From fd8ad2ff7503e3e14608fd4c4e77d2b1c0268244 Mon Sep 17 00:00:00 2001 From: pjbartlein Date: Tue, 28 Sep 2021 16:56:33 -0700 Subject: [PATCH] updated README.md files --- README.md | 25 ++++++++++++++++--------- animations/README.md | 13 +++++++------ data/nc_files/README.md | 8 ++++---- f90/README.md | 8 ++++---- figures/supplemental_figures/README.md | 6 +++--- 5 files changed, 34 insertions(+), 26 deletions(-) diff --git a/README.md b/README.md index 4b97781..99e57a3 100644 --- a/README.md +++ b/README.md @@ -3,7 +3,7 @@ PaleoCalAdjust This is the repository that accompanies the paper: -Bartlein, P. J. and Shafer, S. L.: Paleo calendar-effect adjustments in time-slice and transient climate-model simulations (PaleoCalAdjust v1.0): impact and strategies for data analysis, *Geosci. Model Dev.*, [https://doi.org/10.5194/gmd-12-3889-2019](https://doi.org/10.5194/gmd-12-3889-2019), 2019. +Bartlein, P. J. and Shafer, S. L.: Paleo calendar-effect adjustments in time-slice and transient climate-model simulations (PaleoCalAdjust v1.0): impact and strategies for data analysis, *Geosci. Model Dev., 12, 3889-3913*, [https://doi.org/10.5194/gmd-12-3889-2019](https://doi.org/10.5194/gmd-12-3889-2019), 2019. ## Abstract ## @@ -11,9 +11,9 @@ The “paleo calendar effect” is a common expression for the impact that the c ## Programs ## -The current version of the key program is `cal_adjust.f90` (in the folder `/f90/main_programs`), which applies the paleo calendar-effect adjustment to CMIP/PMIP-type netCDF files, but should also work in the case of files that are "CF-compliant" (or nearly so), and they are also known to work with netCDF files of transient paleoclimatic simulations. There is a related program, `month_length.f90`, that can be used to produce tables of the changing length of months over time that are used by `cal_adjust.f90`. Figures illustrating the paleo calendar effect are in the folder `/figures`, and relevant data sets for exercising the programs are in the folder `/data`. +The current version of the key program is `cal_adjust.f90` (in the folder `/f90/main_programs`). This program applies the paleo calendar-effect adjustment to CMIP/PMIP-type netCDF files, but should also work in the case of files that are "CF-compliant" (or nearly so), and is known to work with netCDF files of transient paleoclimatic simulations. There is a related program, `month_length.f90`, that can be used to produce tables of the changing length of months over time that are used by `cal_adjust.f90`. Figures illustrating the paleo calendar effect are in the folder `/figures`, and relevant data sets for exercising the programs are in the folder `/data`. -Several minor modifications to the main program and its modules were made since the original *Geoscientific Model Development Discussions (GMDD)* manuscript submission to accommodate the adjustment of CMIP6-PMIP4 files. Additionally, following a referee's suggestion, we replaced the approach used in the initial submission of the paper for calculating month lengths (i.e., the approximation of Kutzbach and Gallimore (1988, *J. Geophys. Res.* 93(D1):803-821)), with a direct approach based on Kepler's equation. This substitution of approaches had no practical significance. Several other code modifications were made over time in the interests of transparency. +Several minor modifications to the main program and its modules were made since the original *Geoscientific Model Development Discussions (GMDD)* manuscript submission to accommodate the adjustment of CMIP6-PMIP4 files. Additionally, following a referee's suggestion, we replaced the approach used in the initial submission of the paper for calculating month lengths (i.e., the approximation of Kutzbach and Gallimore (198)), with a direct approach based on Kepler's (1609) equation. This substitution of approaches had no practical significance. Several other code modifications were made over time in the interests of transparency. The current version is v1.1. Relative to previous versions, this version includes: @@ -22,23 +22,30 @@ The current version is v1.1. Relative to previous versions, this version includ - specification of the infofile path and name on the command line, so that once built locally, the programs can be run in a terminal window; - addition of the adjusted month lengths, and beginning, middle, and ending dates to the output file; - a choice of two mean-preserving interpolation methods, including the Epstein (1991) approach implemented in v1.0, as well as the Harzallah (1995) iterated-spline approach; -- the inclusion of a subroutine, `enforce_mean()` that requires the pseudo-daily