Field Crops Research

Twenty-six CIMMYT bread (Triticum aestivum) and durum (Triticum turgidum) wheat genotypes, created between 1964 and 2009, were tested for yield and agronomic performance at CIMMYT’s experimental station near Ciudad Obregon, Mexico, over six years. Treatments included conventional and permanent raised beds with full and reduced irrigation. The objectives were to study breeding progress in distinct agronomic systems and to elucidate the importance of tillage and genotype by tillage interaction for yield and agronomic traits.

The first 5 columns in every sheet indicate the Experiment (year), Sowing date, Block or repetition (according to the experimental design), Radiation and Genotype. // "NILs" Sheet: Maturity group, Stewart coefficient and E allele composition of the genotypes used in this work. // "Ontogeny" Sheet: Date each ontogenic stage was attained, crop phase duration (expressed in calendar days and thermal days). // "PAR Photop Tº" Sheet: Photosynthetically Active Radiation (PAR) captured, mean incident PAR, mean air temperature, mean photoperiod and mean vapour pressure deficit during different crop phases. // "Biomass" Sheet: Aerial biomass (dry weight) for vegetative (leaves, petioles and stems) and reproductive (flowers, shells and seeds) organs. Partition coefficient. Seed number, Yield and individual seed weight. // "Numeric" Sheet: Number of nodes (in main stems and branches) and nodes with pods at maturity. Number of pods, pods per node and seeds per pod on primary and lateral racemes, total main stem, branches and total. // "Analisis" Sheet: For each node position (basal, central and apical) and raceme (primary and lateral): Number of opened flowers, final number of pods, Set, date first flower was opened, date dominant pod reached 2 cm, date dominant pod began seed filling, date last flower was opened, same stages expressed in thermal days after flowering (R1). Photoperiod explored when flowering began. Lag phase and flowering duration. // "Pod developmental rate" Sheet: Days after flowering when different flower/pod stages were first observed at each node position (basal, central and apical) and raceme (primary and lateral). For stage description please refer to Nico et. al (2016, JXB 67(1):365–377).

On-farm Performance and Farmers’ Participatory Assessment of New Stress-Tolerant Maize Hybrids in Eastern Africa Two sets of varieties, one with 12 early-to-intermediate maturing (EI) hybrids and the other with 13 intermediate-to-late maturing (IL) hybrids, were evaluated on-farm under smallholder farmers’ conditions. The varieties included pre-commercial CIMMYT test hybrids, internal genetic gain checks, popular commercial hybrid checks, and one farmer’s check variety. The on-farm trials were conducted in 2016 and 2017 in Kenya, Uganda, Tanzania, and Rwanda under rainfed conditions and farmers’ management, except at KALRO-Kiboko station where supplementary irrigation was used. The EI varieties were evaluated at a total of 60 environments (site-year-management combinations), the IL varieties were evaluated at 54 environments; grain-yield data were collected successfully at 42 and 40 environments, respectively. For each entry, field weight was recorded from all ears in the four central rows (15 m2), and grain yield (t ha-1) was estimated. To evaluate new maize hybrid varieties, farmers were invited to observe them in the field, at mid-season and at harvest. Participants were shown one replication of the trial, containing one variety per plot. The evaluation was double blind, and the order in which plots were evaluated were randomized. The tool a questionnaire with three components: socioeconomic characteristics, plant traits, and overall performance evaluation. The first component included socioeconomic characteristics of the respondent and maize production characteristics. The second component was a table with a list of traits. For each trait, participants quantified its importance on a scale of 0 (not important) to 3 (very important). The third component was a continuation of the same table, with one column for each variety, identified by the plot number in the column head, and a row for each trait. The participant was asked to score each variety on a five-point Likert scale, where A = like strongly, B = like, C = neither like nor dislike, D = dislike, E = dislike strongly. In the final row, participants scored the variety for overall performance. A total of 2,025 farmers took part in the evaluations. The data of these trials were assembled in five data files (Table 1). Since the data cover different years and sites, where site numbers may have been repeated and randomized across the districts, key variables are to be considered together in the identification of unique records, as explained in the table below. # File Key variables 1 Respondent characteristics END_SERIAL_NO and SITE_NUMBER 2 Criteria for importance END_SERIAL_NO and SITE_NUMBER 3 Participatory evaluation END_SERIAL_NO, SITE_NUMBER, CATEGORY, YEAR and VARIETY 4 Plot level data yield and mean pve trait scores SITE_NUMBER, CATEGORY, YEAR and VARIETY 5 Regional On-Farm Trials (roft) early-intermed site and randomization SITE_NUMBER, CATEGORY, YEAR Plot and VARIETY

