Skip to content

Numerical Weather Prediction Short Notes

One-page CSIR NET Part C revision: CFL, time schemes, filters, NWP model hierarchy, forecast ranges, verification, objective analysis, assimilation, satellites.

  • 8 min read
  • Updated
On this page · 8 sections
  1. Part C in one table
  2. Constants used
  3. 1. Computational Instability and Filtering
  4. 2. Barotropic, Baroclinic and Primitive-Equation Models
  5. 3. Short-, Medium- and Long-Range Prediction
  6. 4. Objective Analysis, Initialization, Data Assimilation and Satellite Data
  7. Formula drill
  8. Where to go next

Every formula, range and trap from Computational Instability and Filtering, Barotropic, Baroclinic and Primitive-Equation Models, Short-, Medium- and Long-Range Prediction and Objective Analysis, Initialization, Data Assimilation and Satellite Data, on one page. The chapter overview is at Numerical Weather Prediction.

Part C in one table

Per the NTA information bulletin for December 2026, Earth Sciences Part C:

GivenAttemptCorrectWrongMaximum
8025+4−1.32100

Only the first 25 answered questions are evaluated. Full pattern on the Earth Sciences overview.

Constants used

ConstantValue
Earth’s rotation \( \Omega \)\( 7.292 \times 10^{-5} \) s⁻¹
Earth’s radius \( a \)6371 km
\( \beta \) at 45°\( 2\Omega\cos\phi/a \approx 1.62 \times 10^{-11} \) m⁻¹ s⁻¹
External gravity-wave speed\( \sqrt{gH} \approx 280 \) m/s (\( H = 8 \) km) to 313 m/s (\( H = 10 \) km)
Geostationary altitudeabout 35 800 km
Polar orbiter altitudeabout 800 km (typical)

1. Computational Instability and Filtering

FactDetail
Lax equivalenceconsistency + stability ⇔ convergence (linear, well posed)
von Neumann test\( u_j^n = \lambda^n e^{ikj\Delta x} \); stable if \( \lvert\lambda\rvert \le 1 \)
Oscillation equation\( du/dt = i\omega u \); \( p = \omega\Delta t \)
Forward\( \lvert\lambda\rvert = \sqrt{1 + p^2} \): always unstable
Backward\( 1/\sqrt{1 + p^2} \): stable, damping
Trapezoidal1: stable, neutral, second order
Matsuno\( \sqrt{1 - p^2 + p^4} \); stable \( \lvert p\rvert \le 1 \); damps high frequencies
Heun\( \sqrt{1 + p^4/4} \): weakly unstable
Leapfrog\( \lambda = ip \pm\sqrt{1 - p^2} \); neutral \( \lvert p\rvert \le 1 \); physical mode phase \( \arcsin p \) (fast); computational mode \( \to -1 \)
Robert–Asselin\( \bar u^{n} = u^{n} + \nu(\bar u^{n-1} - 2u^{n} + u^{n+1}) \); damps computational mode; lowers accuracy to first order
CFL\( c_{\max}\Delta t/\Delta x \le 1 \) (1-D); about \( 1/\sqrt2 \) on 2-D square grid
FTCS advectionunconditionally unstable
Upstream\( 0 \le \mu \le 1 \); diffusive (coefficient \( \tfrac{c\Delta x}{2}(1 - \mu) \))
Semi-implicitgravity terms implicit → advective CFL only
Semi-Lagrangianremoves advective CFL; accuracy limits \( \Delta t \)
Pole problemconverging meridians; Fourier filtering, reduced grids, spectral
Centred phase speed\( c^{*}/c = \sin(k\Delta x)/(k\Delta x) \): \( 2\Delta x \) → 0, \( 3\Delta x \) → 0.41, \( 4\Delta x \) → 0.64, \( 8\Delta x \) → 0.90
Numerical group velocity\( c\cos(k\Delta x) \): negative below \( 4\Delta x \)
Shortest wave\( 2\Delta x \)
Aliasing\( k^{*} = 2\pi/\Delta x - k \) for \( k > \pi/\Delta x \)
Nonlinear instabilityaliasing; not cured by smaller \( \Delta t \); Arakawa Jacobian (energy + enstrophy), alias-free grid (\( \ge 3N + 1 \) points), Shapiro filter, \( \nabla^4 \) diffusion
Hydrostaticfilters vertical sound waves (Lamb wave remains)
Anelastic / Boussinesqfilters all sound waves
Quasi-geostrophicfilters sound + gravity; Rossby only