interpolated values to have the same monthly mean as the input monthly values; +- the inclusion in `pseudo_daily_interp_subs.f90` of a subroutine, `enforce_mean()` that requires the pseudo-daily interpolated values to have the same monthly mean as the input monthly values; - some modifications of I/O to accommodate (some) "non-standard" files (e.g. TraCE-21ka). ## Interpolation methods ## -The Epstein (1991) interpolation approach is intrinsically periodic, meaning that when applied to interpolate pseudo-daily values from monthly input values, the interpolated daily values at the end of the year will be consistent with those at the beginning, which is a desirable feature. However, when iteratively applied to multi-annual time series of monthly data, small discontinuities will arise between years. In the v1.0 implementation, this discontinuity was removed by smoothing the interpolated daily values at the end and beginning of the year. The Harzallah (1995) approach, which involves iteratively fitting splines to the input data (and to the residuals from the original fit) is intrinsically not periodic, meaning the interpolated daily values at the end of the year will not be consistent with those at the beginning of a single year. However, because this approach involves local as opposed to global fitting (as in the Epstein approach), the input data can be padded, either cyclically in the case of a single year's data, or with data from adjacent years in the case of time series. This effectively eliminates the discontinuity between years. The Epstein (1991) approach is recommended for adjusting "climatology" data sets (e.g. CMIP/PMIP "Aclim" time-frequency data), while the Harzallah (1995) approach is better suited for adjusting time-series data (e.g. "Amon"-type data sets), or transient-simulation data. +The Epstein (1991) interpolation approach is intrinsically periodic, meaning that when applied to interpolate pseudo-daily values from monthly input values, the interpolated daily values at the end of the year will be consistent with those at the beginning, which is a desirable feature. However, when iteratively applied to multi-annual time series of monthly data, small discontinuities will arise between years. In the v1.0 implementation, this discontinuity was removed by smoothing the interpolated daily values at the end and beginning of the year. The Harzallah (1995) approach, which involves iteratively fitting splines to the input data (and to the residuals from the original fit) is intrinsically not periodic, meaning the interpolated daily values at the end of the year will not be consistent with those at the beginning of a single year. However, because this approach involves local as opposed to global fitting (as in the Epstein approach), the input data can be padded, either cyclically in the case of a single year's data, or with data from adjacent years in the case of time series. This effectively eliminates the discontinuity between years. The Epstein (1991) approach is recommended for adjusting "climatology" data sets (e.g. CMIP/PMIP "Aclim"-type time-frequency data), while the Harzallah (1995) approach is better suited for adjusting time-series data (e.g. "Amon"-type data sets), or transient-simulation data. -Despite the name, "mean-preserving" interpolation methods do not necessarily yield interpolated data that exactly reproduce the input data. This can be addressed by setting a "tolerance" (`tol`) value for reproduction of the input values (typically 0.01 or 0.001 times the mean value of the data), that when exceeded, causes the discrepancy to be redistributed among the interpolated values. Values of `tol` that are too large may lead to differences between the input monthly means and the means of the pseudo-daily interpolated values. Values of `tol` that are too low may result in anomalously large adjusted values. This will be easily seen in maps of the calendar effect, the difference between the adjusted and input monthly values. +Despite the name, "mean-preserving" interpolation methods do not necessarily yield interpolated data that exactly reproduce the input data. This issue can be addressed by setting a "tolerance" (`tol`) value for reproduction of the input values (typically 0.01 or 0.001 times the mean value of the data), that when exceeded, causes the discrepancy to be redistributed among the interpolated values. Values of `tol` that are too large may lead to differences between the input monthly means and the means of the pseudo-daily interpolated values. Values of `tol` that are too low may result in anomalously large adjusted values. This will be easily