Seasonal Changes in Reflectance
Reflectance patterns throughout the growing season reveal a large amount of information about the changes in the visible and near-infrared (NIR) wavelengths. Understanding these patterns is critical to the interpretation of the seasonal patterns in VIs. These wavelengths are used in various VIs; however, it is important to examine their changes throughout the year and especially during the growing season. There are general trends in changes within a crop and also differences among crops. There are similar features in terms of the general response of the blue, red, and green wavelengths during the growing season and the near-infrared increasing and the following reflectance patterns are typical of the different crops throughout each year. Throughout the complete year from harvest of the preceding crop until after harvest of the next crop there were changes over the fall and winter period; however, these affected the bare soil reflectance values. An example of the changes in the reflectance values for corn throughout the growing season is shown in
Figure 1. There is a rapid increase in the NIR values as soon as the crop begins to develop while the changes in the visible wavebands change more slowly and throughout the growing season the NIR wavelengths are more dynamic than the visible wavelengths (
Figure 1). This is due in part the reflectance values for bare soil being closer to the visible reflectances than NIR values (
Figure 1). The bare soil line remained fairly constant across wavelengths throughout the non-growing season period and changes from day to day were affected more by soil water content and disturbance of the soil surface with tillage which in turn affects the distribution of the crop residue or the roughness of the soil surface (
Figure 2). An interesting feature in the corn reflectance values during the growing season is the increase the NIR values when tassels emerge at the upper canopy. Changes in the major phenological stages and agronomic practices for the corn crop are shown in
Figure 3. The sharp change in the visible and NIR patterns during the growing season create the changes that make the VIs a valuable assessment tool for detection of crop changes,
A cropping sequence of wheat followed by soybean within the same growing season reveals the dynamics of the vegetation development of these two crops (
Figure 4). A unique feature in these patterns was found with a increase in reflectance observed in the seasonal trends from Day of Year (DOY) 180 until 200 created by the presence of weeds within the soybean crop that was removed through herbicide application. The reflectance values were similar to wheat and soybean in terms of the visible and NIR wavelengths but occurred very quickly in development and then disappeared after the weeds were removed. These patterns were similar to corn with very stable red and green reflectance compared to the change in NIR wavelengths. There was a noticeable and significant decline in the NIR reflectance over the wheat canopy following the presence of the panicles at the top of the canopy. In wheat, these morphological structures reduce the NIR reflectance values compared to the leaves of the canopy (
Figure 4). As the soybean crop began to develop the same pattern in reflectance change appeared with reflectances in the visible wavelengths decreasing and the NIR increasing (
Figure 4).
Figure 1.
Reflectance over corn obtained throughout the growing season with a CropScan eight-band radiometer over a spring strip tillage system in 2007.
Figure 1.
Reflectance over corn obtained throughout the growing season with a CropScan eight-band radiometer over a spring strip tillage system in 2007.
Figure 2.
Reflectance over corn obtained throughout the year with a CropScan eight-band radiometer over a spring strip tillage system in 2007.
Figure 2.
Reflectance over corn obtained throughout the year with a CropScan eight-band radiometer over a spring strip tillage system in 2007.
Figure 3.
Sequence of agronomic practices and phenological [
24]stages of the corn crop for the 2007 cropping season.
Figure 3.
Sequence of agronomic practices and phenological [
24]stages of the corn crop for the 2007 cropping season.
Figure 4.
Reflectance over wheat (until DOY 180) and soybean (DOY 180 and greater) obtained throughout the 2006–2007 year with a CropScan eight-band radiometer.
Figure 4.
Reflectance over wheat (until DOY 180) and soybean (DOY 180 and greater) obtained throughout the 2006–2007 year with a CropScan eight-band radiometer.
