To use the normal approximation you will need to compute the mean (sample size * proportion "success") and standard deviation (square root of (sample size * proportion "success" * (1 - proportion "success")).
Let X be binomial variable = number of flights on time.
a) You can only compute probabilities under the continuous normal curve for intervals, so the probability of a single point (120) is zero.
b) You want to compute P(X > 120). Standardizing to z-score you get: P(Z > (120 - mean)/standard deviation), which you can compute from normal distribution tables or by using a calculator.
c) Here you want to compute P(X < 98). You standardize to z-score and compute probability as in b).
d) Here you want to compute P(98 < X 120). Just compute 1 - the sum of probabilities from b) and c).