seen in maps of the calendar effect, that display the difference between the adjusted and input monthly values. Further discussion of mean-preserving interpolation, and comparisons among several practical approaches for its application can be found in the GitHub repository at [[https://github.com/pjbartlein/mp-interp]](https://github.com/pjbartlein/mp-interp). Contact Pat Bartlein (bartlein@uoregon.edu) for further information. Animations used in a presentation at the Fall 2019 AGU Meeting can be found in the `/animations` folder. References: -Epstein, E), On obtaining daily climatological values from monthly means, *J. Climate* 4:365-368. -Harzallah, A. (1995) The interpolation of data series using a constrained iterating technique *Monthly Weather Review* 123:2251-2254. -Bartlein, P.J. and S.L. Shafer, 2019, Paleo calendar effects on radiation, atmospheric circulation, and surface temperature, moisture, and energy-balance variables can produce interpretable but spurious large-scale patterns and trends in analyses of paleoclimatic simulations. PP31A-08, AGU 2019 Fall Meeting. [[https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/525140]](https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/525140) +Bartlein, P.J. and S.L. Shafer: Paleo calendar effects on radiation, atmospheric circulation, and surface temperature, moisture, and energy-balance variables can produce interpretable but spurious large-scale patterns and trends in analyses of paleoclimatic simulations, PP31A-08, AGU 2019 Fall Meeting, https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/525140, 2019. + +Epstein, E.S.: On obtaining daily climatological values from monthly means, *J. Climate*, 4, 365-368, https://doi.org/10.1175/1520-0442(1991)004<0365:OODCVF>2.0.CO;2, 1991. + +Harzallah, A.: The interpolation of data series using a constrained iterating technique, *Monthly Weather Review*, 123, 2251-2254, https://doi.org/10.1175/1520-0493(1995)123<2251:TIODSU>2.0.CO;2, 1995. + +Kepler, J.: *New Astronomy (Astronomia Nova)*, translated from the Latin by: Donahue, W. H., Cambridge University Press, Cambridge, UK, 681 pp., 1992, 1609.  + +Kutzbach, J.E. and Gallimore, R.G.: Sensitivity of a coupled atmosphere/mixed layer ocean model to changes in orbital forcing at 9000 years B.P., *J. Geophys. Res.-Atmos.*, 93, 803–821, https://doi.org/10.1029/JD093iD01p00803, 1988.  + ## Version history ## diff --git a/animations/README.md b/animations/README.md index 88c950d..83e2b7b 100644 --- a/animations/README.md +++ b/animations/README.md @@ -12,11 +12,12 @@ The folders include: Six subfolders, containing individual .pngs: / orb_0ka + / orb_0ka_misc / orb_insol65n_anm / orb_monangle / orb_monlen / orb_position - / orb_solsitice_anm + / orb_solstice_anm /files ! text files with ImageMagick 7 command lines and file lists /gifs ! animated .gifs @@ -28,14 +29,14 @@ The folders include: The animations include: - `orb_0ka.gif` The present-day elliptical orbit, showing the position of the Earth at 5-degree intervals during the year. -- `orb_insol65n_anm` Insolation long-term mean differences (from 0 ka, also known as "anomalies"). -- `orb_position.gif` Earth's orbit over at 1 kyr intervals over the past 150 kyr. +- `orb_insol65n_anm.gif` Insolation long-term mean differences (from 0 ka, also known as "anomalies"). +- `orb_position.gif` Earth's orbit at 1 kyr intervals over the past 150 kyr. - `orb_monlen.gif` An animated version of Fig. 1 in Bartlein and Shafer (2019, *GMD*), showing the orbit, and month-length "anomalies" (long-term mean differences between 0 ka (1950 CE) and each time. -- `orb_solstice_anm` An animated version of Fig. 2 in Bartlein and Shafer (2019), showing the orbit, and differences between the middle day of each month and the June solstice. -- `orb_monangle.gif` As alternative illustration of the month-lenth differences, where January 1 (as opposed to the March equinox is the reference point. This animation illustrates the conservation of the "angular" or celestial definitions of the months. +- `orb_solstice_anm.gif` An animated version of Fig. 2 in Bartlein and Shafer (2019), showing the orbit, and differences between the middle day of each month and the June solstice. +- `orb_monangle.gif` As alternative illustration of the month-lenth differences, where January 1 (as opposed to the March) equinox is the reference point. This animation illustrates the conservation of the "angular" or celestial definitions of the months. For the sake of illustration, eccentricity has been exaggerated by ten for plotting in each animation, but the geometric positions, and the orbital speed in `orb_0ka.gif`, have