On the Optimal Design of Field Significance Tests for Changes in Climate Extremes

Field significance tests have been widely used to detect climate change. In most cases, a local test is used to identify significant changes at individual locations, which is then followed by a field significance test that considers the number of locations in a region with locally significant changes. The choice of local test can affect the result, potentially leading to conflicting assessments of the impact of climate change on a region. We demonstrate that when considering changes in the annual extremes of daily precipitation, the simple Mann-Kendall trend test is preferred as the local test over more complex likelihood ratio tests that compare the fits of stationary and nonstationary generalized extreme value distributions. This lesson allows us to report, with enhanced confidence, that the intensification of annual extremes of daily precipitation in China since 1961 became field significant much earlier than previously reported. Plain Language Summary Changes to weather and climate extremes at individual locations can be highly uncertain due to natural variability. Much of the natural variability in precipitation extremes occurs on small spatial scales, and thus analyzing changes at different locations in a region with a field significance test can help extract information about changes in the region that is less affected by natural variability. An important component of doing so is the local test that is used to identify significant changes at individual locations and a field significance test that evaluates whether such changes are found at more locations than would be expected from natural variability in an unchanged climate. By contrasting several common local test methods with varying complexity, we find that the simple Mann-Kendall test tends to yield a field significance test with high power of detection. Based on these lessons, we find that the intensification of extreme precipitation in China became field significant much earlier than previously reported, thereby resolving uncertainty about whether intensification is in fact discernable in China. Key Points Field significance is generally determined by summarizing the results of individual tests conducted at different locations in a domain Inconsistent field significance conclusions can be drawn when using different local test methods For extreme precipitation, field significance determined from the simple Mann-Kendall test performs better than other commonly used ones

Automated summary

Field significance tests are commonly used in climate change detection, with the choice of local test method impacting the results significantly. For extreme precipitation in China, the simple Mann-Kendall trend test is preferred for detecting changes accurately.