satellite extrapolation, lightning detection, AWS, conceptual models of storms
Very short rangeup to about 12 hnowcasting tools plus high-resolution NWP
Short rangeabout 1–3 dayssynoptic analysis, deterministic NWP
Medium rangeabout 3–10 daysglobal NWP, ensembles
Extended rangeabout 10–30 daysensembles and coupled models; skill from slowly varying modes (MJO, MISO)
Long range / seasonala month to a seasoncoupled ocean–atmosphere models; statistical models using predictors such as ENSO and the IOD

Classical methods still used for checking and for short lead times:

MethodIdeaWorks when
Persistencetomorrow = todayslowly changing regimes; very short range
Climatologyforecast the long-term averagelong range; baseline for skill
Trend (extrapolation)systems keep moving at their present speed and directionnowcasting fronts, squall lines, cyclones
Analoguefind a past situation like today and use what followedrarely exact; atmosphere seldom repeats
Synoptic / conceptualapply models of fronts, cyclones, monsoon systemsshort range
Numerical weather predictionintegrate the governing equations from an analysed initial stateall ranges beyond nowcasting

Numerical weather prediction (NWP)

graph LR
    A["Observations
surface, upper air,
satellite, radar, aircraft"] --> B["Quality control"] B --> C["Data assimilation
3D-Var, 4D-Var,
ensemble Kalman filter"] C --> D["Analysis
initial state"] D --> E["Model integration
primitive equations +
parameterised physics"] E --> F["Post-processing
MOS, bias correction,
downscaling"] F --> G["Forecast products
and warnings"] E --> H["Ensemble members
perturbed initial states
and physics"] H --> G

Key ideas:

  • Primitive equations: momentum (usually hydrostatic for global models, non-hydrostatic at high resolution), continuity, thermodynamic energy, moisture, equation of state.
  • Parameterisation: processes smaller than the grid (convection, boundary-layer turbulence, radiation, cloud microphysics, surface exchange) are represented statistically. Convection parameterisation is a large source of error in the tropics.
  • Numerical stability: explicit time stepping must satisfy the CFL condition \( c\,\Delta t/\Delta x \le 1 \), where \( c \) is the fastest signal speed resolved; finer grids need shorter time steps.
  • Data assimilation blends a short-range model forecast (the background) with observations, weighted by their error statistics, to produce the analysis.
  • Predictability: small errors in the initial state grow (Lorenz’s sensitive dependence on initial conditions), so deterministic skill for synoptic weather is lost after roughly two weeks. Ensemble prediction runs many forecasts from perturbed initial states; the spread estimates uncertainty and gives probabilities.
  • Model output statistics (MOS): regression relations between model output and observed station weather that correct systematic biases.

Forecasting Weather Elements

Temperature

FactorEffect on maximum or minimum temperature
Advection \( -\mathbf{V}\cdot\nabla T \)warm advection raises, cold advection lowers; dominant with frontal and western disturbance passages
Cloud coverby day cuts maximum; by night raises minimum (less longwave loss)
Windat night mixes warmer air down and keeps minimum up; calm nights allow strong inversions
Humiditymoist air keeps nights warmer (water-vapour greenhouse); dry desert air has a large diurnal range
Soil moisture and snowwet soil and snow cover lower the maximum
Thickness and model fieldslayer-mean temperature from forecast thickness; model 2 m temperature with MOS

A rough rule for clear, calm nights in a stable air mass: the minimum temperature is not much below the evening dew point, because condensation (dew, fog) releases latent heat once saturation is reached.

Precipitation

  • Quantitative precipitation forecasts (QPF) come from NWP and are calibrated with radar and gauge data.
  • Probability of precipitation (PoP) expresses uncertainty from ensembles or statistics.
  • Ingredients for heavy rain: deep moisture (high precipitable water), lift (convergence, orography, fronts, low-pressure systems), instability and a long duration (slow movement or repeated cells).