been calculated using the true eccentricity values. -Citation: If you use the animations, please cite: Bartlein, P.J. and S.L. Shafer, 2019, Paleo calendar effects on radiation, atmospheric circulation, and surface temperature, moisture, and energy-balance variables can produce interpretable but spurious large-scale patterns and trends in analyses of paleoclimatic simulations. PP31A-08, AGU 2019 Fall Meeting. [[https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/525140]](https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/525140), as will as the GMD paper. +Citation: If you use the animations, please cite: Bartlein, P.J. and S.L. Shafer, 2019, Paleo calendar effects on radiation, atmospheric circulation, and surface temperature, moisture, and energy-balance variables can produce interpretable but spurious large-scale patterns and trends in analyses of paleoclimatic simulations. PP31A-08, AGU 2019 Fall Meeting. [[https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/525140]](https://agu.confex.com/agu/fm19/meetingapp.cgi/Paper/525140), as well as the Bartlein and Shafer *GMD* paper. diff --git a/data/nc_files/README.md b/data/nc_files/README.md index 7e6766d..277f433 100644 --- a/data/nc_files/README.md +++ b/data/nc_files/README.md @@ -12,7 +12,7 @@ Example CMIP/PMIP-formatted netCDF source files (and resulting month-length adju There are three sets of folders that contain model output for the *midHolocene* experiment that can be used to demonstrate `cal_adjust.f90`: `/data/nc_files/test1/`, `/data/nc_files/test2/`, and `/data/nc_files/test3/` each of which contains a `source` and `adjusted` folder. In addition, the folder `/data/nc_files/ctrl_nc_files/` contains *piControl* simulations that can be used to illustrate the impact of the calendar effect on long-term mean differences. -- `/data/nc_files/test1/` contains some typical CMIP5/PMIP3 netCDF files, including long-term means (`Aclim`) files, monthly time series (`Amon` files), and daily time series (`day`) files): +- `/data/nc_files/test1/source/` contains some typical CMIP5/PMIP3 netCDF files, including long-term means (`Aclim`) files, monthly time series (`Amon` files), and daily time series (`day`) files): tas_Amon_MPI-ESM-P_midHolocene_r1i1p1_185001-194912.nc tas_Amon_IPSL-CM6A-LR_midHolocene_r1i1p1f1_gr_185001-204912.nc @@ -29,7 +29,7 @@ There are three sets of folders that contain model output for the *midHolocene* pr_Amon_CCSM4_midHolocene_r1i1p1_100001-130012.nc pr_Aclim_CCSM4_midHolocene_r1i1p1_100001-130012-clim.nc -- `/data/nc_files/test2/` contains contains a variety of CMIP5/PMIP3 and CMIP6/PMIP4 3-D and 4-D files, including some on rotated-pole ocean grids: +- `/data/nc_files/test2/source/` contains contains a variety of CMIP5/PMIP3 and CMIP6/PMIP4 3-D and 4-D files, including some on rotated-pole ocean grids: tas_Amon_MPI-ESM-P_midHolocene_r1i1p1_185001-194912.nc tas_Amon_CCSM4_midHolocene_r1i1p1_100001-130012.nc @@ -47,7 +47,7 @@ There are three sets of folders that contain model output for the *midHolocene* msftmyz_Oclim_MPI-ESM-P_midHolocene_r1i1p1_185001-194912-clim.nc msftmyz_Oclim_CCSM4_midHolocene_r1i1p1_100001-130012-clim.nc -- `/data/nc_files/test3/` contains additional files that show the impact of the calendar effect on a range of other variables: +- `/data/nc_files/test3/source/` contains additional files that show the impact of the calendar effect on a range of other variables: clt_Aclim_CCSM4_midHolocene_r1i1p1_100001-130012-clim.nc hfls_Aclim_CCSM4_midHolocene_r1i1p1_100001-130012-clim.nc @@ -106,4 +106,4 @@ Source data for calendar-effects figures: ana4mips_URL.txt cdo_ltm.txt ESRL-PSD_CMAP_URL.txt - tas_Aclim_CFSR_reanalysis_ana4mips_198101-201012-clim_cal_adj.nc \ No newline at end of file + tas_Aclim_CFSR_reanalysis_ana4mips_198101-201012-clim_cal_adj.nc diff --git a/f90/README.md b/f90/README.md index 2bfa39e..fe10850 100644 --- a/f90/README.md +++ b/f90/README.md @@ -1,6 +1,6 @@ ## Programs (PaleoCalAdjust v1.1) ## -Main programs, including `month_length.f90` and `cal_adjust_PMIP.f90`, plus additional driver and demonstration programs are in the folder `/main_programs`: +Main programs, including `month_length.f90` and `cal_adjust.f90`, plus additional driver and demonstration programs are in the folder `/main_programs`: month_length.f90 ! month-length tables cal_adjust.f90 ! paleo calendar adjustment @@ -21,7 +21,7 @@ The `/modules` folder contains the following: mp_interp_harzallah_subs.f90 ! Harzallah (1995) iterative-spline interpolation spline_subs.f90 ! Burkhardt spline-interpolation subroutines -The module `spline_subs.f90` contains several subroutines