2. Barotropic, Baroclinic and Primitive-Equation Models

FactDetail
Barotropic atmosphere\( \rho = \rho(p) \); no thermal wind; no APE to release
Barotropic vorticity equation\( \partial_t\nabla^2\psi = -J(\psi, \nabla^2\psi + f) \); absolute vorticity conserved
First computer forecasts1950, ENIAC; Charney, Fjørtoft, von Neumann; barotropic vorticity equation
Rossby–Haurwitz\( c = U - \beta/(k^2 + l^2) \); westward relative to flow
Stationary wavelength\( L_s = 2\pi\sqrt{U/\beta} \) (about 7000 km for 20 m/s at 45°)
Equivalent barotropicnear level of non-divergence (about 500–600 hPa)
Barotropic model limitsno development, no vertical motion, ultralong waves retrogress too fast
Shallow-water modeldivergence + gravity waves \( \sqrt{gH} \); conserves \( (\zeta + f)/h \)
QGsmall Rossby number; geostrophic advection; \( \omega \) diagnostic (omega equation); conserves QGPV
Two-level modelvorticity at 250 and 750 hPa, \( \omega \) at 500 hPa; minimum for baroclinic instability
QG limitstropics, fronts, steep terrain; constant \( f_0 \), \( \sigma \)
PE prognostic\( u, v, T, q, p_s \)
PE diagnostic\( \omega \) (continuity), \( \Phi \) (hydrostatic)
σ coordinate\( p/p_s \); \( \dot\sigma = 0 \) at ground; pressure-gradient error over slopes
Hybrid σ–pterrain-following low, pressure aloft
Spectralspherical harmonics; T = triangular (isotropic); shortest wavelength \( \approx 2\pi a/N \) (T639 → about 63 km)
Parameterizationsradiation, convection, microphysics, PBL, land surface, gravity-wave drag
Non-hydrostaticneeded at a few km grid length (convection-permitting)
Limited-area modellateral boundary relaxation zone; one-way or two-way nesting

3. Short-, Medium- and Long-Range Prediction

FactDetail
WMO rangesnowcast ≤ 2 h; very short ≤ 12 h; short 12–72 h; medium 3–10 d; extended 10–30 d; long 30 d–2 yr; climate > 2 yr
ChaosLorenz (1963); sensitive dependence on initial conditions
Intrinsic limitorder of two weeks for daily weather
Error growth\( E = E_0 2^{t/\tau_d} \); halving \( E_0 \) gains one \( \tau_d \); tenfold gains \( 3.32\tau_d \)
First / second kindinitial conditions / boundary forcing
Long-range sourcesSST (ENSO, IOD), soil moisture, snow cover, sea ice
Short-range toolsconvection-permitting regional models, rapid-update DA, radar nowcasting
MOS vs perfect progMOS corrects model bias; perfect prog assumes perfect model
Medium-range skill500 hPa ACC; 60% = useful
EPS perturbationssingular vectors, bred vectors, ensemble DA; stochastic physics; multi-model
Ensemblemean = best estimate (smooth); spread = uncertainty; fraction of members = probability
Extended rangeMJO (about 30–60 days), monsoon intraseasonal oscillation; weekly anomalies
Seasonal forecasttercile probabilities; statistical + coupled dynamical; hindcasts; multi-model ensembles
IMD monsoon categories< 90 deficient; 90–95 below normal; 96–104 normal; 105–110 above normal; > 110 excess (% of LPA)