IMD rainfall intensity terms (24-hour accumulation at a station):

CategoryRainfall (mm)
Very lighttrace to 2.4
Light2.5–15.5
Moderate15.6–64.4
Heavy64.5–115.5
Very heavy115.6–204.4
Extremely heavy204.5 or more

Wind

Gradient and geostrophic wind from the pressure analysis, adjusted for friction and stability (strong daytime mixing brings momentum down; nocturnal inversions decouple the surface). Local effects: land and sea breezes, mountain and valley winds, channelling, gusts in convective downdrafts.

Visibility and fog

Fog is a suspension of water droplets that reduces horizontal visibility below 1 km; mist is the same with visibility of 1 km or more.

Fog typeFormationFavourable conditionsIndian example
Radiation fogsurface cools by longwave emission at nightclear sky, light wind (about 1–3 m s⁻¹), moist surface air, long nights; often after rainIndo-Gangetic plains in winter, especially after a western disturbance has moistened the ground and cleared out
Advection fogwarm moist air flows over a colder surfacemoderate wind; sharp surface temperature contrastcoastal areas over cold water; less common in India
Steam (evaporation) fogcold air over much warmer waterlarge air–water temperature differencerivers and lakes on cold mornings
Upslope fogmoist air cooled adiabatically as it risesgentle upslope flowhill slopes
Frontal fograin falling into cold air saturates itahead of warm frontsmid-latitudes

Radiation fog dissipates after sunrise as the surface warms and mixing increases; a stronger wind or a cloud deck at night prevents it. Fog forecasting uses the evening dew point, the expected minimum temperature, wind and cloud forecasts, and satellite night-time fog products (difference of IR channels).

IMD grades fog by visibility from shallow and moderate fog through dense and very dense fog; dense fog reduces visibility to a few hundred metres or less and disrupts aviation and rail traffic.


Forecasting Thunderstorms: Stability Indices

Thunderstorms need moisture, instability and a lifting mechanism. Indices computed from a sounding (see Observations and Synoptic Charts) summarise the first two.

IndexFormulaInterpretation
K index\( K = (T_{850} - T_{500}) + T_{d,850} - (T_{700} - T_{d,700}) \)higher = more likely air-mass thunderstorms; values of about 35 and above indicate high potential (operational tables vary)
Total Totals\( TT = T_{850} + T_{d,850} - 2T_{500} \)about 50 or more suggests severe thunderstorms are possible
Lifted index\( LI = T_{500,\mathrm{env}} - T_{500,\mathrm{parcel}} \) (parcel lifted from the surface or mixed layer)negative = unstable; more negative = more vigorous
Showalter indexas LI but parcel lifted from 850 hPanegative = unstable
CAPE\( \int_{LFC}^{EL} g\,(T_{v,p} - T_{v,e})/T_{v,e}\,dz \)energy available to updrafts; maximum updraft \( \approx \sqrt{2\,\mathrm{CAPE}} \)
CINnegative area below the LFCthe “lid”; large CIN suppresses storms until it is broken

The K index includes a dryness penalty (the 700 hPa dew-point depression), so it is suited to air-mass (non-severe) storms in moist tropical air; Total Totals and LI weight lapse rate and instability. Vertical wind shear (from the hodograph) decides storm organisation: weak shear gives short-lived single cells; moderate shear multicells and squall lines; strong, veering shear supercells.

Convective hazards over India

HazardSeason and regionFeatures
Nor’westers (kalbaisakhi)April–May; West Bengal, Odisha, Jharkhand, Assam, Bangladeshsquall lines approaching from the north-west; hail, gusts, lightning; dry-line setting with cT air over moist mT air
Dust storms (andhi)April–June; Rajasthan, Punjab, Haryana, Delhi, west UPconvective downdrafts and gust fronts lift dust; can be violent
Pre-monsoon thunderstorms in the southApril–May; Kerala, Karnataka‘mango showers’, ‘blossom showers’
Haillate winter and pre-monsoon; north and central Indianeeds strong updrafts and a low freezing level
Lightningpre-monsoon and monsoon; east and central India especiallya major cause of weather-related deaths in India
Cloudburstmonsoon; Himalaya and hillsextreme short-duration rain over a small area, enhanced by orographic lifting; often tied to a monsoon system or a WD–monsoon interaction
Squallswith thunderstorms and frontsa sudden, sharp increase of wind speed lasting minutes, then dying away