and functions from John Burkhardt's library of Fortran90 spline-fitting subroutines, and is used by `mp_interp_harzallah_subs.f90` [[https://people.sc.fsu.edu/~jburkardt/f_src/spline/spline.html)]](https://people.sc.fsu.edu/~jburkardt/f_src/spline/spline.html)) +The module `spline_subs.f90` contains several subroutines and functions from John Burkhardt's library of Fortran90 spline-fitting subroutines, and is used by `mp_interp_harzallah_subs.f90` [[https://people.sc.fsu.edu/~jburkardt/f_src/spline/spline.html)]](https://people.sc.fsu.edu/~jburkardt/f_src/spline/spline.html) The `/projects` folder contains a set of subfolders, one for each main program, containing example GNU Make makefiles for the individual main programs. The makefiles must be localized for a particular operating system or file structure (to correctly point to the compiler and source code). The examples are for macOS. @@ -34,7 +34,7 @@ The programs are used as follows: - `GISS_orpar_driver.f90` and `GISS_srevents_driver.f90` write orbital-parameter output to the folder `/GISS_orbital`, using specific parameter values set in the programs; - `month_length.f90` reads the info file `month_length_info.csv` in the folder `/info_files` and writes month-length tables to the folder `/month_lengths`; -- `cal_adjust_PMIP.f90` reads the info file `cal_adj_info.csv` in the folder `/info_files`, and source netCDF files from the folders `/nc_files/PMIP3_source` and `/nc_files/PMIP4_source` and writes paleo calendar-adjusted netCDF files to the folder `/nc_files/PMIP3_adjusted` and `/nc_files/PMIP4_adjusted`; +- `cal_adjust_PMIP.f90` reads the info file specified on the command line, and source netCDF files from the `/nc_files/test1/source/`, `/nc_files/test2/source/`, and `/nc_files/test3/source/` folders and writes paleo calendar-adjusted netCDF files to the folder `/nc_files/test1/adjusted/`, `/nc_files/test2/adjusted/` and `/nc_files/test3/adjusted/`; - `demo_01_pseudo_daily_interp.f90`, and `demo_02_adjust_1yr.f90`, are stand-alone programs, writing only to the console; - `demo_03_adjust_TraCE_ts.f90` reads, for example, the file `TraCE_c30r40_tas_land_monlen0ka_Jan-Dec.csv` in the folder `/TraCE_example` and writes a paleo calendar-adjusted output file into the same folder. @@ -44,4 +44,4 @@ The same code compiles and runs on the following systems: - Windows 10: Intel Parallel Studio XE 2019 Update 4 Composer Edition for Fortran, with netCDF version 4.1.3, using the Visual Studio 2019 IDE; - MacOS: gfortran version 10.2.0 (Homebrew GCC 10.2.0_4), with netCDF version 4.8.0 (from Homebrew), using the Eclipse IDE for Scientific Computing, Version 4.19.0 (2021-03), with Eclipse for Parallel Application Developers, Version 9.4.0. -- Linux: (Ubuntu 18.04.5 LTS (bionic)): gfortran version 7.5.0, with netcdf version 4.6.0. \ No newline at end of file +- Linux: (Ubuntu 18.04.5 LTS (bionic)): gfortran version 7.5.0, with netcdf version 4.6.0. diff --git a/figures/supplemental_figures/README.md b/figures/supplemental_figures/README.md index 3697c1f..b5f7d12 100644 --- a/figures/supplemental_figures/README.md +++ b/figures/supplemental_figures/README.md @@ -1,7 +1,7 @@ Supplemental Figures ----------------- -The following figures illustrate the calendar effects on a selection of simulations used to exercise the calendar-effect adjustment program (`cal_adjust_PMIP.f90`). Each figure has four panels: +The following figures illustrate the calendar effects on a selection of simulations used to exercise the calendar-effect adjustment program (`cal_adjust_PMIP.f90` (version 1). Each figure has four panels: - the long-term mean differences (experiment minus control, sometimes referred to as "anomalies") for each simulation calculated using adjusted paleo data (upper left); - the long-term mean differences (experiment minus control) calculated using unadjusted data (upper right); @@ -28,10 +28,10 @@ For *AClim* files, with time dimension lengths of 12, the calendar-effect maps a ## Discussion -There are broad similarities across individual models of the calendar effects in the various simulations (lower left), that resemble the "pure" calendar effects shown in Fig. 11 and 12, including those for the *midHolocene* simulations as will as the single *lig127k* simulation of near-surface air temperature (and that one is also consistent with the *midHolocene* and *lig127k* temperature-effect difference on Fig. 11). The larger calendar effects can be seen to rival in size the paleo minus *piControl* long-term mean differences (upper left and upper right In many cases the calendar effects simply reinforce the long-term mean differences, but