The Berne dataset contains soil responses and a large set of explanatory covariates. The study area is located to the Northwest of the city of Berne and covers agricultural area. Soil responses included are soil pH (4 depth intervals calculated from soil horizon), drainage classes (3 ordered classes) and presence of waterlogging characteristics down to a specified depth (binary response).
Covariates cover environmental conditions by representing climate, topography, parent material and soil.
data("berne")A data frame with 1052 observations on the following 238 variables.
site_id_uniqueID of original profile sampling
xeasting, Swiss grid in m, EPSG: 21781 (CH1903/LV03)
ynorthing, Swiss grid in m, EPSG: 21781 (CH1903/LV03)
datasetFactor splitting dataset for calibration and independent validation. validation was assigned at random by using weights to ensure even spatial coverage of the agricultural area.
dclassDrainage class, ordered Factor.
waterlog.30Presence of waterlogging characteristics down to 30 cm (1: presence, 0: absence)
waterlog.50Presence of waterlogging characteristics down to 50 cm (1: presence, 0: absence)
waterlog.100Presence of waterlogging characteristics down to 100 cm (1: presence, 0: absence)
ph.0.10Soil pH in 0-10 cm depth.
ph.10.30Soil pH in 10-30 cm depth.
ph.30.50Soil pH in 30-50 cm depth.
ph.50.100Soil pH in 50-100 cm depth.
timesetFactor with range of sampling year and label for sampling type for soil pH. no label: \(CaCl_{2}\) laboratory measurements, field: field estimate by indicator solution, ptf: \(H_{2}0\) laboratory measurements transferred by pedotransfer function (univariate linear regression) to level of \(CaCl_{2}\) measures.
cl_mt_etap_pecolumns 14 to 238 contain environmental covariates representing soil forming factors. For more information see Details below.
cl_mt_etap_ro
cl_mt_gh_1
cl_mt_gh_10
cl_mt_gh_11
cl_mt_gh_12
cl_mt_gh_2
cl_mt_gh_3
cl_mt_gh_4
cl_mt_gh_5
cl_mt_gh_6
cl_mt_gh_7
cl_mt_gh_8
cl_mt_gh_9
cl_mt_gh_y
cl_mt_pet_pe
cl_mt_pet_ro
cl_mt_rr_1
cl_mt_rr_10
cl_mt_rr_11
cl_mt_rr_12
cl_mt_rr_2
cl_mt_rr_3
cl_mt_rr_4
cl_mt_rr_5
cl_mt_rr_6
cl_mt_rr_7
cl_mt_rr_8
cl_mt_rr_9
cl_mt_rr_y
cl_mt_swb_pe
cl_mt_swb_ro
cl_mt_td_1
cl_mt_td_10
cl_mt_td_11
cl_mt_td_12
cl_mt_td_2
cl_mt_tt_1
cl_mt_tt_11
cl_mt_tt_12
cl_mt_tt_3
cl_mt_tt_4
cl_mt_tt_5
cl_mt_tt_6
cl_mt_tt_7
cl_mt_tt_8
cl_mt_tt_9
cl_mt_tt_y
ge_caco3
ge_geo500h1id
ge_geo500h3id
ge_gt_ch_fil
ge_lgm
ge_vszone
sl_nutr_fil
sl_physio_neu
sl_retention_fil
sl_skelett_r_fil
sl_wet_fil
tr_be_gwn25_hdist
tr_be_gwn25_vdist
tr_be_twi2m_7s_tcilow
tr_be_twi2m_s60_tcilow
tr_ch_3_80_10
tr_ch_3_80_10s
tr_ch_3_80_20s
tr_cindx10_25
tr_cindx50_25
tr_curv_all
tr_curv_plan
tr_curv_prof
tr_enessk
tr_es25
tr_flowlength_up
tr_global_rad_ch
tr_lsf
tr_mrrtf25
tr_mrvbf25
tr_ndom_veg2m_fm
tr_nego
tr_nnessk