Observations over the canola canopy had a different temporal pattern because of the early season development and maturity and harvest in mid-summer (
Figure 5). Patterns of reflectance are similar to the other crops observed in the study with the largest temporal change in the NIR wavelengths; however, the NIR values showed a broader peak throughout the season compared to other crops (
Figure 5). Reflectance values in the red region were often less than 0.05 during the growing season while NIR values ranged from 0.2 to 0.4 (
Figure 5). In these four canopies the changes in the reflectances across the wavelengths show the dynamic nature of the reflectance values observed over agronomic crops and indicate that for accurate development of VIs there is a need for frequent observations during the growing season while during the off-season there is little change. The patterns among years were very similar; however, the rapid changes within the season demonstrate the need for frequent measurements to capture the dynamics of cropping systems within a field. Observations of canopies within fields for precision agriculture applications will have to recognize these temporal patterns in reflectance. Changes in the visible and near-infrared wavelengths provide information about the canopy level response that forms the foundation for assessing differences in crop growth or response to different practices applied within a field.
Figure 5.
Reflectance over canola during the 2007 growing season obtained with a CropScan eight-band radiometer.
Figure 5.
Reflectance over canola during the 2007 growing season obtained with a CropScan eight-band radiometer.
Seasonal changes in VIs
Seasonal changes in the VIs showed patterns indicative of the seasonal patterns in the wavebands as shown in
Figure 1,
Figure 2,
Figure 4, and
Figure 5. Values for NDVI showed a bare soil value near 0.15 and increased to 0.9 during the period of maximum vegetative growth with this value maintained for a long period until the crop began to senesce (
Figure 6). The values of NDVI in these studies saturated when the LAI approached a value of 4 and then remained high until midway through the reproductive phase of development. The seasonal patterns of NDVI reveal the long period of the season when the values are above 0.9 even though leaf area is changing in the canopies. This pattern is evident in NDVI data collected over corn, canola, soybean, or wheat canopies over the course of this study. This pattern for NDVI was found for all of the seasons we examined in this study and the results are very consistent among years. To evaluate NDVI patterns over the years we conducted an ANOVA on the complete set of observations across the tillage systems for the corn and soybean crops. This was conducted for seven years of observations at four different periods during the growing season with the results shown in
Table 2. There was a significant effect of years on NDVI at the mid-vegetative stages (approximately V6 to V7 in corn,
Figure 3), while crop and tillage was significant only at the mid-vegetative stage. The reason for this is very clear given the patterns shown in
Figure 6 where the variation is most evident in the early stages and then little variation after the NDVI values saturate in the end of the season. This is similar to tillage systems which were significantly different at planting and at mid-vegetative stages because of the differences in rates of plant development.
Figure 6.
Values for NDVI and standard deviation for three tillage systems for corn grown in 2007.
Figure 6.
Values for NDVI and standard deviation for three tillage systems for corn grown in 2007.
Table 2.
Mean values of NDVI observations over corn and soybean crops grown in four tillage systems for the period from 2002 to 2006 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Table 2.
Mean values of NDVI observations over corn and soybean crops grown in four tillage systems for the period from 2002 to 2006 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Factor | Degrees of Freedom | Development Stage |
---|
Planting | Mid-Vegetative | Reproductive Onset | Mid-grain Fill |
---|
Year | 4 | 0.12ns1 | 0.80* | 0.94ns | 0.85ns |
Crop | 1 | 0.12ns | 0.81* | 0.93ns | 0.85ns |
Tillage | 3 | 0.13* | 0.83* | 0.93ns | 0.84ns |
Year × Crop | 4 | 0.12ns | 0.81ns | 0.94ns | 0.85ns |
Year × Tillage | 3 | 0.12ns | 0.82ns | 0.94ns | 0.84ns |
Crop × Tillage | 3 | 0.12* | 0.83* | 0.93** | 0.86** |
Year × Crop × Tillage | 12 | 0.13ns | 0.82ns | 0.93ns | 0.85ns |
The tillage term was significant because of the effect of the different tillage systems on the residue remaining on the soil surface. The only interaction term that was significant was the crop × tillage interaction caused by the effect of the tillage system on crop response that was detected in the NDVI values (
Table 2). There has been extensive use of the NDVI approach to estimate yield potential in corn [
19] and our results would support this approach since the most sensitive period to detect differences is in the vegetative stage.