Temperature Hazards: IMD Heat and Cold Wave Criteria

Heat wave

A heat wave is considered when the maximum temperature of a station reaches at least 40 °C in the plains, 37 °C at coastal stations or 30 °C in hilly regions, and then:

BasisHeat waveSevere heat wave
Departure from normal maximum4.5 °C to 6.4 °C above normalmore than 6.4 °C above normal
Actual maximum (plains)45 °C or more47 °C or more

Heat waves in India are typical of April–June over the north-west, central and east India and the east coast, under subsidence, clear skies and hot, dry cT air (the loo). A delayed monsoon onset prolongs them. Humidity makes coastal heat more dangerous at lower temperatures.

Cold wave

For plains stations with a minimum temperature of 10 °C or less:

BasisCold waveSevere cold wave
Departure from normal minimum4.5 °C to 6.4 °C below normalmore than 6.4 °C below normal
Actual minimum (plains)4 °C or less2 °C or less

Cold waves follow western disturbances: clearing skies and cold north-westerlies after passage, with snow-covered hills upstream. Cold day conditions occur when fog or low cloud persists all day and keeps the maximum far below normal.


Cyclone and Impact-Based Warnings

Tropical cyclone warnings. IMD issues cyclone bulletins in stages as a system approaches the coast, typically: a pre-cyclone watch, a cyclone alert, a cyclone warning, and a post-landfall outlook, each given progressively closer to landfall, with track and intensity forecasts shown as a cone of uncertainty. Storm-surge forecasts are issued with them (see Tropical Weather Systems).

Colour-coded impact-based warnings. IMD’s district-level warnings use a four-colour scale:

ColourMeaning
Greenno warning; no action
Yellowwatch: be aware, stay updated
Orangealert: be prepared
Redwarning: take action

Forecast Verification

For a yes/no event (for example, “heavy rain at a district”), forecasts and observations fill a 2 × 2 contingency table:

Observed yesObserved no
Forecast yes\( a \) hits\( b \) false alarms
Forecast no\( c \) misses\( d \) correct negatives
ScoreFormulaPerfectNote
Probability of detection (POD) / hit rate\( a/(a+c) \)1ignores false alarms
False alarm ratio (FAR)\( b/(a+b) \)0fraction of “yes” forecasts that were wrong
Probability of false detection (POFD)\( b/(b+d) \)0used in ROC curves
Critical success index (CSI) / threat score\( a/(a+b+c) \)1ignores correct negatives; good for rare events
Frequency bias\( (a+b)/(a+c) \)1> 1 over-forecasting; < 1 under-forecasting
Proportion correct\( (a+d)/n \)1inflated by many correct “no"s for rare events
Heidke skill score (HSS)\( \dfrac{2(ad-bc)}{(a+c)(c+d)+(a+b)(b+d)} \)1skill relative to random chance; 0 = no skill

For probability forecasts, the Brier score is

$$BS = \frac{1}{N}\sum_{i=1}^{N}(p_i - o_i)^2$$

with \( p_i \) the forecast probability and \( o_i \) = 1 or 0. It ranges from 0 (perfect) to 1. A skill score compares any score with a reference (climatology or persistence):

$$SS = 1 - \frac{\text{score}_{\text{forecast}}}{\text{score}_{\text{reference}}}$$

for error-type scores such as MSE or the Brier score. SS = 1 is perfect, 0 is no better than the reference, negative is worse.


Memory Tricks

Don’t Confuse

ThisvsThat
Nowcasting: 0 to a few hours, extrapolationvsShort range: 1–3 days, NWP
Deterministic forecast: one runvsEnsemble: many runs, probabilities
POD: hits / observed eventsvsFAR: false alarms / forecast events
FAR (ratio, over yes forecasts)vsPOFD (over observed non-events)
CSI: ignores correct negativesvsProportion correct: includes them, flatters rare-event forecasts
Radiation fog: light wind, clear night, landvsAdvection fog: warm moist air over a cold surface, needs wind
K index: high = unstablevsLifted index: negative = unstable
Heat wave criteria: departure or absolute maximumvsCold wave criteria: departure or absolute minimum
Fog: visibility below 1 kmvsMist: visibility 1 km or more

Common Mistakes


Practice

All questions below are original practice questions written in CSIR NET style. They are not previous-year questions.