tr_ns25
tr_ns25_145mn
tr_ns25_145sd
tr_ns25_75mn
tr_ns25_75sd
tr_poso
tr_protindx
tr_se_alti10m_c
tr_se_alti25m_c
tr_se_alti2m_fmean_10c
tr_se_alti2m_fmean_25c
tr_se_alti2m_fmean_50c
tr_se_alti2m_fmean_5c
tr_se_alti2m_std_10c
tr_se_alti2m_std_25c
tr_se_alti2m_std_50c
tr_se_alti2m_std_5c
tr_se_alti50m_c
tr_se_alti6m_c
tr_se_conv2m
tr_se_curv10m
tr_se_curv25m
tr_se_curv2m
tr_se_curv2m_s15
tr_se_curv2m_s30
tr_se_curv2m_s60
tr_se_curv2m_s7
tr_se_curv2m_std_10c
tr_se_curv2m_std_25c
tr_se_curv2m_std_50c
tr_se_curv2m_std_5c
tr_se_curv50m
tr_se_curv6m
tr_se_curvplan10m
tr_se_curvplan25m
tr_se_curvplan2m
tr_se_curvplan2m_grass_17c
tr_se_curvplan2m_grass_45c
tr_se_curvplan2m_grass_9c
tr_se_curvplan2m_s15
tr_se_curvplan2m_s30
tr_se_curvplan2m_s60
tr_se_curvplan2m_s7
tr_se_curvplan2m_std_10c
tr_se_curvplan2m_std_25c
tr_se_curvplan2m_std_50c
tr_se_curvplan2m_std_5c
tr_se_curvplan50m
tr_se_curvplan6m
tr_se_curvprof10m
tr_se_curvprof25m
tr_se_curvprof2m
tr_se_curvprof2m_grass_17c
tr_se_curvprof2m_grass_45c
tr_se_curvprof2m_grass_9c
tr_se_curvprof2m_s15
tr_se_curvprof2m_s30
tr_se_curvprof2m_s60
tr_se_curvprof2m_s7
tr_se_curvprof2m_std_10c
tr_se_curvprof2m_std_25c
tr_se_curvprof2m_std_50c
tr_se_curvprof2m_std_5c
tr_se_curvprof50m
tr_se_curvprof6m
tr_se_diss2m_10c
tr_se_diss2m_25c
tr_se_diss2m_50c
tr_se_diss2m_5c
tr_se_e_aspect10m
tr_se_e_aspect25m
tr_se_e_aspect2m
tr_se_e_aspect2m_10c
tr_se_e_aspect2m_25c
tr_se_e_aspect2m_50c
tr_se_e_aspect2m_5c
tr_se_e_aspect2m_grass_17c
tr_se_e_aspect2m_grass_45c
tr_se_e_aspect2m_grass_9c
tr_se_e_aspect50m
tr_se_e_aspect6m
tr_se_mrrtf2m
tr_se_mrvbf2m
tr_se_n_aspect10m
tr_se_n_aspect25m
tr_se_n_aspect2m
tr_se_n_aspect2m_10c
tr_se_n_aspect2m_25c
tr_se_n_aspect2m_50c
tr_se_n_aspect2m_5c
tr_se_n_aspect2m_grass_17c
tr_se_n_aspect2m_grass_45c
tr_se_n_aspect2m_grass_9c
tr_se_n_aspect50m
tr_se_n_aspect6m
tr_se_no2m_r500
tr_se_po2m_r500
tr_se_rough2m_10c
tr_se_rough2m_25c
tr_se_rough2m_50c
tr_se_rough2m_5c
tr_se_rough2m_rect3c
tr_se_sar2m
tr_se_sca2m
tr_se_slope10m
tr_se_slope25m
tr_se_slope2m
tr_se_slope2m_grass_17c
tr_se_slope2m_grass_45c
tr_se_slope2m_grass_9c
tr_se_slope2m_s15
tr_se_slope2m_s30
tr_se_slope2m_s60
tr_se_slope2m_s7
tr_se_slope2m_std_10c
tr_se_slope2m_std_25c
tr_se_slope2m_std_50c
tr_se_slope2m_std_5c
tr_se_slope50m
tr_se_slope6m
tr_se_toposcale2m_r3_r50_i10s
tr_se_tpi_2m_10c
tr_se_tpi_2m_25c
tr_se_tpi_2m_50c
tr_se_tpi_2m_5c
tr_se_tri2m_altern_3c
tr_se_tsc10_2m
tr_se_twi2m
tr_se_twi2m_s15
tr_se_twi2m_s30
tr_se_twi2m_s60
tr_se_twi2m_s7
tr_se_vrm2m
tr_se_vrm2m_r10c
tr_slope25_l2g
tr_terrtextur
tr_tpi2000c
tr_tpi5000c
tr_tpi500c
tr_tsc25_18
tr_tsc25_40
tr_twi2
tr_twi_normal
tr_vdcn25
Soil data