The analysis was extended to examine the differences among corn hybrids grown with a fall strip
versus a spring strip tillage system for NDVI values for the same four development stages (
Table 3). There was no effect of years and hybrids were significantly different at all times except at the planting observation. Tillage was significant at only the planting and mid-vegetative stage because of the effect of the tillage practice on the crop residue that was remaining on the soil surface. The only other significant effect was the hybrid x tillage interaction for the mid-grain fill period. This was caused by the tillage systems interacting with hybrids in terms of the senescence of the crop (
Table 3).
Table 3.
Mean values of NDVI observations over six corn hybrids grown in two tillage systems for the period from 2002 to 2004 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Table 3.
Mean values of NDVI observations over six corn hybrids grown in two tillage systems for the period from 2002 to 2004 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Factor | Degrees of Freedom | Development Stage |
---|
Planting | Mid-Vegetative | Reproductive Onset | Mid-grain Fill |
---|
Year | 2 | 0.15ns1 | 0.84ns | 0.92ns | 0.84ns |
Hybrid | 5 | 0.14ns | 0.83* | 0.93ns | 0.85* |
Tillage | 1 | 0.13* | 0.84* | 0.92ns | 0.84ns |
Year × Hybrid | 10 | 0.14ns | 0.84ns | 0.92ns | 0.83ns |
Year × Tillage | 2 | 0.15ns | 0.84ns | 0.93ns | 0.84ns |
Hybrid × Tillage | 5 | 0.14ns | 0.83ns | 0.91ns | 0.85* |
Year × Hybrid × Tillage | 10 | 0.14ns | 0.85ns | 0.93ns | 0.84ns |
One of the applications of NDVI is the estimation of intercepted Photosynthetically Active Radiation (iPAR) as shown by [
6]. In both corn and soybean canopies we were able to use iPAR derived from the NDVI values to predict the seasonal changes in the canopies in intercepted PAR (
Figure 7 and
Figure 8).
Figure 7.
Cumulative intercepted solar radiation for corn in two tillage systems, fall strip and spring strip with two different N treatments in 2007.
Figure 7.
Cumulative intercepted solar radiation for corn in two tillage systems, fall strip and spring strip with two different N treatments in 2007.
Figure 8.
Cumulative intercepted solar radiation for two varieties of soybean in three tillage systems, nu-till, fall strip and no-till in 2006.
Figure 8.
Cumulative intercepted solar radiation for two varieties of soybean in three tillage systems, nu-till, fall strip and no-till in 2006.
For the corn canopies there was no significant difference in the seasonal NDVI values for the growing season in these systems (
Figure 7) as would be expected from the lack of significance among tillage systems (
Table 2). For the soybean canopies there were significant differences in the cumulative iPAR interception values with the no-till practice showing the lowest cumulative value and also the lowest grain yield (
Figure 8). The use of the iPAR values derived from NDVI provides a stable measure of crop growth among years and can be easily derived from NDVI values and consistently shows differences among agronomic practices related to crop yield. Utilization of the iPAR values derived from NDVI offers a valuable method for assessing differences among different systems.