Data for model calibration and evaluation

The dataset contains the yield predictions from different statistical models in the paper. 1. The leave_one_year_out test from 2003 to 2016 (Figs. 2,3,5,6,7): (1) (Tgs + Pgs)linear model: prediction_Tgs_linear_rainfed_leave_one_year_out.csv (2) (T + P)linear model: prediction_tave_linear_rainfed_leave_one_year_out.csv (3) (VPD + P)linear model: prediction_vpd_linear_rainfed_leave_one_year_out.csv (4) (VPD + P)poly model: prediction_vpd_poly_rainfed_leave_one_year_out.csv (5) (VPD + P)spline model: prediction_vpd_spline_rainfed_leave_one_year_out.csv (6) LSTspline model: prediction_lstmax_spline_only_rainfed_leave_one_year_out.csv (7) EVIlinear model: prediction_evi_linear_only_rainfed_leave_one_year_out.csv (8) EVIpoly model: prediction_evi_poly_only_rainfed_leave_one_year_out.csv (9) LSTspline + EVIpoly model: prediction_lstmax_spline_evi_poly_only_rainfed_leave_one_year_out.csv (10) (VPD + P)spline + EVIlinear model: prediction_vpd_spline_evi_rainfed_leave_one_year_out.csv (11) (T + P)spline + EVIpoly model: prediction_tave_spline_evi_poly_rainfed_leave_one_year_out.cs (12) (LST + P)spline + EVIpoly model: prediction_lstmax_spline_evi_poly_rainfed_leave_one_year_out.csv (13) (VPD + P)spline + EVIpoly model: prediction_vpd_spline_evi_poly_rainfed_leave_one_year_out.csv (14) (VPD + P)poly + EVIpoly model: prediction_vpd_poly_evi_poly_rainfed_leave_one_year_out.csv 2. The leave_one_year_out test for state-specific models from 2003 to 2016 (Fig. 4): (15) (VPD + P)poly model: prediction_vpd_poly_evi_poly_rainfed_leave_one_year_out_state.csv 3. The leave_one_year_out test from 1981 to 2016 (Fig. S4): (16) (VPD + P)spline model: prediction_vpd_spline_allyears_rainfed_leave_one_year_out.csv 4. The forward test from 2005 to 2016 (Fig. 8): (17) (VPD + P)spline model: prediction_vpd_spline_rainfed_forward.csv (18) (VPD + P)spline + EVIpoly model: prediction_vpd_spline_evi_poly_rainfed_forward.csv 5. The leave_one_year_out test for irrigated yield and comebined yieldfrom 2003 to 2016 (Fig. S6): (19) (VPD + P)poly + EVIpoly model for irrigated yield: prediction_vpd_poly_evi_poly_irrigated_leave_one_year_out.csv (20) (VPD + P)poly + EVIpoly model for combined yield: prediction_vpd_poly_evi_poly_all_leave_one_year_out.csv

The information of yield, ET, soil water recharge during through fallow, and WUE was included in this file. It is the data foundation of Figs.3 and 4 and Tables 3-5 in the manuscript.

Temporal monthly temperature/precipitation distribution from 2007 to 2017 at the Heyang Dryland Agriculture Research Station. It is the data foundation of Figs. 2 and 3 and Table 2 in the submitted manuscript.

An Excel spreadsheet containing plot means for treatments applied in sorghum agronomy trials conducted across the Queensland production region

Excel file contains the 3 supplemental tables in the manuscript. Table S1 contains individual tabs for each of the eight study sites (S1 to S8). Tables S2-S3 are individual tabs. Supplementary Table S1. 10 m resolution QA/QC data for each study site including: location, elevation, hydrogeophyscial surveys of EMI and neutron intensity, Landsat GCVI by year, crop yield by year, and First EOF of each covariate. Supplementary Table S2. Summary of MLR and RF fitting coefficients and statistical metrics by site and crop type. Supplementary Table S3. Summary of MLR and RF fitting coefficients and statistical metrics by site, crop type, and year. Supplementary R Code and results for sites S1 to S8 by crop.

CIMMYT’s durum wheat breeding program performed parallel selection in conventional tilled (CT) and ZT soils with the aim to compare the effect of selection under either CT or ZT on the performance of selected progenies. From 16 initial crosses, 234 lines were selected under CT and 250 under ZT. All 484 lines were subsequently tested for yield and growth traits during three seasons (winter 2012-2013, 2013-2014 and 2014-2015) near Ciudad Obregon, Sonora, Mexico in three different testing environments. Those included ZT and CT with full irrigation and CT with reduced irrigation. The experiment was set-up as an alpha lattice design with three replications for each testing environment. Within each replication, genotypes were arranged randomly in three blocks of 160 and 170 genotypes. The dataset includes the following data: days to heading (DH), plant height (PHT), grain yield (GY) and two NDVI values (NDVI1 and NDVI2). Throughout the experiment, NDVI readings were recorded at regular intervals and growth curves were created based on the obtained data. For analysis, two values were selected, one measurement during early vegetative growth (NDVI1), around four weeks after planting, and the second at maximum growth (NDVI2).