The soil data originates from various soil sampling campaigns since 1968. Most of the data was collected in conventional soil surveys in the 1970ties in the course of amelioration and farm land exchanges. As frequently observed in legacy soil data sampling site allocation followed a purposive sampling strategy identifying typical soils in an area in the course of polygon soil mapping.
dclass contains drainage classes of three levels.
Swiss soil classification differentiates stagnic (I), gleyic (G) and anoxic/reduced (R) soil profile qualifiers with each 4, 6 resp. 5 levels. To reduce complexity the qualifiers I, G and R were aggregated to the degree of hydromorphic
characteristic of a site with the ordered levels well (qualifier labels I1--I2, G1--G3, R1 and no hydromorphic qualifier), moderate well drained (I3--I4, G4) and poor drained (G5--G6, R2--R5).
waterlog indicates the presence or absence of waterlogging characteristics down 30, 50 and 100 cm soil depth. The responses were based on horizon qualifiers ‘gg’ or ‘r’ of the Swiss classification (Brunner et al. 1997) as those were considered to limit plant growth. A horizon was given the qualifier ‘gg’ if it was strongly gleyic predominantly oxidized (rich in \(Fe^{3+}\)) and ‘r’ if it was anoxic predominantly reduced (poor in \(Fe^{3+}\)).
pH was mostly sampled following genetic soil horizons. To ensure comparability between sites pH was transferred to fixed depth intervals of 0--10, 10--30, 30--50 and 50--100 cm by weighting soil horizons falling into a given interval. The data contains laboratory measurements that solved samples in \(CaCl_{2}\) or \(H_{2}0\). The latter were transferred to the level of \(CaCl_{2}\) measurements by univariate linear regression (label ptf in timeset). Further, the dataset contains estimates of pH in the field by an indicator solution (Hellige pH, label field in timeset).
The column timeset can be used to partly correct for the long sampling period and the different sampling methods.
Environmental covariates
The numerous covariates were assembled from the available spatial data in the case study area. Each covariate name was given a prefix:
cl_ climate covariates as precipitation, temperature, radiation
tr_ terrain attributes, covariates derived from digital elevation models
ge_ covariates from geological maps
sl_ covariates from an overview soil map
References to the used datasets can be found in Nussbaum et al. 2017b.
Brunner, J., Jaeggli, F., Nievergelt, J., and Peyer, K. (1997). Kartieren und Beurteilen von Landwirtschaftsboeden. FAL Schriftenreihe 24, Eidgenoessische Forschungsanstalt fuer Agraroekologie und Landbau, Zuerich-Reckenholz (FAL).
Nussbaum, M., Spiess, K., Baltensweiler, A., Grob, U., Keller, A., Greiner, L., Schaepman, M. E., and Papritz, A., 2017b. Evaluation of digital soil mapping approaches with large sets of environmental covariates, SOIL Discuss., https://www.soil-discuss.net/soil-2017-14/, in review.