Verification

ScoreFormulaPerfect
Bias (mean error)\( \overline{F - O} \)0
RMSE\( \sqrt{\overline{(F - O)^2}} \)0
ACC\( \sum f'o'/\sqrt{\sum f'^2\sum o'^2} \)1
POD\( a/(a + c) \)1
FAR (ratio)\( b/(a + b) \)0
CSI (threat score)\( a/(a + b + c) \)1
Frequency bias\( (a + b)/(a + c) \)1
ETS\( (a - a_r)/(a + b + c - a_r) \), \( a_r = (a + b)(a + c)/n \)1
Brier score\( \frac1N\sum(p - o)^2 \)0
Skill score\( 1 - S/S_{\text{ref}} \)1

4. Objective Analysis, Initialization, Data Assimilation and Satellite Data

FactDetail
Ingredientsbackground \( \mathbf{x}_b \) (\( \mathbf{B} \)), observations \( \mathbf{y} \) (\( \mathbf{R} \), includes representativeness), operator \( H \)
Innovation / increment\( \mathbf{y} - H\mathbf{x}_b \) / \( \mathbf{x}_a - \mathbf{x}_b \)
Analysis historypolynomial fitting → successive correction → OI → 3D/4D-Var, EnKF
Cressman\( w = (R^2 - r^2)/(R^2 + r^2) \), 0 beyond \( R \); passes with shrinking \( R \)
Barnes\( w = e^{-r^2/\kappa} \)
Scalar OI\( x_a = x_b + \frac{\sigma_b^2}{\sigma_b^2 + \sigma_o^2}(y - x_b) \); \( 1/\sigma_a^2 = 1/\sigma_b^2 + 1/\sigma_o^2 \)
BLUE\( \mathbf{K} = \mathbf{B}H^{\mathsf T}(H\mathbf{B}H^{\mathsf T} + \mathbf{R})^{-1} \)
Why initializeimbalance → spurious gravity waves, spin-up
Staticbalance relation (geostrophic, nonlinear balance)
Dynamicforward–backward integration with damping scheme (Matsuno)
Linear vs nonlinear NMIgravity-mode amplitudes = 0 / gravity-mode tendencies = 0 (Machenhauer)
DFIlow-pass time filter on short forward–backward run
3D-Var\( J = \tfrac12(\mathbf{x}-\mathbf{x}_b)^{\mathsf T}\mathbf{B}^{-1}(\mathbf{x}-\mathbf{x}_b) + \tfrac12(\mathbf{y}-H\mathbf{x})^{\mathsf T}\mathbf{R}^{-1}(\mathbf{y}-H\mathbf{x}) \); = OI for linear \( H \)
4D-Varwindow; model strong constraint; adjoint gives gradient; incremental
Kalman / EnKFforecasts error covariance; ensemble estimate with localization and inflation
Quality controlgross, background, buddy checks; thinning; variational bias correction
Polar vs geostationaryabout 800 km, global, twice daily / about 35 800 km, fixed disc, minutes
Temperature soundingCO₂ 15 μm (IR, clear sky); O₂ 50–60 GHz (microwave, through cloud)
Humidity soundingH₂O 6.3 μm; 183 GHz
Other satellite dataAMVs (geostationary), scatterometer surface winds, GNSS RO (bending angle, refractivity), microwave imager water vapour and rain
Weighting function\( d\tau/d\ln p \); broad, overlapping → retrieval ill-posed
Direct radiance assimilationradiative-transfer operator + bias correction; preferred over retrievals
Impactlargest over oceans and southern hemisphere

Formula drill

$$\mu = \frac{c\,\Delta t}{\Delta x} \le 1 \qquad \lvert\lambda_{\text{forward}}\rvert = \sqrt{1 + p^2} \qquad \lambda_{\text{leapfrog}} = ip \pm \sqrt{1 - p^2}$$$$\frac{c^{*}}{c} = \frac{\sin k\Delta x}{k\Delta x} \qquad k^{*} = \frac{2\pi}{\Delta x} - k \qquad c = U - \frac{\beta}{k^2 + l^2}$$$$t = \tau_d\log_2\frac{E}{E_0} \qquad x_a = x_b + \frac{\sigma_b^2}{\sigma_b^2 + \sigma_o^2}(y - x_b) \qquad w = \frac{R^2 - r^2}{R^2 + r^2}$$

Where to go next

Advertisement
Advertisement