In nowadays, wavelength-division multiplexing (WDM) networks, on the one hand, increasingly more users expect the network to provide high-priority QoS services demanding no congestion and low latency. On the other hand, it is significantly more difficult for network operators to forecast future traffic demands, as the packet traffic running over WDM networks fluctuates over time for a variety of reasons. Confronted with a rough understanding of traffic patterns as well as the increasing number of time-sensitive applications, most networks today are grossly over-provisioned. Thus, designing cost-effective WDM networks in an uncertain traffic environment, which includes network planning and robust routing, is both an important and a challenging task. In this paper, we explore adaptive load-balancing to investigate the problems of network planning and robust routing for WDM mesh networks under varying traffic matrices. We first propose an efficient heuristic algorithm called Maximizing Network Capability (MNC) to provision congestion-free and cost-effective WDM networks based on load-balancing to deal with traffic uncertainty. Then, a novel traffic grooming algorithm called Adding Direct Traffic (ADT) is proposed to implement robust routing with partial traffic information. Finally, we demonstrate by simulation that MNC consumes less resources than previous methods and performs quite close to the optimal solution, while ADT achieves the desirable performance in delay, jitter (delay variation), and throughput compared with existing robust routing and traffic grooming algorithms.