detrending data

[monsoon]

It is always suggested that we should detrend data before calibration and estimation in a dsge model. But detrending leaves the series with the cycle and the averages may not be meaningful at all. For example, I detrended a time series of real gdp and the averages are negative which intuitively does not make sense. The data is attached for your perusal. It is detrended using hpfilter.

There is no problem when we use interest rates for calibration because they are not detrended. But using per capita consumption or gdp in calibration can be difficult.

How do we deal with this?

Thanks