A "dry day" is a day without any agriculturally meaningful rainfall, which is generally defined by a threshold of 0.1 mm/day. The maximum number of consecutive dry days is an important metric for rain-fed agriculture as it directly impacts soil moisture, and thus crop growth. As climate warms, one of the signals is the increase in contrast: when it rains, it might rain harder, but when its dry it might get drier. The trend toward more consecutive dry days and higher temperatures will increase evaporation and add stress to limited water resources, affecting irrigation and other water uses. Long periods of consecutive days with little or no precipitation also can lead to drought. In general, the average annual maximum number of consecutive dry days are projected to increase for the higher emissions scenarios. Some crops, however, might benefit from this change, particularly when the dry conditions exist in specific parts of the crop cycle.
The graph shows the recorded maximum number of Consecutive Dry Days (CDD) per year for 1986-2005, and projected maximum number of CDD for 2020-2100 under all RCPs of CIMP5 ensemble modeling. Note, the shaded ranges (or model spread) illustrate the inter-model differences, here using the +/- one standard deviation. The reason for using a narrower metric compared to the 10th and 90th percentile is that the inter-model difference is large for precipitation, and in particular for the count of days with rainfall.
Temporal variations in the projections of the maximum length of consecutive dry days reflect the multi-decadal variability in weather and climate. At regional scales, that inter-annual to decadal variability is often larger than the difference between emission scenarios, particularly in the first half of the 21st century. Depending on the size of the domain being averaged, the separation between emission scenarios is not always clearly recognizable. One difficulty is that not all modeling groups contributed data from all the different RCPs, and thus the multi-model median values shown in the dashboard figure are based on a different set of underlying models. Particularly RCP6.0 generally is represented by the fewest members, which leads to a somewhat noisier projection. The general tendencies over time joined by the increasing separation of the RCPs is the most robust representation of the expected changes.