The influence of the North Sea on coastal precipitation in the Netherlands in the present-day and future climate

Abstract

Coastal precipitation in the Netherlands for present-day and future climate using observations and simulations with the KNMI regional climate model RACMO2 is studied. Overall a good correspondence between RACMO2 and the observations is found for the present-day climate. Yet, the model generally simulates too sharp gradients in precipitation amounts near the coast line, and the maximum in precipitation is shifted towards the sea compared to the observations. The seasonal cycle of coastal precipitation (here the difference between precipitation near the coast and inland) is well linked to the land sea temperature contrast. Yet, this is not a one-to-one relation and other effects (like atmospheric stability) are likely to be important as well. Variations from year to year in monthly coastal precipitation correlate well with variations in strength of the cyclonic westerly circulation in the periods August to November. A moderate sensitivity to sea surface temperatures (≈6 mm month⁻¹ K⁻¹) has also been found. In simulations of climate change a small increase in coastal precipitation is generally found, except for the spring period. Using a more realistic prescription of North Sea temperatures (from a slab ocean model) only marginally changes these results. The increase in coastal precipitation is tempered by a decrease in land sea temperature contrast, and considerably larger increases of coastal precipitation could be expected if the temperature rise over sea equalled the temperature rise over land. Despite the fact that the overall changes in coastal effect appear relatively moderate, impact on more extreme events could be considerable, increasing the probability of exceeding extreme thresholds by a factor two or more.

Slab model

The ocean slab model is given by the following equation:

$$\frac{\partial T}{\partial t} = \frac{1}{h c_{p} \rho}( Q + F ) - \frac{1}{\beta} ( T - T_{ext}). $$

(3)

The slab layer has a thickness h. Experience shows that optimal depths are between 5 and 35 m; smaller values lead to too large fluctuations with the weather, whereas a too large depth effectively turns off the model. The thickness is limited by the observed ocean depth, which is in particular relevant near the coast. The change in slab temperature T is then given by the energy balance at the surface F and a prescribed energy flux Q (the Q-flux). The last relaxation term is used to prevent a too strong divergence between the simulated T and the temperature provided by the boundary (T _ext ). The relaxation time (β) is set to 1 month for the North Sea, and smoothly drops to 1 week elsewhere.

Using the slab model is a two step process. First, a normal run with an accurate, high resolution SST is done from which T _high and F _high are used to determine Q:

$$Q = h c_p \rho \frac{\partial T_{high}}{\partial t}- F_{high} $$

(4)

To get good statistics a sufficiently long period should be used; in the literature, typically a 15 years run with a climatological SST is taken. Here, we use a satellite derived SST for the period 1996–2010. The Q-flux is determined per month, using monthly averages. Note that the Q-flux is model specific: it depends on any model bias or error in the radiation.

The Q-flux can then be used in a run with the slab model turned on. This effectively means running with a perturbation on the climatology of the high resolution SST. Combining 3 and 4 we get:

$$\frac{\partial}{\partial t} (T - T_{high} )= \frac{1}{h c_p \rho} ( F - F_{high} ) - \frac{1}{\beta} ( T - T_{ext} ) $$

(5)

The term on the left represents the deviation from the climatological temperature; this is determined by the local energy balance (influenced by the weather). The last term is a relaxation back to the external SST as set in the boundaries.

With the slab model, and using a coarse resolution SST as input, a large part of the coastal warming as visible in the high resolution satellite observations is retrieved (see the bottom part of Fig. 2). Although ocean currents, and thus advection of heat, are not included explicitly in the slab model, they are included indirectly via the prescribed Q-flux. Consequently, the slab model can only be used for climate integrations when it is assumed that no large changes in ocean circulation occur.