in some cases they are large enough to change the sign of the difference (see *Long-term mean difference sign reversals* maps below). The spatial patterns of the sign-change maps have two components: 1) large, or continental-scale patches that indicate where and when during the year the calendar effect could completely reverse the sense of change between a paleo and control experiment, and 2) long looping, or hollow cell-like features that indicate where the location of zero isopleths of the long-term mean differences are sensitive to the calendar effect. +There are broad similarities across individual models of the calendar effects in the various simulations (lower left), that resemble the "pure" calendar effects shown in Fig. 11 and 12, including those for the *midHolocene* simulations as well as a single *lig127k* simulation of near-surface air temperature (and that one is also consistent with the *midHolocene* and *lig127k* temperature-effect difference on Fig. 11). The larger calendar effects can be seen to rival in size the paleo minus *piControl* long-term mean differences (upper left and upper right). In many cases the calendar effects simply reinforce the long-term mean differences, but in some cases they are large enough to change the sign of the difference (see the long-term mean difference sign reversal panels). The spatial patterns of the sign-change maps have two components: 1) large, or continental-scale patches that indicate where and when during the year the calendar effect could completely reverse the sense of change between a paleo and control experiment, and 2) long looping, or hollow cell-like features that indicate where the location of zero isopleths of the long-term mean differences are sensitive to the calendar effect. The three *midHolocene* simulations of *tas* with *Aclim* files (CCSM4, CNRM-CM5, and MPI-ESM-P) can be compared with those with *Amon* files. In general, the calendar-effect patterns are quite similar, with perhaps slightly stronger map patterns expressed by data from the *Amon* files, which is probably related to the order of differencing and averaging. (This is more evident in the two *pr* simulations with both *Aclim* and *Amon* files.) However, there is little practical difference in patterns that emerge from the two kinds of files. The two PMIP4 simulations with IPSL-CM6A-LR, *midHolocene* and *lig12ka* show calendar-effect differences that strongly resemble those on Fig. 11 despite large differences in the amplitude and sign of the long-term mean differences, particularly in May and June. Comparisons among the different *tas* simulations suggests that the calendar effects vary little between models with larger amplitude "experiment minus control" long-term mean differences and those with smaller amplitude anomalies. This implies that adjusting for the calendar effect does not suppress inter-model (or inter "warm-climate" experiment) differences (although this is a very small sample, both in terms of number of models and variables). -Among the three *midHolocene* precipitation-rate simulations generally show smaller spatial-scale long-term mean difference sign-reversal patterns than temperature, reflecting both the smaller spatial scale of the map-patterns of precipitation, and the tendency for the long-term mean differences to show a combination of amplification and damping of broad-scale features (like the monsoons), as well as some shifts in the location of features. For example, the calendar-effect maps show a consistent pattern of decreases in N. African monsoon precipitation in the adjusted data in July and August, and increases in September, October and November. In contrast to temperature, these changes do not produce sign changes in the long-term mean differences that are of large scale and consistent from month to month. +The three *midHolocene* precipitation-rate simulations generally show smaller spatial-scale long-term mean difference sign-reversal patterns than temperature, reflecting both the smaller spatial scale of the map-patterns of precipitation, and the tendency for the long-term mean differences to show a combination of amplification and damping of broad-scale features (like the monsoons), as well as some shifts in the location of features. For example, the calendar-effect maps show a consistent pattern of decreases in N. African monsoon precipitation in the adjusted data in July and August, and increases in September, October and November. In contrast to temperature, these changes do not produce sign changes in the long-term mean differences that are of large scale and consistent from month to month.