Temporal patterns of other VIs showed different seasonal trends than did NDVI. This is to be expected since each of the VIs has their own unique combination of wavebands that have been related to a specific canopy parameter. It is not possible to show all of the different VIs that are available for crop assessment. However, the concern about improved N management based on chlorophyll content of leaves draws attention to indices with the ability to estimate chlorophyll content. One of those indices is the NPCI and patterns for the NPCI index showed a slight but significant change in the early season and then an increase as the crop matured when leave color began to change. This seasonal pattern is interesting since the NPCI increased when the canopy was in the senescence phase and losing chlorophyll and the NPCI was more sensitive to changes in chlorophyll in the later season when N decisions are not longer needed as compared to the early season when N rates could be adjusted. These difficulties in using independent measures of VIs for N management is the reason why high N rates strips are placed within fields as a direct comparison within the field between a high rate and the area to be managed. Thus, this index has the ability to detect leaf chlorophyll differences in canopies. Ratios of NIR/green or NIR/red edge have been proposed as chlorophyll indices [
25]. An ANOVA analysis for the corn hybrids for the two tillage systems showed a difference in the development periods in which there were significant differences (
Table 4).
Table 4.
Mean values of Chlorophyll Index (CI) observations over six corn hybrids grown in two tillage systems for the period from 2002 to 2004 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Table 4.
Mean values of Chlorophyll Index (CI) observations over six corn hybrids grown in two tillage systems for the period from 2002 to 2004 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Factor | Degrees of Freedom | Development Stage |
---|
Planting | Mid-Vegetative | Reproductive Onset | Mid-grain Fill |
---|
Year | 2 | 0.80ns1 | 6.5ns | 9.5* | 7.7* |
Hybrid | 5 | 0.79ns | 6.7* | 9.6* | 7.6** |
Tillage | 1 | 0.82ns | 6.3* | 9.7* | 7.5* |
Year × Hybrid | 10 | 0.83ns | 6.4ns | 9.5ns | 7.6ns |
Year × Tillage | 2 | 0.78ns | 6.6ns | 9.5ns | 7.7ns |
Hybrid × Tillage | 5 | 0.79ns | 6.5* | 9.8** | 7.6** |
Year × Hybrid × Tillage | 10 | 0.80ns | 6.6ns | 9.5ns | 7.6ns |
There significant differences among the years for the reproductive stages because the years were different in their growth patterns that were evident during these growth stages. Hybrids and tillage systems were significantly different at all stages (
Table 4). The hybrid × tillage term was also significant across the vegetative, reproductive onset, and mid-grain filling periods which indicates that hybrids reacted differently to the nitrogen associated with a particular tillage system and the CI is able to detect those responses. The CI is effective at assessing nitrogen status in corn leaves. There were no significant differences in any of the factors for the CI when this analysis was applied to the soybean varieties. This is expected since soybean canopies obtain their nitrogen from fixation. The use of simple ratios of wavebands early in the growing season will create indices with a large amount of variation. If an index like CI is used to estimate the shortage of N in a crop canopy then these results will have to be used with caution. Simple ratio indices are responding to changes in either one of the wavebands and early in the season are affected by the soil background. The transformation of this ratio as suggested by [
12] did not affect the standard deviation and we did not find this degree of variation in the individual leaf chlorophyll readings collected in these treatments; however, the leaf chlorophyll readings are collected on individual leaves and don’t have the spatial variation of canopies. When the NPCI was compared for soybean canopies there was no significant difference and a minimum standard deviation among the four tillage systems. This is expected since there are no expected differences in chlorophyll content of the soybean canopies.
The temporal pattern of VIs depends upon the phenological stage or plant parameters to which the index is most closely related. An example of this change throughout the season related to a different sequence is the PSRI which is related to the senescence of the canopies [
13]. We evaluated the PSRI for the six corn hybrids grown in two tillage systems using an ANOVA on the different effects similar to the analysis for the other VIs. There were no significant differences for the PSRI until the onset of the reproductive period and mid-grain fill. The differences at the onset of the reproductive stage can be attributed to the change in leaf color induced by late season N stress which created a change in the reflectance in the canopies. This was found to be significant in the hybrid × tillage interaction term (
Table 5).
Table 5.
Mean values of PSRI observations over six corn hybrids grown in two tillage systems for the period from 2002–2004 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Table 5.
