Geohydrology and safety of Dutch canal dikes: threats from above

Abstract

Flood risk is of key interest to societies in low-lying areas such as the Netherlands, where a wide variety of dikes protect people and property from flooding. This thesis focuses on canal dikes, located along drainage canals in polders, mainly in the northern and western regions of the Netherlands. Safety assessment outcomes and recent dike failures indicate that inner-slope instability is a key failure mechanism affecting canal-dike safety, occurring under both very dry and very wet conditions. Assessing inner-slope instability remains challenging due to large uncertainties in geotechnical and geohydrological parameters. The aim of this thesis is to enhance the geohydrological understanding of canal dikes through monitoring and observation-based analysis, thereby improving the assessment of their stability and the safety of dike systems as a whole. Two unique datasets were established: a multiyear detailed monitoring series of soil moisture levels and hydraulic heads at ten canal dikes, and a nationwide collection of over one hundred hydraulic head time series from about 50 monitoring sites.

The detailed monitoring series included two extremely dry summers (2020 and 2022). Hydraulic heads at inner-slope and toe monitoring points showed strong seasonal variation. During winter, conditions were near saturation, while during dry summers heads dropped down to nearly 2 m, despite outside water levels remaining nearly constant. Non-hydrostatic hydraulic head levels were observed within dike bodies, conditions often not accounted for in safety assessments for drought situations. The precipitation deficit proved the most reliable meteorological drought indicator, outperforming the standardised drought indicators (SPEI and SPI).

Analysing the nationwide dataset with time-series models, a non-linear model performed best, resulting in 35 reliable time-series models. Four clusters of dikes were identified, differentiated by response times, defined as the time required for 95% of an impulse's influence to dissipate. Longer response times cause peak heads to occur later in the winter. Peak head statistics indicated a that extreme heads are close to yearly occuring heads, with a median decimate height of 15 cm and a range of 5 to 50 cm. By 2100, extreme peak heads are expected to occur between three times less frequently and eight times more frequently, depending on the climate scenario and the type of canal dike.

The time-series models were used to hindcast 60 years of hydraulic head levels and estimate spatial dependencies, quantified using the length-effect factor of peak heads. At the polder scale, variation in head responses can increase the factor by a factor of four to five, whereas spatial weather variability can double it. At the scale of the entire canal-dike system, the length-effect of peak heads is dominated by spatial weather variability, increasing from about 6 for annually occurring heads to around 40 for heads with an exceedance frequency of 1/100 per year.

The impact of small load variations on reliability updating was analysed. The impact of reliability updating increases as load variations decrease, regardless of the prior failure probability. The two canal dikes, with decimate heights below 10 cm, benefited most from reliability updating, whereas updated failure probabilities of the river dikes changed only slightly. A credibility check was introduced: when uncertainty in dike strength is more than 1.5 times larger than the variation in loads, estimated failure probabilities are not credible. Reducing uncertainties in dike strength, through monitoring with reliability updating or more detailed soil strength data, leads to more credible failure probability estimates.