FieldDino: High-throughput physio-morphological phenotyping of stomatal characteristics for plant breeding research

Key results

1. 01

   ShownThe combined porometer and leaf-clip microscope workflow scaled to a replicated field trial. It collected paired physiology and anatomy data much faster than screening based on gas exchange.

   2,400 coupled measurements on 200 wheat genotypes over 5 days (09:30 to 15:00); < 15 s per measurement, about 1 min including leaf selection and plot moves; over 60 plants per hour vs less than 10 for IRGAs; microscope capture < 10 s per leaf at under AU $2,000 per unit. · 3 Results; 4 Discussion

2. 02

   ShownA COCO-pretrained YOLOv8 medium segmentation model, chosen from five model sizes, accurately outlined whole stomatal complexes in field microscope tiles.

   Fine-tuned on 518 images covering all 200 genotypes, tiled into 14,928 sub-images (13,068 train, 1,236 validation, 624 test); YOLOv8-M reached mask mAP@0.5 of 97.1% and mAP@[0.5:0.95] of 65.9%; the N, S, L and X variants were compared on precision, recall, F1, FLOPs and parameter count. · Image annotation; Model training and validation; Tables

3. 03

   ShownGuard cell length predicts pore length closely enough to derive anatomical maximum conductance (gsmax) from segmented stomata. The authors report that the resulting gsmax values align closely with previously published values.

   ImageJ measurements of 40 stomata on each of 50 adaxial and 50 abaxial images gave R^2=0.98, with pore length = 0.543 x guard cell length; gsmax spanned 0.58 to 2.22 mol m^-2 s^-1 (adaxial tending higher), slightly wider than the 0.7-2.0 mol m^-2 s^-1 range cited from prior studies. · Guard cell – pore length relationship &…; 3.2 Stomatal Anatomy; 4 Discussion

4. 04

   SuggestedThe automated anatomy measurements reproduced expected patterns: higher stomatal density on the adaxial than the abaxial surface, denser but smaller stomata under rainfed conditions, and a negative relationship between stomatal size and density.

   Mean density 50.31 (adaxial) vs 42.87 (abaxial) per mm^2, and the Discussion separately cites a 31.5% higher adaxial density; guard cell length 68.39/68.55 µm irrigated vs 64.30/66.04 µm rainfed (abaxial/adaxial); overall means 66.8 µm length, 40.2 µm width, 46.6 per mm^2; no statistical tests were run. · Analyses and visualisation of data; 3.2 Stomatal Anatomy; 4 Discussion

5. 05

   SuggestedPairing operating conductance with gsmax showed that stomatal operating efficiency varies with genotype and water supply. Efficiency fell under rainfed conditions, together with lower conductance and lower ΦPSII (quantum yield of photosystem II).

   Rainfed conductance was 24.4% (abaxial) and 28.2% (adaxial) below irrigated, vs 26% - 50% drought reductions reported previously; median adaxial gsop/gsmax across genotypes 0.03 to 0.25 irrigated vs 0.01 to 0.17 rainfed; median abaxial ΦPSII 0.61 irrigated vs 0.46 rainfed. · 3.1 Stomatal conductance and chlorophyll…; 3.3 Combined stomatal physiology and anatomy…; 4 Discussion