Mean values of PSRI observations over six corn hybrids grown in two tillage systems for the period from 2002–2004 for four development periods and the ANOVA for the comparison of the treatments and interactions.
Factor | Degrees of Freedom | Development Stage |
---|
Planting | Mid-Vegetative | Reproductive Onset | Mid-grain Fill |
---|
Year | 2 | 0.22ns1 | 0.02ns | 0.01* | 0.32* |
Hybrid | 5 | 0.23ns | 0.02ns | 0.03* | 0.33** |
Tillage | 1 | 0.21ns | 0.02ns | 0.02ns | 0.32ns |
Year x Hybrid | 10 | 0.22ns | 0.01ns | 0.02ns | 0.35** |
Year x Tillage | 2 | 0.22ns | 0.02ns | 0.02ns | 0.33ns |
Hybrid x Tillage | 5 | 0.23ns | 0.02ns | 0.03* | 0.35** |
Year x Hybrid x Tillage | 10 | 0.23ns | 0.02ns | 0.02ns | 0.33ns |
This was similar to the observations from the soybean canopies in which there were significant differences among tillage systems only after DOY 250 (data not shown) in which there was beginning leaf drop and yellowing of the leaves. The utility of this index is to provide a VI to detect differences in plant senescence across fields. This index is most sensitive to the senescence phase in being able to differentiate among treatments. This index pattern can be used to determine the rate of senescence in different crops by comparing the rate of change from a green canopy to a senesced canopy. Combination of the different indices throughout the growing season can be used to define the points at which the indices can be most useful in their comparison.
Other plant parameters that can be derived from remote sensing data are LAI and plant biomass. Leaf area index was derived from the EVI as proposed by [
13] and for the corn canopies there was a significant difference among the treatments (
Figure 9).
Figure 9.
Values for LAI derived from an Enhanced Vegetative Index (EVI) and the standard deviation over the treatment means throughout the growing season for corn in 2007 for combinations of tillage and nitrogen fertilizer systems.
Figure 9.
Values for LAI derived from an Enhanced Vegetative Index (EVI) and the standard deviation over the treatment means throughout the growing season for corn in 2007 for combinations of tillage and nitrogen fertilizer systems.
The ability to estimate LAI provides a useful tool for crop assessment and the variation seen in this figure is not different than direct observations collected with either destructive or non-destructive leaf area meters. In this analysis, we were able to detect differences in LAI induced by the increase in N addition to the plants (191
vs. 135 kg ha
−1N) which were detectable with the EVI values and significantly different (
Figure 9). These patterns are typical of the different cropping seasons demonstrating the potential utility of this type of approach. Biomass patterns showed even a larger difference during the growing season and the differences shown in
Figure 10 demonstrate the ability to detect treatment differences in green biomass for soybean. It is encouraging that we can detect the differences in green biomass throughout the growing season and the comparison between these values derived from the remote sensing, using the NIR/Red ratio, and the destructive sampling were correlated with a value of r
2 = 0.93. There were differences in biomass throughout the season that were induced by the different management practices and these were detectable using the VI (
Figure 10).
Figure 10.
Values for above ground biomass derived from the nir/red ratio and the standard deviation over the treatment means throughout the growing season for soybean in 2007 for combinations of tillage and folair fertilizer application.
Figure 10.
Values for above ground biomass derived from the nir/red ratio and the standard deviation over the treatment means throughout the growing season for soybean in 2007 for combinations of tillage and folair fertilizer application.
Similar observations were found in corn canopies in which we could detect significant differences among treatments that affected green biomass accumulation. Plant parameters can be detected with remote sensing with sufficient accuracy to provide an indication of the growth patterns of crops. The seasonal variation does provide a measure of the variation which exists within treatments that is often missed by destructive sampling because of the limitations imposed by the time and human resources required for intensive sampling. The use of remote sensing methods allows for more frequent and less costly methods for obtaining LAI and biomass values across more treatments and within fields.