Level 1: Foundation

  1. In what IMD category is a 24-hour station rainfall of 120 mm?
  2. Write the formula for the critical success index.
  3. Which fog type is typical of the Indo-Gangetic plains on clear winter nights?
Answers
  1. Very heavy (115.6–204.4 mm).
  2. CSI = hits/(hits + false alarms + misses) = \( a/(a+b+c) \).
  3. Radiation fog.

Level 2: Part B Style

Q1 (practice). A sounding gives \( T_{850} = 24 \) °C, \( T_{d,850} = 18 \) °C, \( T_{700} = 10 \) °C, \( T_{d,700} = 4 \) °C and \( T_{500} = -8 \) °C. The K index and Total Totals are:

  • (a) 44 and 58
  • (b) 32 and 42
  • (c) 44 and 42
  • (d) 50 and 58
Answer and explanation

(a). \( K = (24 - (-8)) + 18 - (10 - 4) = 32 + 18 - 6 = 44 \). \( TT = 24 + 18 - 2(-8) = 58 \). Both indicate high thunderstorm potential, with TT in the severe range.

Q2 (practice). A plains station with a normal maximum of 39.5 °C records 45.0 °C. Under IMD criteria this is:

  • (a) no heat wave, because the departure is below 6.5 °C
  • (b) a heat wave
  • (c) a severe heat wave
  • (d) a heat wave only if it persists for a week
Answer and explanation

(b). The station is above 40 °C. The departure is 5.5 °C (heat wave band, 4.5–6.4 °C), and the actual maximum of 45 °C also meets the absolute heat-wave criterion. It is not severe: departure is below 6.5 °C and the maximum is below 47 °C.

Q3 (practice). A southerly wind of 10 m s⁻¹ blows across isotherms in which temperature decreases northward by 5 K per 500 km. Ignoring other effects, the local temperature change due to advection is about:

  • (a) −0.36 K h⁻¹
  • (b) +0.36 K h⁻¹
  • (c) +3.6 K h⁻¹
  • (d) +8.6 K h⁻¹
Answer and explanation

(b). \( -\mathbf{V}\cdot\nabla T = -v\,\partial T/\partial y = -(10)(-5/5\times10^5) = +1\times10^{-4} \) K s⁻¹ = +0.36 K h⁻¹ (about 8.6 K per day): warm advection. Option (d) is the daily value given with the wrong unit.

Q4 (practice). An explicit model with 10 km grid spacing must resolve signals travelling at up to 300 m s⁻¹. The largest stable time step under the CFL condition is about:

  • (a) 3 s
  • (b) 33 s
  • (c) 330 s
  • (d) 3000 s
Answer and explanation

(b). \( \Delta t \le \Delta x/c = 10^4/300 = 33 \) s. Halving the grid spacing halves the allowed time step, so cost rises steeply with resolution.

Level 3: Part C Style

Q1 (practice). Over one monsoon season, district heavy-rain warnings gave the following results:

Heavy rain observedNot observed
Warning issued3010
No warning20140

Compute POD, FAR, CSI, frequency bias, proportion correct and the Heidke skill score. Comment on the forecasts.

Answer and worked solution

\( a = 30, b = 10, c = 20, d = 140, n = 200 \).

  • POD = 30/50 = 0.60
  • FAR = 10/40 = 0.25
  • CSI = 30/60 = 0.50
  • Bias = 40/50 = 0.80 (under-forecasting: warnings issued less often than events occurred)
  • Proportion correct = 170/200 = 0.85
  • HSS = \( \dfrac{2(30\times140 - 10\times20)}{(50)(160) + (40)(150)} = \dfrac{8000}{14000} = \) 0.57

Comment: warnings caught 60% of events with a modest false-alarm ratio, but the bias below 1 shows the forecasters were cautious; issuing a few more warnings would raise POD at the cost of FAR. The 85% proportion correct looks better than the CSI of 0.5 because of the 140 easy correct negatives.

Q2 (practice). A forecaster issued probability-of-rain forecasts on five days: 0.9, 0.7, 0.2, 0.1 and 0.6. Rain occurred on the first two days only.

(i) Compute the Brier score. (ii) Compute the Brier skill score against a climatological forecast of 0.4 on every day. (iii) Interpret.

Answer and worked solution
  1. \( BS = \frac{1}{5}[(0.9-1)^2 + (0.7-1)^2 + (0.2)^2 + (0.1)^2 + (0.6)^2] = \frac{1}{5}[0.01 + 0.09 + 0.04 + 0.01 + 0.36] = 0.51/5 = \) 0.102.
  2. Climatology: \( \frac{1}{5}[2(0.4-1)^2 + 3(0.4)^2] = \frac{1}{5}[0.72 + 0.48] = 0.24 \). \( BSS = 1 - 0.102/0.24 = \) 0.575.
  3. The forecasts have clear skill over climatology. Most of the error comes from day 5 (0.6 with no rain); without it, the score would be much better.

Q3 (practice). On a December evening over the Indo-Gangetic plains, a western disturbance has just passed. Observations at 12 UTC: sky clearing, surface wind 2 m s⁻¹ and expected to fall to 1–2 m s⁻¹ overnight, temperature 16 °C, dew point 11 °C, soil moist after rain the previous day. Normal minimum at the station is 8 °C.

Which of the following statements are correct?

(A) Radiation fog is likely by early morning. (B) The minimum temperature is unlikely to fall far below about 10–11 °C, because the air will saturate near the dew point and latent heat release and fog will slow further cooling. (C) If the minimum were instead 3.5 °C, a cold wave would apply on the absolute criterion. (D) A strong north-westerly wind of 8 m s⁻¹ overnight would make fog more likely.

  • (a) A, B and C only
  • (b) A and D only
  • (c) B, C and D only
  • (d) A, B, C and D
Answer and worked solution

(a) A, B and C.

  • (A) Clear sky, light wind, moist ground and a dew-point depression of 5 °C that the night will remove: classic radiation-fog setting after a WD. Correct.
  • (B) Correct as a forecasting rule: once the air reaches its dew point, condensation (dew and fog) releases latent heat and the fog layer itself reduces longwave loss.
  • (C) A plains minimum of 4 °C or less is a cold wave on the absolute criterion (2 °C or less would be severe). Correct.
  • (D) Incorrect: strong wind mixes the boundary layer and prevents radiation fog.

Quick Revision Table

ItemFact
Rangesnowcast (0 to a few h), very short (to 12 h), short (1–3 d), medium (3–10 d), extended (10–30 d), long/seasonal
Classical methodspersistence, climatology, trend, analogue, synoptic
NWP chainobs → QC → assimilation (3D/4D-Var, EnKF) → model → MOS → products; ensembles for uncertainty
CFL\( c\Delta t/\Delta x \le 1 \)
Predictabilityabout two weeks for synoptic weather
IMD rain (24 h)light 2.5–15.5; moderate 15.6–64.4; heavy 64.5–115.5; very heavy 115.6–204.4; extremely heavy ≥ 204.5 mm
Heat wave≥ 40 °C plains (37 coast, 30 hills); +4.5 to +6.4 °C = HW, > 6.4 = severe; or ≥ 45 / ≥ 47 °C
Cold wavemin ≤ 10 °C plains; −4.5 to −6.4 °C = CW, beyond −6.4 = severe; or ≤ 4 / ≤ 2 °C
Fog< 1 km; radiation (clear, light wind, moist); advection (warm over cold, windy)
IndicesK = (T850 − T500) + Td850 − (T700 − Td700); TT = T850 + Td850 − 2T500; LI < 0 unstable; \( w_{max} \approx \sqrt{2\,\mathrm{CAPE}} \)
Hazardsnor’westers, andhi, hail, lightning, cloudbursts, heat and cold waves, fog, cyclones
Warningscyclone: watch → alert → warning → post-landfall outlook; colours green/yellow/orange/red
VerificationPOD a/(a+c); FAR b/(a+b); CSI a/(a+b+c); bias (a+b)/(a+c); HSS; Brier score; SS = 1 − S/S_ref

Key Takeaways

What’s Next

Advertisement
Advertisement